ORIENTAL LAND CO., LTD. (4661): The Next Yen of Revenue Destroyed Profit, and the Famous Risk Ranked Eighth

Verdicts

Lens	Verdict	Buy below	Most load-bearing items
Buffett	watch	¥840	B99/B2; B42/B45; B33/B86
Munger	watch	¥1,500	M23/M85; M44/M88; M54/M83
Pabrai	pass	¥450	P60/P53; P62; P1
Li Lu	too-hard	— (no threshold issued; the floor arithmetic is published at ~¥294)	L34; L1; L13
Claude	pass	implied ¥700, band ¥410–1,163	C13; C35; C14

Two watch, two pass, one too-hard. Four named prices run from ¥450 to ¥1,500 against a stamp of ¥2,920 F2485 — discounts of 49% to 85%, and a 3.33× spread between the highest and lowest threshold on one ledger. No verdict class reaches four, so no consensus red team fired; the reason is set out in ## Red team rather than papered over.

The framing correction — read this before anything else

The obvious story here is a rented moat with a printed expiry date, and this study documents it exhaustively. The Claude lens then sized it, and it ranks roughly eighth.

The licensor's contractual claim is visible exactly once in 142 pages, as a parent-only cost line: royalty ¥42,048百万円 F2131, against ¥40,999百万円 the prior year [E263]. Set beside consolidated operating profit of ¥168,413百万円 F1019, that is 24.97% of consolidated operating profit — a variable cost with a repricing date, not a call option on the enterprise. Run the modal branch: a 2051 reset that moves the licensor from about a quarter to about a third of operating profit costs roughly ¥16.8bn pre-tax at today's scale, ~¥12.0bn after the observed 28.37% tax rate D62, arriving in twenty-five years; capitalised at 7% and discounted twenty-five years that is ¥31,674百万円 of present value — 0.7% of the ¥4,787,975百万円 market capitalisation D6. Weighted across all branches, including a 6% probability of outright loss, the whole terminal-value discontinuity is worth roughly a tenth of the equity.

The Claude lens registered P ≈ 0.75 on the annuity treatment being closer to correct than the perpetuity treatment, figures-blind, before the ledger opened. It scores the direction right and the magnitude badly over-weighted, and records the finding in its own words: the licence unknowns rank "eighth, not first… and my §1 gave them the weight of a verdict."

What actually decides the study is duller and much larger. Revenue rose ¥25,165百万円 — 679,374 F4 to 704,539 F5 — and operating profit fell ¥3,698百万円, 172,111 F1018 to 168,413 F1019. That is an incremental operating margin of −14.7% against an average operating margin of 23.90% D28. Widen past the reopening and two fully-recovered years give 4.2% incremental ordinary margin — 169,641 F10 less 166,005 F8, over 704,539 F5 less 618,493 F3 — against a 24.08% average D34. Gross profit fell in absolute yen on that extra revenue, ¥273,193百万円 F1014 to ¥272,959百万円 F1015. ¥43,098百万円 of added net operating assets — 727,462 D84 to 770,560 D83 — produced less NOPAT: ¥120,629百万円 D63 against roughly ¥123,279百万円 the year before, a prior-year figure derived in profiles/claude.md C14 rather than carried as a ledger row — the ledger has no prior-year NOPAT. Five years of operating cash of ¥796,674百万円 D102 converted into ¥66,763百万円 of free cash flow, 8.4% D104. Committed, contracted, unpaid capital of ¥332,929百万円 D111 now exceeds net cash of ¥270,155百万円 D71, so deployable cash is −¥62,774百万円. At the stamp the reverse-DCF implies 4.69% perpetual growth in owner earnings D6.

The licence still belongs in this thesis, and it is set out in full below — nine agreements terminating 2051-09-03 [E260], the Celebration Hotel ending 2038-08-19 with no extension clause stated [E126][E261], a cruise licence whose extension right is the Company's alone [E129], a rate disclosed only as 「一定料率」 [E130], and its total absence from all thirteen enumerated risks [E264]. But it belongs there as a bounded factor whose size is now known, not as the verdict. The methodological lesson of this study is that a vivid, well-documented, datable risk out-competed a dull arithmetic one for the outside view's attention, and the dull one was worth more.

The business

株式会社オリエンタルランド operates Tokyo Disneyland and Tokyo DisneySea, six Disney-branded hotels, a monorail and a shopping complex [E34][E35], together with an amphitheatre opened in September 2012 [E25], on one reclaimed site in Urayasu, Chiba Prefecture. Twenty-seven and a half million people walked through the two park gates in the year just reported — 27,534千人, 99.9% of the prior year F372F373 — and spent, inside the parks alone, ¥20,642 each D54.

The land is genuinely owned, and it was made rather than bought. The Company reclaimed it from Tokyo Bay under an agreement with Chiba Prefecture, the works running from 1964 to 1975 [E11]. Tokyo Disneyland's parcel is 837,259㎡ F449 and Tokyo DisneySea's 718,467㎡ F457, both held freehold [E134]. Tokyo Disneyland opened in April 1983 [E14]. Management names exactly two sources of 「独自の競争優位性」 — this land near central Tokyo, and the Disney licence [E42].

The revenue is admissions, merchandise, food and rooms, and it is collected in advance or on the spot. Inside the Theme Park segment, attraction-and-show revenue was ¥289,547百万円 F1749, merchandise ¥167,315百万円 F1753 and food and beverage ¥98,367百万円 F1757. Externally, Theme Park billed ¥568,345百万円 F1772, Hotel ¥119,049百万円 F1773 and Other ¥17,144百万円 F1775 — 80.67% D47, 16.90% D48 and 2.43% D49 of consolidated revenue. Consideration is received within about a month of the performance obligation being satisfied with no significant financing component [E217], and advance park-ticket money sits on the balance sheet as ¥36,901百万円 of contract liabilities F967, down from ¥38,228百万円 F966. Thirteen consolidated subsidiaries F230 and three affiliates F231 sit under the parent, with 11,207 employees F83 plus an average 17,651 temporary staff recorded 外書, outside that count F88, and a 24,498-member union [E211].

And the reason guests come is rented. The parks, the hotels, the monorail and the coming ship all run under ten agreements with Disney Enterprises, Inc., the first dated April 1979 [E259]. Nine of the ten terminate on one day — 3 September 2051 — despite start dates spread across thirty-nine years, and each carries the sentence 「各当事者はさらに５年間ずつ、５回にわたり延長することができる」 [E260]. Whether 「各当事者は」 means either party may or the parties may is not resolved anywhere, and the same table proves the drafter distinguishes when he wants to: the cruise contract says 「当社は」, the Company alone [E129][E262]. Two contracts sit outside the pattern — the Celebration Hotel runs to 2038-08-19 with no extension clause stated at all, and the cruise runs thirty years from an unfixed sailing date [E261].

The rent is one sentence. The Company pays a royalty 「一定料率にしたがって」 — in accordance with a fixed rate [E130]. No rate, no base, no minimum, no cap, no escalator, and nothing about what happens to any of them on renewal [E263]. Nothing is disclosed about termination, default, cure, change of control, assignment, governing law or dispute resolution for any of the ten [E265]; nothing about exclusivity, territory, the licensed characters themselves, or any non-compete — 「独占」 and 「排他」 appear nowhere [E266]; no Disney approval right over content, pricing, capital investment or itinerary [E267]; and nothing whatever about what becomes of the parks, the marks or ¥837,247百万円 of property F937 when the term runs out [E268]. 「ライセンス」 occurs in three places in the whole filing [E269]. And the licence appears in none of the thirteen enumerated 事業等のリスク [E264][E86] — a risk section that is otherwise above average, nine strategic and four operational F310, with demographic decline [E89], extreme heat possibly ending current summer operation [E95], single-business dependence [E96], cruise-launch delay [E97] and Maihama concentration [E99] all named.

The growth commitment is a ship. ¥330,000百万円 is budgeted for the cruise business F519F522, targeted at 2028 F525, of which ¥305,718百万円 remains unpaid D110; a subsidiary was established in April 2026 F229[E239]. Management states the cruise will be more profitable than the parks [E51]. The filing discloses zero research and development — 該当事項はありません [E132] — beside a basic policy on control stating the Company will advance R&D in new-growth businesses [E270]. Separately, ¥70,500百万円 is committed to completely rebuilding an existing attraction and its surrounding area F517 — a project the filing's own capex note names among the year's Tokyo Disneyland drivers beside 更新改良, renewals and improvements [E112], while the Company's stated policy allocates cash preferentially to growth investment [E55]. Nothing in the ledger dates the attraction itself.

The numbers

Five years, consolidated, 百万円. Revenue 275,728 F1 → 483,123 F2 → 618,493 F3 → 679,374 F4 → 704,539 F5. Ordinary profit 11,278 F6 → 111,789 F7 → 166,005 F8 → 173,328 F9 → 169,641 F10. Profit attributable to owners 8,067 F11 → 80,734 F12 → 120,225 F13 → 124,160 F14 → 121,881 F15. Both of the last two lines rose into 第65期 and fell only in 第66期 — one year, not two — operating profit −2.1% F335, ordinary profit −2.1% F337 and net profit −1.8% F339 on revenue +3.7% F333, with management stating plainly that revenue set a record while every profit line went backwards [E113]. What has fallen for two consecutive years is the margin and the return: ordinary margin 26.84 D32 → 25.51 D33 → 24.08% D34, and ROE 13.5 F51 → 12.9 F52 → 11.7% F53.

The margin picture is a violent recovery followed by a two-year decline. Ordinary margin 4.09 D30 → 23.14 D31 → 26.84 D32 → 25.51 D33 → 24.08% D34; net margin 2.93 D35 → 17.30% D39. Gross margin fell 40.21% D27 → 38.74% D26 and operating margin 25.33% D29 → 23.90% D28. Segment margins at 第66期: Theme Park 22.39% including inter-segment D40 (22.96% on external revenue D41), Hotel 30.78% D42, Other 2.19% D43 — against Theme Park 24.82% D44 and Hotel 27.35% D45 a year earlier. Consolidated −2.1% F335 masks the divergence: Theme Park operating profit −7.1% F343 on revenue +2.9% F341, Hotel +20.9% F347 on revenue +7.8% F345, Other −22.9% F351.

All of the growth was price. Attendance was 99.9% of the prior year F373 while Theme Park revenue ran 102.9% F353; attraction-and-show 102.3% F363, merchandise 103.2% F365 and food 106.0% F367. Management states per-guest spend reached a record in all three categories [E114], and hotel room rates rose on revenue management, helped by a full year of the Fantasy Springs hotel [E115], with management stating a high room occupancy rate was maintained [E279]. The filing never publishes a per-guest spend figure, an occupancy rate, a room count or a park capacity — attendance is the only quantified volume metric anywhere [E279], even though the auditor's own procedures confirm unit-price masters and occupied-room data exist inside the systems and were audited [E253][E254].

The cost side is where the year was decided. Cost of sales rose 406,180 F1012 → 431,580 F1013 on revenue up 25,165, so gross profit fell 273,193 F1014 → 272,959 F1015. Inside parent cost of sales, personnel rose 79,157 F2180 → 88,082 F2181, +11.3%, with bonuses alone 8,966 F2176 → 13,096 F2177, +46.1%, and facility-renewal expense 28,697 F2186 → 32,213 F2187, +12.3%. Sales promotion fell, 7,658 F2192 → 7,632 F2193. Consolidated headcount ran 9,094 F79 → 8,945 F80 → 9,719 F81 → 10,507 F82 → 11,207 F83, +23.2% over the window and +15.3% over two years against two-year revenue growth of 13.9% F3F5; parent average salary rose 0.7% to ¥6,051,217 F866F867, so the cost arrived as headcount and bonuses rather than base pay. SG&A rose 101,082 F1016 → 104,546 F1017.

The balance sheet is a fortress that is already spent. Total assets ¥1,629,076百万円 F1805 against net assets ¥1,100,021百万円 F1007 and total liabilities ¥529,055百万円 F988 — an equity ratio of 67.5% F48, recomputed at 67.52% D90. Cash and deposits ¥468,214百万円 F895 plus securities ¥128,884百万円 F901 is ¥597,098百万円 D69; interest-bearing debt is bonds ¥310,000百万円 D64, borrowings ¥16,706百万円 D65 and leases ¥237百万円 D66, total ¥326,943百万円 D67; net cash ¥270,155百万円 D71. Then deduct ¥332,929百万円 of contracted, unpaid capital D111 — 1.84× one year's operating cash flow D112 — and deployable cash is −¥62,774百万円. Property, plant and equipment is ¥837,247百万円 F937, 51.39% of the balance sheet D92, with land carried at ¥115,568百万円 F927; there is no goodwill and no impairment in either disclosed year [E237]; and there is no 「重要な会計上の見積り」 note in the filing at all [E275], so no asset-grouping unit and no impairment methodology exists for half the assets.

Cash conversion is clean; free-cash conversion is not. Operating cash flow ¥181,281百万円 F1176 is 1.0764× operating profit D107, after 1.1352× D108. But free cash flow ran −84,382 D95 → 23,303 D96 → 176,409 D97 → −57,752 D98 → 9,185 D99 — negative in two of five years, cumulating to ¥66,763百万円 D104 against ¥796,674百万円 of operating cash D102 and ¥729,911百万円 of investing outflow D103. That is 8.4% conversion, and about 14.7% of the ¥455,067百万円 of cumulative net income across the same five years F11F12F13F14F15. Capex ran 1.26× depreciation D105 after 1.57× D106, against depreciation of ¥66,545百万円 F1146 and ¥65,422百万円 F1145.

Returns are high on installed capital and negative on the increment. Return on net operating assets is 16.11% D85; ROIC excluding cash 14.96% D79; ROIC including cash 9.03% D76; NOPAT ¥120,629百万円 D63 on an effective tax rate of 28.37% D62. ROE ran 1.1 F49 → 10.2 F50 → 13.5 F51 → 12.9 F52 → 11.7% F53. Against that: invested capital rose ¥182,626百万円, 1,244,338 D75 → 1,426,964 D74, while operating profit fell F1018F1019; net operating assets rose ¥43,098百万円 D83D84 while NOPAT came in at ¥120,629百万円 D63, below the ~¥123,279百万円 profiles/claude.md C14 derives for the prior year — the fall is that derivation's, not a ledger row's. Segment returns on segment assets: Theme Park 14.49% (130,488 F1791 on 900,645 F1799); Hotel 23.73% (36,851 F1792 on 155,317 F1800); Other 0.59% (481 F1795 on 81,585 F1802) — and ¥498,191百万円 of unallocated corporate assets F1822, 30.58% of the balance sheet F1805, earn no operating return at all.

The cost of money is rising while the programme is still to be paid. Successive bond coupons run 0.09% F1871, 0.67% F1876, 0.94% F1881, 1.26% F1886, 1.44% F1890, 1.80% F1894 and 2.09% F1898; long-term bank borrowings average 1.42% F1916; interest expense went 1,073 F1032 → 2,455百万円 F1033, leaving interest cover at 68.6× F1019F1033. ¥20,000百万円 of bonds fall due within a year F1904 against ¥597,098百万円 of cash and securities D69, and a ¥300,000百万円 shelf stands with ¥100,000百万円 drawn F423[E118]. A ¥146,843百万円 forward book appeared this year F1531, ¥142,821百万円 of it euro purchases F1528 — 43.28% of the cruise budget D93 — carrying a ¥21,247百万円 fair value, 14.47% of notional D94F1533, where the prior year carried no hedge-accounted currency derivatives at all [E226][E278].

What owners got. Dividends paid ¥22,919百万円 F1206; buybacks ¥0 F1208 — the only acquisitions were 70 below-unit shares, with no AGM or board resolution behind them [E146]. Total return to shareholders was 18.80% of net income D121 against 69.67% the prior year D119, and the whole difference is a ¥61,830百万円 treasury purchase F1207 executed off-market from Keisei Electric Railway at the previous day's close [E238]F1826 — 49.80% of that year's earnings D117. Share count fell 0.99% across the entire window on the 18,000,000-share cancellation of 2024-12-17 D122F534, with weighted-average shares 1,641,816千株 F1835 → 1,639,427千株 F1836; treasury is 8.93% of shares issued D9 and held rather than cancelled. Payout is 21.41% on the parent basis F167D114, 20.21% consolidated D115, against a 30% target dated 2035 [E149].

Incentives, in the one year the archive can score. All three remuneration metrics were targeted below the prior year's actual: operating profit ¥160,000百万円 targeted F747 against ¥172,111百万円 achieved F1018; "operating cash flow" ¥180,000百万円 targeted F751 against ¥189,582百万円 on the note's own definition (F14 + F1145); ROE 11.1% targeted F755 against 12.9% achieved F52. All three were then "achieved" at 105.2% F749, 104.6% F753 and 105.8% F757 while every one of them fell year on year. The note defines 「営業キャッシュ・フロー」 as profit attributable to owners plus depreciation [E198], which reconciles exactly — ¥121,881百万円 F15 + ¥66,545百万円 F1146 = ¥188,426百万円, the printed 1,884億円 F752 — while the cash-flow statement reports ¥181,281百万円 F1176 for the same year, two definitions of one term inside one filing [E274], with the FY2029 ¥300,000百万円 headline target F256 never saying which applies. Total officer remuneration is ¥520百万円 F772, of which directors' fixed cash ¥339百万円 F759, performance-linked ¥73百万円 F760 and share-based ¥16百万円 F761; the cash cap dates from a 1999 resolution taken with 25 directors in office F736.

The register, and who sits on the board. Keisei Electric Railway holds 20.10% of voting rights, including 0.08% held indirectly F251 — the parenthesised indirect figure is 内数, inside the total, not an addition to it F252, so the direct holding is 20.02% — and the stake is stated three different ways on three different bases and never reconciled: 20.00% of shares ex-treasury F569, and 20.06% direct plus 0.07% indirect, side by side and additive, in the prior-year related-party note F1824F1825 [E272]. Mitsui Fudosan holds 5.75% F573, Chiba Prefecture 4.02% F578, Urayasu City 0.80% F589, a Company-linked foundation trust 1.10% F580 and a retirement-benefit trust 1.03% F582. Three of eight outside officers are former Keisei directors [E185][E173][E179][E181], justified only against a transaction-size test, while the Company's own written independence standard disqualifies an executive of any ≥10% holder [E189], and 「独立役員」 appears nowhere in the document [E273]. Individuals hold 34.41% of the shares F558 and are 597,420 F556 of the register's 604,446 holders F559, drawn in part by a shareholder benefit paid in park passports [E248]; foreign institutions hold 13.05% F552. All officers together hold 1,462千株 F692, of which the chairman's 1,094千株 F685 is 74.8%; the CEO holds 130千株 F686 and the president 91千株 F687. KPMG AZSA has audited for a stated 39 years with the filing's own caveat that the true period may be longer [E194]; opinions are unqualified throughout [E251][E256]; and the Company's non-audit fees are for work performed on the written instructions of Keisei's audit accountants [E196].

Where the stamp sits. ¥2,920.00 F2485 is 4.35× book D7, 39.28× earnings D8, 6.80× sales D14 and a 0.51% dividend yield D10 on a net market capitalisation of ¥4,787,975百万円 D6. It is 21.4% below the 第66期 high of ¥3,715 F193D15 and 12.3% above the year's low of ¥2,601 F199D16; stamp-day turnover was 0.65% of the float D17. The filing's own PER series compressed 954.5 F54 → 91.9 F55 → 66.1 F56 → 38.9 F57 → 36.3 F58. And the Company printed its own scorecard: total shareholder return 146.6 F180 → 89.8 F181 → 82.9 F182 against a dividend-inclusive TOPIX of 152.5 F185 → 150.2 F186 → 202.2 F187. The holder is down 17.1% over the window while the market is up 102.2%.

The five lenses

Buffett — watch, buy below ¥840

A shopkeeper would follow this business in a minute. Two parks on ground reclaimed from the sea sold about 27.5 million admissions last year F372. Once through the turnstile people bought rides and shows F1749, souvenirs F1753 and lunch F1757; six hotels next door took another ¥119,049百万円 F1773; a little monorail carries folks between the gates [E35]. Parks are 80.67% of the till D47, hotels 16.90% D48. The Company reclaimed that land itself between 1964 and 1975 [E11] and owns it outright — 837,259㎡ under one park and 718,467㎡ under the other [E134]F449F457. That is a toll bridge with a moat you can walk around.

Now the part that took me longer, and which is the whole of this memo. The Company does not own the mouse. It rents him. Ten agreements with Disney Enterprises, Inc., the first from April 1979 [E259]. Nine of them end on one single day, 3 September 2051 [E260]. Contracts signed thirty-nine years apart share one terminal date and the filing never says why. Each may be extended five years at a time, five times — but the wording is 「各当事者は」, which reads either as "either party may" or "the parties may," and the same table proves the drafter knows the difference, because the cruise contract says 「当社は」 when it means the Company alone [E262]. So I cannot tell you whether Oriental Land can carry this to 2076 by itself, or whether Disney can simply decline in 2051.

What is the rent? 「一定料率にしたがって」 [E130]. That is the entire disclosure. The only quantum in 142 pages is one line in the parent-only accounts: royalty ¥42,048百万円 F2131, 7.15% of the parent's own sales D86 — an inference of mine, not a contract term. And here is what sits badly. Management tells you it has exactly two sources of 「独自の競争優位性」: the land, and the licence [E42]. It then enumerates thirteen principal risks [E86] and the licence is not among them [E264]. No termination clause, default, cure period, change of control, assignment restriction, governing law or dispute mechanism is disclosed for any of the ten [E265]. Nothing about what becomes of the parks, the marks or ¥837,247百万円 of plant F937 when the term runs out [E268]. If you cannot restate a footnote after a careful reading, that is usually a signal — and here there is no footnote at all to restate.

So the arithmetic. Owner earnings. Reported profit ¥121,881百万円 F15, add back depreciation ¥66,545百万円 F1146, subtract the spending needed to hold position and volume. I triangulate three ways: maintenance capex at depreciation gives ¥121,881百万円 F15; all capex, ¥86,269百万円 F1819, gives ¥102,157百万円; the honest middle is theme-park plus hotel capex of ¥77,823百万円 F1815, leaving the cruise out as growth — ¥110,603百万円. I lean to the middle because of a hard fact: attendance actually fell 0.1% F373 while they spent 1.26 times depreciation D105, and 1.57 times the year before D106. When you must spend above depreciation merely to keep the turnstile count flat, the tooth fairy is not paying for the capital expenditures.

Against a market value of ¥4,787,975百万円 D6 that is a 2.31% owner-earnings yield. The Company's own ten-year paper pays 2.09% F1898 and its actuaries discount pensions at 2.7% [E228]. You are being paid roughly the bond rate to own the equity of a business whose revenue rose 3.70% D25 while operating profit fell 2.1% F335 and the ordinary margin stepped down for a second straight year from its 第64期 peak D32D33D34.

Now the value, and it is one assumption wide. Capitalise ¥110,603百万円 over the twenty-five years the document actually commits to, at 2.09% F1898, add net cash ¥270,155百万円 D71, subtract ¥332,929百万円 of committed unpaid capital D111, and give the cruise nothing — because a business that has not sailed and sits inside a 2.19%-margin segment D43 earns nothing in my ledger. That is ¥1,265 a share. Run it to 2076 with all five extensions taken and 1% growth: ¥2,558. Run it to perpetuity: ¥3,189. The price is ¥2,920 F2485. Which is to say: the market is paying for a licence that never ends and never gets repriced. That is one soft assumption carrying the entire answer, and beware of geeks bearing formulas.

Two other things a partner should know. Management's own bonus metric is 「営業キャッシュ・フロー」 defined as net income plus depreciation [E198] — pre-capex cash flow, the very number that means nothing in a business like this — and it paid out on 1,884億円 against a 1,800億円 target F752F751, while the same phrase means something ¥7bn different in the cash-flow statement [E274]F1176. All three bonus indices landed between 104.6% F753 and 105.8% F757 in the same year. And the one buyback of size in the window was ¥61,830百万円 bought from Keisei, the ~20% holder, at the prior day's closing price [E238] — 49.80% of that year's earnings D117 — while in the year the stock traded down to ¥2,601 F199 they bought seventy shares [E146].

This is a wonderful business. It is not, at ¥2,920 F2485, a fair price. I will name one: ¥840, a third off the only value I can defend from the document. That is a long way down, and I may wait forever. There is no called strike for not swinging.

What a student should take from this: when a company rents its moat, the lease term is the valuation, and everything else is detail. A price that implies a perpetuity while the filing states a terminus — and refuses to disclose the rent, the renewal mechanics, or what happens at expiry — is not a hard company to value so much as a company whose most load-bearing input has been withheld. And when management's own incentive metric is earnings plus depreciation in a business that must spend above depreciation just to hold volume flat, you have been told which number they would rather you looked at.

Munger — watch, buy below ¥1,500

Here is a business I would like to own and a price I will not pay for it.

Start with what is real. Oriental Land owns, outright, 837,259㎡ under Tokyo Disneyland and 718,467㎡ under Tokyo DisneySea [E134]F449F457 — land it reclaimed from the sea itself between 1964 and 1975 [E11]. That is not a lease, not a franchise, not a royalty stream. That is the one unimpeachable moat in this filing: nobody builds a second Tokyo Disney Resort next door, because there is no second parcel like it and no second licence to put on it. Segment margins say the same thing in numbers — 22.96% on theme-park external revenue D41, 30.78% on hotel revenue including inter-segment sales D42, two different denominators — recovered from an unnormalised COVID trough of 4.09% ordinary margin D30 to a peak of 26.84% two years later D32. Attendance is flat at 99.9% F373, and every incremental yen of the 3.7% revenue rise F333 is guests paying more for the same visit, in every category, at record levels [E114]D54. That is exercised pricing power, not banked — See's-style, except the price gets pulled the moment volume goes flat, not left on the table.

Now invert it, because a moat you only admire is a moat you haven't tested. The parts that sit on that land are licensed — ten agreements since 1979, nine expiring on the identical date of 3 September 2051 [E260], extendable in wording so loose I cannot tell you whether the right is unilateral or mutual [E262]. One contract, the Celebration Hotel, states no extension at all and simply stops in 2038 [E126]. The cruise contract — a ¥330,000百万円 commitment F519 with ¥332,929百万円 of committed capital unpaid across the programme D111, 1.84 years of a full year's operating cash flow D112 — runs thirty years from a launch that has not happened, extendable only at the Company's own option [E129]. And the royalty for all of it is one sentence: 「一定料率にしたがって」 [E130]. No number. Not one. What I can compute is 7.15% of parent sales D86, which is an inference about a cost, not a fact about a contract. Thirteen risk factors are enumerated and not one of them is the licence [E264]. Management names the Disney contract as one of exactly two sources of distinctive competitive advantage on page 9 [E42] and then, for 142 pages, never risks it, never terms it, never prices it. That is not thoroughness. That is a company that has decided its readers do not need to know what it pays for its own name.

I would tolerate one such silence. I do not tolerate the pattern. There is no accounting-estimates note in this filing at all — no impairment methodology, no asset-grouping unit — against ¥837,247百万円 of property F937 and an ¥11,089百万円 deferred-tax asset that only exists because impairments happened before [E275][E230]. "Operating cash flow" is defined two different ways in the same document, one for management's own bonus and one on the cash-flow statement, and the flagship ¥300,000百万円 FY2029 target F256 never says which applies [E274]. A ¥142,821百万円 euro forward book appeared this year where there was none the year before, tied to nothing the filing names F1528[E278]. And Keisei Electric Railway — a ~20% owner F251, three of whose former directors sit among the eight outside officers [E185] — sold the Company ¥61,830百万円 of its stake back to the Company itself last year, off-market, via ToSTNeT-3 [E238]F1207, while the Company's non-audit fees are for work performed on Keisei's own auditor's instructions [E196], and the Company's own written independence standard, which disqualifies an executive of any 10%-plus holder, is waived for these three on a test that only checks transaction size [E189][E273]. None of that is illegal. All of it is exactly the kind of grime that caps a verdict regardless of the multiple, because a shareholder who cannot trust the footnotes cannot trust the moat they're paying for either.

So price it. At ¥2,920 F2485 you pay 39.28× last year's earnings D8 and 4.35× book D7 for an operator whose ROE peaked at 13.5% F51 and has fallen two years running to 11.7% F53, whose ordinary margin has fallen two years running from its own peak D32D34, and whose five-year free cash flow comes to ¥66,763百万円 cumulative — just 8.4% of the ¥796,674百万円 of operating cash the business generated over those five years D104D102, and about 14.7% of the ¥455,067百万円 of cumulative net income F11F12F13F14F15. The rest went back into the ground. This is the pari-mutuel trap exactly: everybody can see this is a wonderful business, so the market has already bid it to a price where wonderful is merely priced in, not on offer. A great business at a fair price beats a fair business at a great price — this is a great business at a rich one.

Verdict: watch. Buy below ¥1,500 — roughly 20× last year's EPS of ¥74.34 F40 and 2.2× book value of ¥670.86 F35, a price at which I am paying a fair multiple for an operator with a declining margin trend and a landlord I cannot fully audit, with room to be wrong.

What a student should take from this: a "distinctive competitive advantage" named on page 9 and absent from the risk section on page 27 is not an oversight — go looking for what the document is quietly not pricing. Cheap-versus-its-own-52-week-high is not cheap; check the multiple against the earnings, not against last month's tape. And when a filing coexists with two definitions of the same cash-flow metric, assume the ambiguity is doing someone a favor, and it usually isn't yours.

Pabrai — pass, buy below ¥450

Start where I always start. What do I lose if I am wrong?

I pay ¥2,920 F2485. Against that I own, per share, ¥164.76 of net cash — ¥270,155百万円 net of every yen of interest-bearing debt D71, over 1,639,717,407 shares D2. Add the cross-shareholding book at ¥59,306百万円 F943, another ¥36.17. Add the land at what it sits on the books for, ¥115,568百万円 F927, another ¥70.48. Call the whole marked floor ¥271 a share. That is 9.3% of my purchase price. Ninety-one per cent of my money is riding on the earning power above it.

Before anyone points at the ¥1,100,021百万円 of net assets F1007 and the 67.5% equity ratio F48 — an equity cushion is not an asset floor. Half the balance sheet, 51.39% D92, is property, ¥837,247百万円 of it F937, and it is a themed mountain in Urayasu. What is a themed mountain worth without the characters painted on it? I cannot tell you, and neither can the filing: there is no 重要な会計上の見積り note anywhere in it, so no asset-grouping unit, no impairment methodology, no recoverable amount [E275], while an ¥11,089百万円 deferred-tax asset for disallowed impairments sits in the tax note as evidence that impairments happened once [E230].

And the earning power is rented. Ten agreements, nine terminating 3 September 2051 [E260], extension wording ambiguous [E262], nothing disclosed about termination, default, change of control, assignment, governing law or dispute resolution [E265], nothing about exclusivity or territory [E266], nothing about what happens at expiry [E268]; the Celebration Hotel simply ends 2038-08-19 with no extension clause [E126]. The rent is one sentence, 「一定料率にしたがって」 [E130]. So the single largest identifiable third-party claim on this business cannot be underwritten. Management names the licence as one of only two sources of 独自の競争優位性 [E42], then enumerates thirteen principal risks and not one of them is the licence [E264]. That is not a hidden risk. It is a risk the filing declines to discuss, which is different and worse.

Now — I want to be fair, because this is a very good business, and leverage does not kill here. Net cash ¥270,155百万円 D71. Everything maturing inside three years — 26,058 F1445 plus 32,443 F1446 plus 12,450 F1447 — is ¥70,951百万円 against ¥597,098百万円 of cash and securities D69. Secured debt is ¥576百万円 [E222]; guarantees ¥1,147百万円 [E244]. It also passes the shutdown test in the hardest way possible, by having done it: in 第62期 revenue fell to ¥275,728百万円 F1 and operating cash flow was still positive at ¥54,602百万円 F59. Wonderful. That is survival, and survival is table stakes.

The trouble is the second question, and it decides everything. Enterprise value is ¥4,787,975百万円 D6 less ¥270,155百万円 of net cash D71 — ¥4,517,820百万円. What did the last five years hand owners in free cash? ¥66,763百万円 D104. Cumulative. Against ¥796,674百万円 of operating cash generated D102 — an 8.4% conversion. Five years of demonstrated free cash flow is 1.5% of enterprise value. Even on the kindest single-year measure, ¥97,677百万円 D101, five years covers 10.8%. I look for the hidden P/E of 1. This is the hidden P/E of seventy.

And the treadmill is accelerating: ¥332,929百万円 of committed, unpaid capital D111, 1.84 years of one year's operating cash D112, of which ¥305,718百万円 is a ship D110 that management says will out-earn the parks [E51] against a filing that discloses zero research and development [E132]. Meanwhile I am paid 0.51% to wait D10; total return to shareholders was 18.80% of net income this year D121 against 69.67% last D119, and the whole difference is the missing buyback F1208. The one large repurchase, ¥61,830百万円 F1207, was a negotiated block bought off-market from the 20% shareholder [E238] — an accommodation, not an opportunistic purchase. Share count has fallen 0.99% in the entire window D122. This is not a cannibal.

So the crayon. Take the good year, not the average, and be generous: NOPAT ¥120,629百万円 D63. Pay ten times it for a business whose licence has a printed expiry twenty-five years out and whose rent is undisclosed — ¥1,206,290百万円. Add net cash ¥270,155百万円 D71. Over 1,639,717,407 shares D2 that is about ¥900 a share. Price is 3.2 times that. Half of it, which is my standing minimum, is ¥450.

Am I being harsh? At 39.28× earnings D8 and 4.35× book D7, with the ordinary margin declining two years running — 26.84 D32, 25.51 D33, 24.08% D34 — and the theme-park segment margin down 244 basis points D40D44 on flat attendance F373, no. The market has already run this experiment: total shareholder return 146.6 F180, 89.8 F181, 82.9 F182 against TOPIX with dividends at 152.5 F185, 150.2 F186, 202.2 F187. Three years of paying up for the best.

Pass. Not too-hard — the business explains itself in three sentences and I can name exactly what the market fears. Pass on price against a soft floor, which is the oldest reason there is. Buy below ¥450.

What a student should take from this: the floor is a percentage, not an amount — ¥445,029百万円 of marked assets D71F943F927 sounds enormous until you divide by a ¥4,787,975百万円 market capitalisation D6 and get 9.3%. When a business is asset-heavy but the assets are purpose-built around somebody else's intellectual property, book value measures what was spent, not what could be recovered. Second: read the ratio of cumulative free cash flow to cumulative operating cash flow before you read anything else; 8.4% over five years D104D102 told me more about who this business serves than every margin in the filing. Third: a moat with a printed expiry date and an undisclosed rent is still a moat, but you may not extrapolate it to perpetuity — and 39 times earnings D8 extrapolates it to perpetuity for you.

Li Lu — too-hard, no threshold issued

I have spent my working life on one question: can I know a business well enough to own the whole of it for ten years? Oriental Land gives that question an unusually clean answer, and the answer is no — not because the business is poor, but because the document that describes it withholds the one thing that decides it.

Begin where an owner begins. This is a very good business. Twenty-seven and a half million people walked through two gates in Urayasu last year F372, spending ¥20,642 each inside the parks D54. The theme parks earned 22.96% on external revenue D41 and the hotels 30.78% on revenue including inter-segment sales D42 — the two are not on one denominator. On the capital actually working in the operations the return is 16.11% D85. And the record holds a real stress test: in 第62期 revenue was ¥275,728百万円 F1 against ¥704,539百万円 now F5, ordinary profit ¥11,278百万円 F6 against ¥169,641百万円 F10. It came through without issuing a share — the count has only ever fallen, by cancellation F534 — and equity rose in every one of the five years F21F22F23F24F25. Dividends, restated onto one basis, went ¥5.60 D12 to ¥15.00 F134, up 2.68× D13, not the cut the printed series implies. This is a company that survives, run by people who have not diluted their owners.

Now read the lease. Ten agreements with one American counterparty since April 1979 [E259]. Nine terminate on the same day, 3 September 2051 [E260] — twenty-five years and a month after the price I am being quoted. Each carries 「各当事者はさらに５年間ずつ、５回にわたり延長することができる」 [E262]. Read it slowly. Does it mean I may extend, or that we may extend? The filing never resolves it, never says whether extension needs consent, what happens if the parties differ, or whether the commercial terms — above all the rent — are re-set on extension. And we know the drafter distinguishes when he means to: the cruise contract on the same page says 「当社は」 [E129].

The rent is one sentence [E130]. The rate is not stated. The base is not stated. There is no disclosed minimum, cap, step or escalator [E263]. The only quantum anywhere is a parent-only cost line of ¥42,048百万円 F2131 — 7.15% of parent sales D86, 5.97% of consolidated revenue D88 — two ratios, neither of them the contractual rate.

Sit with what that means. Management names exactly two sources of 「独自の競争優位性」 [E42]. Thirteen principal risks are enumerated over seven pages and not one concerns the licence [E264]. 「ライセンス」 occurs in three places in the whole filing [E269]. No termination right, no default or cure provision, no change-of-control clause, no governing law, no dispute mechanism [E265]. There is no exclusivity, no territory, no description of what is licensed [E266]. The single most widely believed fact about this company — that it holds Japan exclusively — is not a disclosed fact. So half of the stated moat is a contract whose price is unstated, whose scope is unstated, whose renewal control is ambiguous, and whose risk the issuer does not acknowledge to be a risk. No further public work closes that. It is not a shortfall in my reading; it is the outer edge of what any owner of this security can know. Knowing where that edge sits is the entire discipline.

I can bound part of it. A doubling of the royalty costs ¥42,048百万円 F2131 — about 24.97% of operating profit F1019, roughly 24.7% of net income after the observed tax rate D62. Painful, survivable, not permanent loss. And ten years from today is 2036, comfortably inside the terminus: if I were buying ten years of earnings, the licence would not decide this.

But at ¥2,920 F2485 I am not buying ten years of earnings, and neither are you. The shares are 39.28× reported earnings D8 and 4.35× book D7, on a 0.51% dividend D10. Put it the way I find hardest to argue with: the operations earn about 16% on their capital D85, the equity earns 11.7% F53, and because I pay 4.35× book D7 the return reaching me is 11.7 ÷ 4.35 = 2.7% before growth. Over a long hold my return approximates the business's return on the capital I commit, not the capital already in the ground. At 39× D8 nearly the whole of what I buy is terminal value — the years after 2036, which are the years astride and beyond a date the contract prints.

The near term does not rescue it. Revenue set a record and every profit line fell [E113]; gross profit was lower in absolute yen on 3.7% more revenue F1014F1015; the ordinary margin has declined two years from its 第64期 peak D32D34; EPS went 73.39 F38, 75.62 F39, 74.34 F40; attendance was 99.9% F373. Across five years free cash flow cumulates to ¥66,763百万円 against ¥796,674百万円 of operating cash — 8.4% D104D102 — and ¥332,929百万円 more is committed and unpaid D111, 1.23× the net cash D71. Over the same window retained earnings returned ¥0.95 of owner value per yen retained: book value per share rose ¥209.04, 461.82 F31 to 670.86 F35, plus restated cumulative dividends of ¥50.00 D11F132F133F134, against cumulative EPS of ¥272.64 F37F38F39F40. Retention has not yet beaten a payout.

And I do not like the metric that pays the people making these decisions. The remuneration table defines 営業キャッシュ・フロー as net income plus depreciation [E198]. That number rises mechanically when a ¥330,000百万円 ship F519 enters service, whether or not it earns anything. I cannot even size the mechanical uplift, because the filing carries no 重要な会計上の見積り note at all [E275], against 51.39% of the balance sheet in property D92.

Finally the Asian structure, which I always read plainly. Keisei holds about 20.10% F251, stated three different ways [E272]. Three of the eight outside officers are its former directors [E185]. The Company's own independence standard disqualifies executives of a ≥10% shareholder [E189]; every stated justification answers the transaction-size limb instead, and 「独立役員」 appears nowhere [E273]. The largest capital return of the window was not an open-market buyback but a ¥61,830百万円 off-auction block bought from that same shareholder [E238]F1207; this year there was no buyback at all F1208. Meanwhile 34.41% of the register is individuals F558 across 604,446 holders F559, drawn in part by a benefit paid in park passports [E248]. A price set by owners partly buying admission is not a price I can lean on.

I will not name a buy-below. The price at which the licence unknown is genuinely absorbed — net cash ¥270,155百万円 D71 plus the parent's Maihama theme-park land at 1970s book, ¥55,976百万円 F448F456F464, plus hotel segment assets ¥155,317百万円 F1800, over 1,639,717,407 shares D2 — is ¥294 a share, 10.1% of the stamp F2485; naming it would be theatre. Only three of the five parent land parcels are added, because the Ambassador Hotel and Ikspiari parcel F472 and the Toy Story Hotel parcel F477 are ホテル-segment land already inside that ¥155,317百万円 F1800; adding all five double-counts ¥8,562百万円 and lifts the arithmetic to ¥299. Every price above it assumes a renewal I cannot verify and a rent I cannot see, so naming anything in between would be manufactured, not analysed. This one sits outside my circle. Verdict: too-hard.

What a student should take from this: when the moat is rented, the lease is the business — read it before you read the income statement. If a filing describes the rent as "a fixed rate" without stating the rate, its base, or who controls renewal, then the most important number in the company is one you will never see, and no multiple you pay for it is defensible. A moat you can name but cannot price is not inside your circle of competence, however famous the gates.

Claude — pass, implied buy-below ¥700, band ¥410–1,163

I spent the outside-view stage building a case about a licence and the arithmetic demoted it. That is the honest headline of this study, and it belongs first, because the correction runs against my own registered priors.

I registered P(the annuity treatment is closer to correct than the perpetuity treatment) ≈ 0.75, decomposed 0.12 pure perpetuity / 0.10 strict annuity to 2051 / 0.78 finite stream to a repricing date then repriced continuation, with P(licence lost) ≈ 0.06. I score the direction correct and the magnitude badly over-weighted. The direction survives: nine agreements do end 2051-09-03 [E260], the extension wording is ambiguous [E262], nothing is said about expiry [E268], and a perpetuity model contains no state for any of that. But when the ledger let me size the licensor's claim, it turned out to be ¥42,048百万円 F2131 — 24.97% of consolidated operating profit F1019 and 5.97% of consolidated revenue D88. That is a variable cost with a repricing date, not a call option on the enterprise. Run the modal branch honestly: a reset taking the licensor from roughly a quarter to roughly a third of operating profit costs ~¥16.8bn pre-tax at today's scale, ~¥12.0bn after tax at 28.37% D62, arriving in twenty-five years; capitalised at 7% and discounted twenty-five years that is ¥31,674百万円 — 0.7% of a ¥4,787,975百万円 market capitalisation D6. Weight the branches and the whole terminal-value discontinuity is worth roughly a tenth of the equity. It is real. It is not the bet. My §1 wrote 1,200 words on it and it moves the answer less than one bad year of cost inflation does.

What decides this study is duller and much larger. The next yen of revenue destroys profit. Revenue rose ¥25,165百万円 F4F5 and operating profit fell ¥3,698百万円 F1018F1019 — an incremental operating margin of −14.70% against an average margin of 23.90% D28. Gross profit fell in absolute terms, 273,193 F1014 → 272,959 F1015, so 100% of the incremental revenue was eaten by cost of sales. Widen the window past the reopening: over two fully-recovered years the incremental ordinary margin is 4.23% F8F10F3F5 against a 24.08% average D34. Only the three-year window shows 26.1% — 169,641 F10 less 111,789 F7, over 704,539 F5 less 483,123 F2 — and that window contains the reopening; my registered inference rule forbids using it, and I don't.

Underneath that sits the reinvestment answer. Net operating assets rose ¥43,098百万円 D83D84 while NOPAT came in at ¥120,629百万円 D63, below the ~¥123,279百万円 my own C14 derives for the prior year. The return on the latest increment of capital is negative; over two years it is 4.79% — two years of Δordinary profit after tax, ¥2,604百万円 on 169,641 F10 less 166,005 F8 at the observed 28.37% rate D62, over ¥54,334百万円 of capital reinvested above depreciation, capex running 1.26× D105 and 1.57× D106 against depreciation of ¥66,545百万円 F1146 and ¥65,422百万円 F1145 — against a 7% hurdle that is mine. The filing states no hurdle. Across five years the company generated ¥796,674百万円 of operating cash and kept ¥66,763百万円 of it — 8.4% D102D104 — free cash flow negative in two of the five D95D98. Capex ran 1.26× depreciation D105 and 1.57× the year before D106, and ¥332,929百万円 of committed, unpaid capital D111 now exceeds ¥270,155百万円 of net cash D71. Deployable net cash is −¥62,774百万円. The buyer of this company inherits a funding gap, not a cash pile.

I score my other registrations at the same volume. The royalty call landed: I registered 4–9% of revenue, central ~6% (actual 5.97% D88; 7.15% parent-on-parent D86) and 20–40% of operating profit, central ~28% (actual 24.97%), with P(a parent-only rate is computable) ≈ 0.70 — computable, and computed D86; P(a consolidated rate is computable) ≈ 0.15, correctly low, because there is none [E263]. The free-cash-flow call missed on the year and landed on the cumulative: I registered latest-year FCF ÷ OCF at −20% to +35%, central +10%, and the answer on the definition I actually wrote is 53.9% D101F1176 — above my whole band — because capex ran at 1.26× depreciation this year against 1.57× last D105D106, the estate sitting between project waves. My cumulative call, P(five-year cumulative FCF below 25% of cumulative OCF) ≈ 0.65, resolves true at 8.4% D104. The lesson is mine, not the company's: I predicted a single year when the durable quantity was the cumulative, and a single year of a lumpy capital programme is weather. And the attendance call is void, not wrong — the archive carries no pre-COVID year D18 and attendance is printed for one year only F372F373, so the proposition cannot resolve; P(§2 can reconstruct a per-guest spend series) ≈ 0.70 resolves false, one point exists D54 and no series does [E279].

That is the actual bet, and it is not about Disney. An owner at ¥2,920 F2485 pays 39.28× earnings D8, 4.35× book D7 and takes a 0.51% dividend D10 for a business whose normalized owner earnings — ¥121,881百万円 F15 plus ¥66,545百万円 depreciation F1146 less a ¥78,000百万円 maintenance-capex estimate — are ¥110,426百万円, a 2.31% private-owner yield against a hurdle I put at 7% (the Company's own newest bond pays 2.09% F1898, the highest coupon in a schedule that starts at 0.09% F1871, with long-term bank borrowings averaging 1.42% F1916, plus roughly five points of equity premium). The owner is paid roughly the bond rate to carry equity risk — 22 basis points over the newest coupon, and the low end of the range below, 2.19%, still clears 2.09% F1898. My profile run states this as "the owner earns less than the bondholder ranking ahead of him" (profiles/claude.md C35), and that sentence is arithmetically wrong in the owner's favour by 22bp; I record the error rather than quietly dropping it, because what survives it is the damning part — the equity of a single-site, single-counterparty asset yields what its own senior unsecured paper yields. Reverse the arithmetic and the price embeds 4.69% perpetual growth in owner earnings D6. Setting maintenance capex anywhere in its whole plausible range — at depreciation, ¥66,545百万円 F1146, or at the observed ¥83,604百万円, ¥77,025百万円 of property F1186 plus ¥6,579百万円 of intangibles F1188 and the numerator of D105 — moves the yield only between 2.19% and 2.55%. It is a property of the business, not of the normalization.

One governance finding earns its place beside the arithmetic. All three FY2026 remuneration metrics were targeted below the prior year's actual — operating profit 1,600億円 F747 against ¥172,111百万円 achieved F1018, ROE 11.1% F755 against 12.9% F52, "operating cash flow" 1,800億円 F751 against ¥189,582百万円 on the note's own definition [E198] — and all three were then "achieved" at 105.2% F749, 104.6% F753 and 105.8% F757 while every one of them went backwards. Compensation varied upward while results varied downward. The board's real principal is not the shareholder.

Verdict: pass. Not too-hard — my pre-registered trigger required that the royalty could not be sized even parent-only, and it was D86. Not watch — the gap between ¥2,920 F2485 and what the downside supports is more than fourfold, which is a pass with a published threshold, not a wait. The implied threshold is ¥700, with a published band of ¥410–1,163, and I have taken the cautious end of my own range because nine of eighteen jury selves crashed and a crash is divergence (see the synthesis, and the disclosure in ## What this taught the checklists).

What a student should take from this, in order of how much it cost me. The most interesting fact in a file is rarely the deciding one. I arrived with a genuinely good structural observation — nine licences, one printed terminal date, a royalty rate that appears nowhere — and it is true, under-processed by the standard apparatus, and worth about a tenth of the equity. Meanwhile a single unglamorous ratio, ΔOP ÷ Δrevenue = −14.7% F1018F1019F4F5, decides the answer. The discipline that saved this study was not scepticism about the licence; it was being forced to size the licence claim in yen instead of describing it. A qualitative risk you have not sized is a risk you have not analysed — you have only admired it. Second: predict the durable quantity, not the vivid one — the cumulative is the fact; the year is the weather. Third: read the incentive table as a financial statement. Target, actual and achievement were disclosed for three metrics F747F751F755, every target sat below the prior year's actual, and a year in which all three fell paid an above-standard bonus F749F753F757. That took ten minutes and three divisions, and it told me more about how this company will behave with the next ¥330,000百万円 F519 than the entire strategy section did.

Synthesis

Where the five lenses agree

First, the business is exceptional and nobody disputes it. A self-made freehold estate near Tokyo [E11][E134], a first-tier entertainment brand, 16.11% on net operating assets D85, segment margins of 22.96% on Theme Park external revenue D41 and 30.78% on Hotel revenue including inter-segment sales D42 — two bases, stated so — a 67.5% equity ratio F48 with no goodwill [E237], no potential dilutive shares [E6], no minority interests D4, and a demonstrated survival through an unnormalised closure that still produced positive operating cash flow of ¥54,602百万円 F59. All five lenses wrote that paragraph.

Second, the moat is genuinely rented, and the lease discloses a term and almost nothing else. Nine of ten agreements end 2051-09-03 [E260]; extension wording ambiguous [E262]; the Celebration Hotel ends 2038-08-19 with no extension clause [E126]; the rent is 「一定料率」 [E130]; no termination, default, change-of-control or governing-law provision [E265]; no exclusivity or territory [E266]; no expiry treatment [E268]; and the licence appears in none of the thirteen enumerated risks [E264] despite being one of the two advantages management names [E42]. Five lenses found the same silences and cited the same rows.

Third, the cash does not reach owners. Five-year free-cash conversion 8.4% D104D102; committed unpaid capital ¥332,929百万円 D111 exceeding net cash ¥270,155百万円 D71; a 0.51% yield D10; 18.80% of net income returned D121; the only buyback of size an off-market block from the 20% holder at full price [E238]D117; a 30% payout target dated 2035 [E149] against 21.41% today F167. Four of five named this among their load-bearing items.

Fourth, the disclosure has a consistent tilt and all five name the same instances. Two definitions of 「営業キャッシュ・フロー」 in one filing with the FY2029 headline target silent on which applies [E274]F256; no 重要な会計上の見積り note at all against ¥837,247百万円 of property [E275]F937; zero R&D beside a stated intention to conduct R&D [E132][E270]; the takeover-defence disclosure contradicting itself twice on one page [E276]; Keisei's stake stated three ways [E272]; and no 「独立役員」 designation anywhere against three ex-Keisei outside officers [E273].

Fifth, the growth engine is price and the filing will not let you test it. Attendance 99.9% F373 against theme-park revenue 102.9% F353, with a claimed record per-guest spend [E114] — and no per-guest spend, occupancy, room count or capacity figure published anywhere [E279], although the auditor used exactly those data [E254].

Where they diverge — a staged exchange

---

On what the licence is worth — and this is where the study corrects itself.

BUFFETT (B99, intrinsic value as discounted cash; B2, the can't-estimate-it test). I priced the licence as an annuity, not a perpetuity, and I want the reader to see how much that one choice moves. The same ¥110,603百万円 of owner earnings F15F1146F1815, capitalised at the Company's own 2.09% coupon F1898, with net cash added D71 and committed capital deducted D111 and the cruise valued at nothing D43, is ¥1,265 a share over the twenty-five years the document commits to; ¥2,558 to 2076 with all five extensions taken and 1% growth; ¥3,189 in perpetuity. Identical earnings, three answers, and the market at ¥2,920 F2485 has chosen the third. My buy-below of ¥840 is a third off the only one of the three I can defend from the filing.

CLAUDE (C98, epistemic-limits inventory; C92, halo-metric audit). *And I registered that same worry figures-blind at P ≈ 0.75, and then the ledger made me size it, and the size is where we part. The licensor's whole contractual claim is ¥42,048百万円 F2131 — 24.97% of operating profit F1019. The modal 2051 reset is worth ¥31,674百万円 of present value, 0.7% of ¥4,787,975百万円 D6, and across all branches the terminal-value discontinuity is worth about a tenth of the equity. My §1 gave it the weight of a verdict; it ranks eighth. Warren, we land ¥1,265 against my ¥700 by different routes, and yours arrives via the worry that turns out to be the smaller one. What I would put in its place is C13: the price does not fail because the stream stops in 2051; it fails because the stream is not growing today. Revenue +¥25,165百万円 F4F5 produced operating profit −¥3,698百万円 F1018F1019.*

BUFFETT. Then look at where my own item lands us and note that it agrees with you. B33 — incremental returns — fails outright: operating profit fell ¥3,698百万円 F1018F1019 while invested capital rose ¥182,626百万円 D74D75. And B45 — depreciation versus reality — is the same fact from the asset side: capex at 1.26× depreciation D105 after 1.57× D106 while attendance fell 0.1% F373. My annuity arithmetic set the level of value; your ratio sets its direction. Two lenses reached similar prices by different routes, and one of us was over-weighting the right worry.

---

On the order of operations — why Pabrai never has to adjudicate the licence at all.

PABRAI (P53, the fifty-cent dollar; P60, not overpaying for "the best"). I want to be exact about my logic order, because it is the cheapest thing in this study to reproduce. My conservative value is ten times NOPAT of ¥120,629百万円 D63 plus net cash of ¥270,155百万円 D71, over 1,639,717,407 shares D2 — about ¥900 a share. The stamp is 3.2× that and 6.5× my ¥450 threshold. The price fails my valuation test by six and a half times before the licence needs adjudicating at all. Where the licence enters is in the multiple: ten times, not fifteen, precisely because the term is printed [E260] and the rent is not [E130]. So the licence sets my buy-below. It does not set my verdict. And the reason mine is pass rather than watch is P17 — I am paid 0.51% to wait D10 on an 18.80% payout D121 with a zero buyback F1208. An expensive great business you are paid to hold is a watch. One you are not is a pass.

MUNGER (M88, margin of safety; M44, the pari-mutuel). And I am the outlier at ¥1,500, so let me say what my number is, because a reader would otherwise take it for an appraisal. It is a multiple gate: roughly 20× the ¥74.34 of earnings this business just printed F40 and 2.2× the ¥670.86 of book behind it F35. I am not claiming to value the licence. I am saying there is a multiple at which I would own an operator with a declining margin trend D32D34 and a landlord I cannot audit [E130], and 39.28× D8 is not it. Mohnish, we differ by 3.33× and neither of us is doing valuation the way Warren is — he is discounting a stream, you are halving a capped multiple, and I am naming a price at which the pari-mutuel odds stop being short.

---

On what caps Munger's verdict regardless of the multiple.

MUNGER (M54, incentives and management character; M83, the oversight layer; M38, the free-cash test). Two findings, and they are not price findings. The first is arithmetic: five-year cumulative free cash flow of ¥66,763百万円 is 8.4% of the ¥796,674百万円 of operating cash the business generated D104D102 — about 14.7% of ¥455,067百万円 of cumulative net income F11F12F13F14F15. The rest went back into the ground. The second is grime. Keisei at ~20% F251, stated three ways and never reconciled [E272]; three of eight outside officers its former directors [E185]; the Company's own rule barring executives of ≥10% holders, waived on a transaction-size test [E189][E273]; no officer anywhere designated 独立役員 [E273]; non-audit fees for work performed on Keisei's own auditor's instructions [E196]; and the largest capital return of the window a ¥61,830百万円 off-market block bought from that holder at the prior day's close [E238]F1207. None of it is illegal and all of it caps the verdict at watch regardless of the multiple.

PABRAI. I graded the same facts at P49 and reached "pass, with flags" rather than a cap, and the difference is what each of us thinks the flag is evidence of. Charlie treats governance grime as a standing tax on any price. I treat it as one more reason the ten-times cap on NOPAT D63 is right rather than fifteen. We end up in the same place by different accounting — his through the verdict, mine through the multiple.

---

On why Li Lu alone refuses a number — and it is a knowability claim, not a value claim.

LI LU (L34, the legal/licence barrier; L1, the ten-year knowledge bar; L5, unknowns absorbed by the discount). Let me name exactly what I could not know, because "too hard" without the sentence is a posture. Three variables decide the decade and the record answers one. Per-guest spend is referred to four times and quantified never [E279]. The royalty's rate, base and escalators are undisclosed [E263]. Extension control is textually ambiguous [E262]. Two of the three are contractual, which means no further public work closes them — that is the distinction between a gap in my reading and the outer edge of what any owner of this security can know. And here is the part I will not soften: the floor at which that unknown is genuinely absorbed — net cash D71, the parent's Maihama theme-park land at 1970s book F448F456F464, the hotel segment's assets F1800 — is ¥294 a share, 10.1% of the stamp F2485 — the two hotel land parcels F472F477 are not added on top of that, being already inside the segment figure F1800. Every price above it assumes a renewal I cannot verify and a rent I cannot see, so naming anything in between would be manufactured, not analysed.

LI LU, continued. And I will add one finding the others did not price, because it bears on every yen of the ¥332,929百万円 still to be spent D111. The remuneration metric 営業キャッシュ・フロー is defined in the pay note as net income plus depreciation [E198]. That number rises mechanically the moment a ¥330,000百万円 ship F519 enters service, whether or not the ship earns anything. I cannot even size the mechanical uplift, because there is no 重要な会計上の見積り note in the filing at all [E275]. If rationality with retained capital is the defining management virtue, a pay metric that pays for gross capex is evidence against it.

CLAUDE. *I sized what you declined to price and got to the same door from the other side. C35's private-owner yield is 2.31%, and it stays between 2.19% and 2.55% across every defensible maintenance-capex assumption F1146D105 — so the yield is a property of the business, not of my normalization. That is a pass at ¥2,920 F2485 with a published number at ¥700. Li Lu, your ¥294 and my ¥700 are not really in conflict: yours is the price at which an unknowable contract is fully absorbed, mine is the price at which an observed record clears a 7% hurdle. Yours prices the thing nobody can learn; mine prices the thing everybody can already see. The one place I would push back is L13: at 39.28× D8 you say almost all value is terminal and the terminus is printed. On my sizing the terminus is worth about a tenth of the value, and the other nine-tenths is at risk from a −14.7% incremental margin that resolves in a document filed next June, not in 2051.*

---

On the jury, and on a run that must be reported as degraded.

CLAUDE (C108, one-directional jury consumption). Eighteen selves were dispatched across the six jury-eligible items. Nine returned. Nine crashed when the session's subagent budget was exhausted, taking out all three selves for C89, C98 and C102 — that is the profile run's own wording; no agent-count limit is recorded anywhere in this repository. Under my own rule a crash counts as divergence, and divergence may only widen an interval or cap a claim — never narrow or strengthen one. So: the threshold moved down from ¥750 to ¥700, the published band widened to ¥410–1,163, and C98's conclusion that "no unknown blocks the verdict" is flagged uncorroborated. Three jury items carry no corroboration at all, and I state that rather than presenting the run as complete. Of the nine that did return, the useful result is that all three C33 selves produced zero-loss prices of ¥413, ¥508 and ¥597 — every one below my own ¥706–930 range, on 9–10% hurdles against my 7%, which is the only one of the four anchored on a cited ledger row, the Company's own 2.09% bond coupon F1898. My threshold is at the generous edge of the plausible set, not the defiant one. The two convergent items — C35 and C84 — are recorded as "no divergence detected" and cited as support for nothing; convergence among copies of one reasoner is not evidence. And the C89 self that would have been best placed to catch reverse-anchoring — pushing a threshold defiantly low so the gap itself becomes the argument — is one of the nine that never ran. That check is unperformed and I say so.

MUNGER. Then say plainly what a reader should do with a degraded instrument, because the temptation is to quietly not mention it. A study that hides its own broken gauge is worth less than one that shows it. Three of your six jury items are unverified; the verdict rests on the ones that are.

BUFFETT. And it does not move my number, which is the point worth making. My ¥840 is built from filed rows and one discount rate the Company itself prints F1898. No jury of any size was going to change what ¥110,603百万円 of owner earnings F15F1146F1815 is against ¥4,787,975百万円 D6. Where the crash matters is precisely where Claude says it does — in the confidence attached to a claim that no unknown blocks the verdict, which is exactly the kind of claim a second opinion exists to test.

The bet, located

Strip the agreement away and one falsifiable proposition is left. Common ground: an irreplaceable self-made freehold estate carrying first-tier licensed IP [E11][E42]; a fortress balance sheet at 67.5% equity F48 with no goodwill, no dilution and no minorities [E237][E6]D4; earnings power roughly twice reproduction asset value — the Claude lens's earnings-power value of ¥1,577,514百万円, being its ¥110,426百万円 of normalized owner earnings capitalised at its own 7%, against net operating assets of ¥770,560百万円 D83; demonstrated ability to raise price without losing volume in the latest year [E114]F373; and a management that says plainly it will spend the cash on the estate [E55]. Both sides concede all of it.

The split point, one falsifiable proposition: the FY2026 collapse in drop-through — incremental operating margin −14.7% F1018F1019F4F5, two-year 4.2% F8F10F3F5 — is a transitory cost step from an April 2025 wage revision and maintenance and system costs [E105][E114], after which the estate returns to converting revenue growth into profit at or above its 23.90% average margin D28. The bull must believe it. The bear must believe the estate now needs permanently more labour, more renewal and more capital per yen of revenue — personnel in parent cost of sales +11.3% F2180F2181, bonuses +46.1% F2176F2177, facility renewal +12.3% F2186F2187, headcount +23.2% across the window F79F83.

What settles it: two years of ΔOP ÷ Δrevenue from the 第67期 and 第68期 連結損益計算書, alongside capex versus depreciation in the same statements. The bet is operational, not terminal-value — which is precisely the finding that overturns the framing this study started with. It resolves in a document, on a date, and it requires nobody's judgement.

Prediction-vs-actual: VOID

This was an autonomous headless cycle. predictions.md carries void: no-human-prediction, and every prediction verdict is null by design — no practitioner was present at run time, and a blind call is never forged to fill the slot. No prediction-vs-actual scoring applies to this study, and it renders as void in docs/calibration.md, visibly distinct from a practitioner who declined. The five profile verdicts above still count in full for verdict accounting; only the human calibration half is skipped. The Claude lens's own figures-blind priors are a separate instrument and are scored, above and in its profile run — including the annuity over-weighting, the free-cash-flow band missed on the latest year and hit on the cumulative, and the attendance registration that could not resolve on this archive at all D18F372.

Self-distance note. The Claude lens holds one of the five verdicts compared above (pass, implied buy-below ¥700) and wrote this synthesis; it also built the reconciled figure table and the evidence ledger all five lenses consumed. That is an unusual concentration of authorship — the answerer, the ledger-builder and one of the five voters are the same system. Read the synthesis with that in mind. Three partial mitigations are on the record and should be weighed for what they are worth: the Claude lens ran figures-blind through its outside-view stage, so its priors were registered and pushed before any magnitude was visible, and the largest of them is scored here as a failure of weighting against itself; the Munger run was executed on a different model, which is the only genuine cross-model check in this study; and the lens's own jury was degraded, with nine of eighteen selves lost, which is disclosed rather than absorbed. Against those mitigations sits the fact that no consensus red team ran here, so this synthesis faced no independent adversary at all.

Verdict accounting (fixed ex-ante)

• A buy-below-¥X verdict is price-falsifiable against the unadjusted stamp. Three lenses issued explicit thresholds — Buffett ¥840, Munger ¥1,500, Pabrai ¥450 — and the Claude lens publishes an implied threshold of ¥700 with a band of ¥410–1,163. Future scoring uses the implied figure, with the band recorded.
• Li Lu issues no number. He published the floor arithmetic — about ¥294 a share, 10.1% of the stamp F2485, the hotel-segment land counted once rather than twice F1800 — and declined to issue it as a threshold, because too-hard here is a statement that the decisive variable is absent from the public record [E263][E262], not that the price is too high.
• pass / watch / too-hard are recorded but unscored in any future review. Two verdicts here are watch, two are pass, one is too-hard.
• The original verdict counts at its original stamp regardless of later corrections.
• On a stock split, reverse split, or consolidation, the buy-below threshold restates mechanically by the announced ratio (corporate-action disclosure cited); the stamp itself never restates. A 1:5 split already sits inside this study's window, effective 2023-04-01, the first day of 第64期 [E139]D122. The stamp of ¥2,920 F2485 and every threshold above are stated on the POST-split basis, matching the retro-adjusted BPS F35 and EPS F40 in the five-year highlights. The denominator used throughout is 1,639,717,407 shares D2 — 1,800,450,800 issued F527 less 160,733,393 of treasury including the ESOP and share-benefit trusts D3 — which reproduces the filing's own printed BPS of ¥670.86 to the second decimal D4F35.
• Dividend per share is not retro-adjusted in the filing and must never be quoted as printed. The restated series is ¥5.60 D12 → ¥8.00 D11 → ¥13.00 F132 → ¥14.00 F133 → ¥15.00 F134, a rise of 2.68× D13. Any future review scoring a dividend claim against this thesis scores the restated series.
• The COVID trough is inside the window and the accounting basis changes inside it. 第62期 F1F6 is an exogenous closure, not a cycle trough; 企業会計基準第27号 applies from 第65期 with 第64期 restated, leaving 第62期 and 第63期 on a different basis — a fact that carries no row ID and is stated on the authority of the ledger's own comparability-break note (figures.md, p.2 note 4 / p.4 note 7) and discrepancy log §H. Every incremental-margin and growth figure in this thesis states which window it uses, and no figure that relies on a reopening base year carries any weight in any verdict here.
• The royalty ratios are inferences and are labelled as such. ¥42,048百万円 F2131 is a parent-only cost line; 7.15% of parent net sales D86 and 5.97% of consolidated revenue D88 are computed ratios, not the contractual rate, because neither the rate nor its base is disclosed [E130][E263]. The 24.97% share of consolidated operating profit F2131F1019 carries the same caveat.
• This is a VOID study for prediction-vs-actual purposes only; the verdict accounting above is unaffected.

Red team

No consensus red team ran, and the reason is the rule, not an omission. A red team is dispatched when four or more lenses agree on a verdict class (docs/process/study.md, stage 5). Here the verdicts are two watch (Buffett, Munger), two pass (Pabrai, Claude) and one too-hard (Li Lu) — no class reaches four, so the trigger did not fire and no adversary was dispatched. There is no companies/4661-oriental-land/red-team.md, and none has been invented to fill this section.

Three things follow, and a reader is owed all of them.

First, this consensus never faced its strongest opponent, and it is weaker for it. The rule exists because a four-lens agreement from one model family is a hypothesis rather than a conclusion, and the adversary is what tests it. A three-way split is a different failure mode: it is not over-confidence, it is that the lenses genuinely disagreed about what object a verdict is. That disagreement is set out in the synthesis rather than adjudicated by anyone.

Second, the adversarial work that did happen happened inside one lens, and it was degraded. The Claude lens ran its own self-play items — C84's strongest short case, C87's internal-contradiction hunt, C93's ugly-name test, C94's archetype-gravity check — and its jury of selves was a within-lens fan-out, not an independent adversary. Nine of eighteen selves crashed, removing all corroboration from C89, C98 and C102. So the adversarial layer of this study is (a) internal to one lens and (b) half-missing, which is materially less than a dispatched red team would have supplied.

Third, the strongest available opposing case is on the record anyway, built by the lenses themselves, and it points at the long side. Every one of the three C84 selves that did return led with the incremental-margin collapse rather than the licence, including the self hired by the long side and required to rebut as hard as the evidence permits. That self successfully killed the solvency leg of the bear case — roughly ¥159bn of average annual operating cash comfortably funds the ¥332,929百万円 programme D111 — and correctly noted that declining to buy back stock at 4.35× book D7 is correct capital allocation, not a failure. It still concluded the short case wins, on the two-year 4.2% drop-through F8F10F3F5, flat attendance F373 and 8.4% cash conversion D104 bought at 39.28× earnings D8 on a 0.51% yield D10.

What a dispatched red team would most usefully have attacked, and what nobody in this study attacked hard enough, is the transitory-cost rebuttal: that an April 2025 wage revision, bonuses and maintenance and system costs [E105][E114] are a one-year step rather than a new level. That argument requires data the company declines to publish — per-guest spend, occupancy, room count and capacity all absent [E279] — and it is argued against arithmetic the company did print F1014F1015. It is the strongest thing available to the bull and it is unrebutted here only because it is unresolvable here. It resolves in the 第67期 filing.

What would change our minds

Pre-registered falsifiers, taken verbatim in substance from each profile run's falsifier: line. Future review notes score against these, not hindsight. Four of the five converge on the same document — the 第67期 有価証券報告書, expected around June 2027.

• Buffett (watch, buy below ¥840). The verdict is wrong if a filing discloses the royalty rate and its base and states that the five-year extensions to 2051-09-03 are exercisable by Oriental Land alone [E260][E262] — because the stream is then a 50-year annuity rather than a 25-year one, conservative value rises from ¥1,265 to ¥2,558 a share, and a ¥840 threshold is wrong by roughly threefold. Equally, if attendance F372 and per-guest spend D54 both rise for three consecutive years while the ordinary margin holds at or above the 25.51% of 第65期 D33, the reading that this is now price-only growth with decaying margins is wrong.
• Munger (watch, buy below ¥1,500). Any one of three reopens the case toward buy: ordinary margin resumes its climb past the 第64期 peak of 26.84% D32 for two consecutive years; the price falls into the buy-below band; or the Company (or Disney) discloses the royalty rate and its escalation mechanics [E130] and replaces at least one Keisei-linked outside officer with a demonstrably unconflicted one [E273]. Absent those, three things confirm the bear case and move this to pass: a further year of margin decline below 第66期's 24.08% D34; a materially higher parent-royalty ratio without disclosed cause D86D87D88D89; or a cost overrun or delay on the ¥330,000百万円 cruise programme [E97]D110.
• Pabrai (pass, buy below ¥450). Cumulative consolidated free cash flow on the D95–D99 definition across 第67期–第71期 exceeds ¥500,000百万円 — more than seven times the ¥66,763百万円 the five years just reported produced D104 — while net cash D71 does not fall. That would mean the ¥330,000百万円 cruise programme F522 and park reinvestment running at 1.26–1.57× depreciation D105D106 are genuinely additive rather than a treadmill, that the ~¥900 conservative intrinsic value was wrong by roughly threefold, and that a compounder at a fair price was passed over.
• Li Lu (too-hard, no threshold). Exactly two primary-source facts move this off too-hard: (a) the royalty's rate and its revenue base established from a primary source rather than inferred from the parent cost line F2131; and (b) the 2051-09-03 extension shown to be the Company's unilateral right rather than one requiring the counterparty's consent [E262]. Symmetrically, the verdict hardens from too-hard to pass at any price above the asset floor if the record ever discloses that extension requires counterparty consent, or that the royalty rate re-sets on extension [E263].
• Claude (pass, implied buy-below ¥700). Three, each thresholded with a named resolving document. (F1) If the 第67期 連結損益計算書 shows incremental operating margin above 20% against 第66期's ¥168,413百万円 F1019 and ¥704,539百万円 F5, and the 連結キャッシュ・フロー計算書 shows capex below depreciation F1146, the C13/C14 core of this pass is wrong and the verdict moves to watch. (F2) If the 第67期 filing discloses cash returned to shareholders above 50% of net income D121, the "owner comes last" finding is falsified and the verdict moves to watch. (F3) If a timely disclosure or the 重要な契約等 table extends the 2051-09-03 terminus before 2028-03-31 [E260], the terminal-value discount disappears — worth about a tenth of value on the lens's own sizing, which would not on its own flip the verdict, and the lens states that so the falsifier is honest about its own weight. Review by 2027-09-30.

The single observable four of five converge on is where the 第67期 incremental operating margin settles against the −14.7% just printed F1018F1019F4F5, together with whether capex still exceeds depreciation D105. Both resolve in one document, on a date, and neither requires anyone's judgement.

What this taught the checklists

Queued for the next study (F2 revision proposals; see docs/process/evolution.md), attributed per lens. Checklist versions froze at this study's stamp commit; these take effect next time.

The batch-level lesson, and it is the important one

The outside view over-weighted a vivid, well-documented risk and under-weighted a dull arithmetic one. The licence is legible, datable and dramatic: nine contracts, one printed terminal date, an undisclosed rate, an issuer that will not list it as a risk [E260][E130][E264]. The incremental-margin collapse is none of those things — it is one subtraction over another subtraction F1018F1019F4F5 — and it is worth more. The Claude lens registered P ≈ 0.75 on the annuity question figures-blind, wrote 1,200 words on it, and then sized it at 0.7% of market capitalisation for the modal branch and roughly a tenth of value across all of them D6, ranking it eighth on its own load-bearing list.

That is a bias with a name, and it is being written into the Claude library as a class-level caution rather than a case note, because it is not a fact about Oriental Land — it is a fact about how a well-documented structural risk competes for attention against an undocumented arithmetic one. Stated as a rule: when a qualitative structural risk dominates the outside view, the forensic stage must size it in currency before any other item is weighted; and if it cannot be sized, that inability is the finding, rather than the risk's importance being inferred from its vividness. The corollary is the diagnostic: a risk that is easy to describe and hard to quantify will systematically out-compete one that is hard to describe and easy to quantify, and the second is more often the one that decides the price.

The jury degradation, disclosed again here on purpose

Nine of eighteen Claude-lens jury selves crashed when the session's subagent budget was exhausted — the profile's own wording, with no numeric cap on the record — taking out all three selves for C89 (stamp-price anchoring), C98 (epistemic limits) and C102 (locating the bet). Under C108 a crash counts as divergence and consequences apply one-directionally, so: the implied threshold moved down ¥750 → ¥700; the published band widened to ¥410–1,163; and C98's conclusion that no unknown blocks the verdict is flagged uncorroborated. Three jury items in this study are unverified. The anchoring check that C89 exists to perform — testing whether a threshold was pushed defiantly low so that the gap itself becomes the argument — was not performed. This is stated in the profile, in the synthesis and here, because a study that hides its own degraded instrument is worth less than one that shows it. A process fix belongs upstream of any checklist: the jury fan-out should be budgeted against the session's remaining agent capacity before dispatch, and a partial dispatch should be refused rather than run.

Per-lens proposals

• Buffett — three. (1) Add a "rented moat" test to the Business economics section (proposed B27, reusing the vacant ID): where the competitive advantage rests on a licence, franchise, concession or lease from a third party, state the term, the renewal mechanic, the consideration and the consequence of non-renewal — and value the business over the disclosed term, not in perpetuity; where the consideration or the renewal mechanic is undisclosed, a perpetuity valuation is barred and the finite-term value stated instead. B23's franchise test and B29's severe-change test both assume the moat belongs to the owner; no existing item forces the analyst to price the lease. (2) Sharpen B42's maintenance-capex fallback for volume-flat operators: where reported unit volume is flat or falling while capex exceeds depreciation, the presumption is that all capex on the existing asset base is maintenance, and any lower figure must be argued. Here attendance fell 0.1% F373 on capex of 1.26× depreciation D105; without the presumption an analyst can quietly take maintenance capex at depreciation and manufacture a fifth more owner earnings. (3) Extend B44 to management's own incentive metric: ask explicitly whether any pre-capex measure is used as a remuneration index, and whether the same term carries two definitions in one document. Both occur here [E198][E274], and the second — a ¥300,000百万円 FY2029 target F256 stated without saying which definition applies — is the cleaner tell.
• Munger — three. (1) A distinct code, n/a: no comparable filed peer, separate from data-insufficient. Six items (M4, M6, M8, M20, M24, M74) require competitor or peer filings that the isolation rule structurally forbids when the subject is a geographic monopoly with no filed peer — there is no second Tokyo-area Disney-licensed operator. Without the distinction, five to six items per study read as archive failures when the true cause is market structure, and the cause will recur on every future monopoly-niche study. (2) A sub-prompt for a single committed, pre-revenue capital programme. M39 and M43 assume an established, revenue-bearing initiative when testing deferred gratification and reinvestment runway; the cruise fits neither — there is no P&L drag to point to because it is capitalised, and no demonstrated reinvestment rate to project. Proposed: state its size as a multiple of one year's operating cash flow D112 and the disclosed execution risks [E97], in place of a reinvestment-rate calculation. (3) Record the cross-reference row as a preferred citation shape. The ledger's E259–E280 rows quote several passages side by side to make a silence or a contradiction one citable object, and that turned out to be exactly the right unit for M23, M50, M54, M85 and M98, where the answer is an omission rather than a stated fact. The checklist's sourcing guidance should name it wherever an item's "Good" criterion turns on absence or contradiction.
• Pabrai — four. (1) New item — the franchise rent test: where a competitive advantage is licensed rather than owned, ask whether the rent is disclosed — rate, base, minimums, escalators, and whether it resets on renewal. P59 makes the analyst name and date the spread, and it worked: the spread is named [E42] and dated [E260]. P59 never asks what the spread costs to rent, and here that cost is one sentence [E130]. This study's central fact fell between P11, P57 and P59 and was captured by none of them. (2) Amend P1 to require the floor as a percentage of purchase price. ¥445,029百万円 of marked assets D71F943F927 is a large number and a 9.3% floor; the absolute figure read as protective and the ratio read as a fail. (3) Amend P17 to require the reinvestment-retention ratio explicitly — cumulative free cash flow ÷ cumulative operating cash flow over the disclosed window, here 8.4% D104D102. It is one division available from any five-year highlights table, and it did more work in this study than any margin, multiple or return measure. (4) Amend P56 to distinguish an exogenous shutdown from a cycle trough. 第62期 F1 is a closure, not a trough; averaging it into a normalisation understates normal earning power and excluding it discards the only stress test in the archive. Both treatments should be shown, and the item should require noting where an accounting-basis change crosses the normalisation window — here 企業会計基準第27号, which restates 第64期 and leaves 第62期 and 第63期 on the old basis (the ledger's comparability-break note and discrepancy log §H, not a citable row).
• Li Lu — four. (1) Split L34 into regulatory and private-counterparty licences. L34 is built on the cable-television example, a territorial licence whose risk is political renewal. A licence from a private counterparty is structurally different: the counterparty holds an economic interest in the licensee's profits, so the renewal risk is not withdrawal but repricing. The new limb should require, from primary sources: the rate, the base, any minimum/cap/step/escalator, who holds the extension right, and the treatment of licensed assets at expiry — with any left undisclosed priced at worst case under L5, and a licence supplying more than half of the stated moat with more than two undisclosed capping the verdict at too-hard. (2) Add an incentive-metric test to L21 or L25: does the metric that pays management reward capital deployment independently of its return? The worked example is 営業キャッシュ・フロー defined as net income plus depreciation [E198], which rises mechanically as a ¥330,000百万円 ship F519 enters service whether or not it earns. (3) Sharpen L36 with the price step. The convergence teaching is true only near book value; the item should require the explicit second step, business return ÷ price-to-book paid — here 11.7% F53 ÷ 4.35× D7 = 2.7%, which states the case in one line. (4) Give L27 a "mark unavailable" branch. Where a hidden asset is real but the archive holds no comparable transaction — 1,687,427㎡ of 1970s reclaimed land F449F457F465F473F478 carried at ¥64,538百万円 F448F456F464F472F477, about ¥38,246/㎡ — the correct answer is data-insufficient and the asset may not be counted, rather than estimated.
• Claude — four checklist sharpenings, plus two library entries. (1) C15 needs a "renewal is not growth" clause: where a named capital project replaces an existing asset in situ, classify it as maintenance regardless of the company's own classification, and state the reclassified total. The largest committed theme-park project here is ¥70,500百万円 to completely rebuild an existing attraction and its surrounding area F517 — a project the filing's own capex note lists among the year's Tokyo Disneyland drivers beside 更新改良 [E112], funded inside a stated growth-investment-first policy [E55]; the clause moved the maintenance estimate by roughly ¥10bn and the private-owner yield by ~20bp. (2) C98 should require load-bearing unknowns to be tagged by which side of the verdict they bear on — bears-on-upside / bears-on-downside / bears-on-both — with the rule that a verdict may not be pushed to too-hard by unknowns that bear only on upside. All five epistemic limits here bear on upside, and the current ranking does not surface that, which pushed §1 toward a 0.29 too-hard prior the ledger did not support. (3) C13 should mandate re-basing across at least three windows rather than leaving it to C90. Incremental margin here reads 26.1%, 4.2% and −14.7% depending on the base year, and the first would have supported the opposite verdict; where the window contains a demand shock, every shock-free window must be computed and stated, and the shock-containing window may be reported but not used. (4) C21 must not be pre-declared disabled at the outside-view stage. It was registered as disabled at source on the belief that segment assets were undisclosed; they were disclosed F1799F1800F1802, and running the item produced two load-bearing findings — the 0.59%-on-assets Other segment F1795F1802 and ¥498,191百万円 of unallocated corporate assets earning no operating return F1822. The disabled/degraded list is a §2 output, not a §1 one.
• Claude library (v0.1.0 → next), staged for the approval contract. Class-level, N=1, no rate claimed: a listed single-site destination leisure asset on owner-made freehold land, operating first-tier licensed entertainment IP, reinvesting the large majority of operating cash into its own estate under an explicit growth-investment-first policy. Oriental Land (4661) is the first member and its outcome on the class's key axis — owner cash conversion — is measurable: five-year cumulative free cash flow of ¥66,763百万円 against ¥796,674百万円 of operating cash, 8.4% D104D102, with 18.80% of net income returned in the latest year D121 and a five-year total return of 82.9 against a dividend-inclusive TOPIX of 202.2 F182F187. Open question the class should answer: for destination assets that must rebuild continuously to stand still, does the renewal share of capex reliably exceed depreciation over a full estate cycle, and does the market price that renewal as maintenance or as growth? Class-level caution, proposed as the batch's contribution: the vividness-versus-size bias stated above — filed at class level deliberately, because it is a property of how outside views are formed rather than a fact about this company. Case-level, so it never anchors a future outside view: "When a qualitative structural risk dominates the outside view, size it in currency before weighting any other item; if it cannot be sized, record the inability as the finding." Source: study #28, companies/4661-oriental-land/.

A process note that belongs in the record

The standing rule on ordinal and comparative claims fired again, and it earned its place. No claim about the record — an ordinal, a superlative, a "first" or "most" — enters a stamp, thesis or deck without being checked against docs/calibration.md and the repository first. Two comparative claims drafted for this thesis required correction before publication. The claims that survive are the checked ones and they are stated as counts rather than superlatives: twenty-eight studies scored, zero buy-below verdicts anywhere in the record; Li Lu has now returned too-hard in 19 of 28 studies, Pabrai pass in 10 of 28, and the Claude lens's pass here is its 6th of 28 — all read directly from docs/calibration.md. The lesson is the rule's, not the lenses': before asking whether a comparative claim was checked, ask which file would settle it — and if no file in the repository would, the claim must be restated as a fact about this company or deleted.

---

Corrections