Two days after publishing our first Deep Value issue we finished building a harder research system, and the first thing it did was mark our own work. NEOWIZ went through it again, this time verified line by line against the company's own filings. Four things were wrong. Two of them make the company cheaper than we told you. Two of them break the argument we made for it. The score falls from 22 out of 25 to 16, and the rating drops from Conviction Buy to Buy. Everything below is the corrected version, with the old numbers shown next to the real ones.
On July 25 we adopted a stricter research standard that requires every thesis-critical number to be checked against the company's own filing before publication, rather than taken from free data providers. We ran the debut issue through it. It did not pass in its original form.
The honest summary. The company is cheaper than we told you. Once the balance sheet is properly adjusted you are paying under two times operating profit at the current price. But the reason we said it would re-rate has already happened and did not work, and the reason we said it was ignored was not true. The arithmetic was never the problem. The argument was.
We have left the original entry price of ₩17,930 in place for track-record purposes. Under our published rules an entry stands at the closing price on the day it was first published, whatever we learn afterwards. The loss on this position, if it becomes one, belongs to us.
About 69% of Korean listed companies still trade below book value. For decades this "Korea discount" was a trap, because family-controlled groups sat on cash they never paid out and small shareholders had no way to make them. That has genuinely changed. The government's Value-Up programme pushed companies to return cash, and a 2026 amendment to Korea's Commercial Act now forces companies to cancel the shares they buy back, within 12 months for new buybacks and 18 months for older ones. Handing cash back is no longer optional once a buyback is announced.
That much is real and we stand behind it. What we got wrong was assuming the reform automatically means a lot of cash coming back soon at any given company. It does not. It means that when a company chooses to buy back, the shares must be retired. The size of the return is still set by the company. At NEOWIZ the policy pays out the greater of 20% of operating profit or ₩10 billion a year. On this year's earnings that floor is likely to bind, which is a return of roughly 3.7% of the cash pile per year. That is a slow drip, not a forcing mechanism, and we should have said so the first time.
NEOWIZ is a Korean game developer and publisher. Its standout asset is Lies of P, a Pinocchio-themed "Soulslike" action game launched in September 2023 that has sold more than 4 million copies. A prequel expansion, Overture, came out in mid-2025, and a full sequel is in development, though at an earlier stage than we implied: the company's own pipeline chart puts it at "vertical slice", which is the stage after a prototype and well before a finished game. Around that sit the long-running DJMAX rhythm series, the mobile title Brown Dust 2, the Cat & Soup franchise, the Pmang casual games portal, and a publishing arm for outside studios.
In the first quarter of 2026 revenue split 51% mobile, 39% PC and console, 10% other, and 51% overseas. Revenue grew 13.9% year on year. Operating profit was ₩6.95 billion, a 6.9% margin, down sharply from the strong quarters of 2025. The company describes this openly as a gap between releases. That is the honest state of the business: capital-light, genuinely global, and currently between hits.
This is the section we got most wrong, so it is worth being precise about what replaced it.
What we said: NEOWIZ is cheap because nobody has bothered to read the Korean filings.
What is true: eight analysts cover this company. They have a Buy consensus and an average twelve-month price target of ₩27,750, roughly 64% above the current price. The company publishes a full quarterly investor deck and runs an English-language investor site. Domestic Korean brokers write on it regularly. And the stock has still fallen from ₩32,000 in February to under ₩17,000 in July.
That is not a neglected company. That is a company the market has examined and decided to keep discounting anyway. The difference matters more than it might sound. A neglected stock re-rates when someone finally looks. A considered discount only closes when the reason for it goes away, and here the reason is visible: earnings are lumpy, the hit is ageing, the release gap is real, and the cash comes back slowly.
There is still a genuine structural argument, and it is the one this letter should have made. Coverage in Korean, by Korean brokers, for Korean institutions does not put a global bid under a ₩365 billion KOSDAQ stock. Global small-cap value funds do not screen this exchange by default, and at roughly $250 million it is too small for most of them to own meaningfully anyway. So the buyer who would normally arbitrage a two times earnings multiple is structurally absent. That is a real reason for a discount to persist. It is a much weaker reason to expect it to close.
Who is selling, and are they stupid? This is the weakest part of the case and we should be straight about it. We cannot identify a forced or indifferent seller. There is no index deletion, no fund liquidation, no tax-loss window driving this. The stock has fallen steadily from ₩32,000 since February on ordinary volume, which looks like considered selling by informed domestic holders reacting to the earnings gap. When the seller is informed rather than forced, the burden of proof sits with the buyer.
Enter your Cantillon Research subscriber email to unlock the full corrected NEOWIZ issue: the verification log checked line by line against the company's filings, the net cash bridge, the three-scenario valuation with price targets and expected return, the revised catalyst timeline, the full flag register, the IVT institutional-flow read with the invalidation level, the computed position size, and the tear sheet.
Publishing an idea starts the coverage, it does not end it. For this name: the annual report work to close the restricted cash, auditor and related-party lines, published by 15 August. A verified peer comparison to follow. An earnings note within 72 hours of the second-quarter results. An update on every checkpoint date whether or not anything happened. A timing alert within 24 hours if the gate moves either way. And an exit note if the thesis breaks, published whether or not it lost money.
Every position is tracked in the open in the Deep Value Portfolio, entered at the closing price on the day it was first published, under rules set in advance and not changed afterwards.