After some thought, I decided that this would violate my rules.nczempin wrote:So perhaps even for uMax it would be "legal" (cost of 0 characters) to keep them out of the move generation, but allow it to be a legal move (without resigning).
The claim is that micro-Max is the smallest Chess _program_ (or that with best Elo per character, depending on which version we are talking about). Not just the smallest subroutine that comes up with moves.
So the character counts must include thuse of an interface that allows you to play chess with it, that can read, parse and execute user moves, and that somehow informs the user about the computer move. And preferably set up a position. It can be a minimal interface, but it has to be there, and the characters of it must be fully included in the count.
To solve the under-promotion problem in that context (by accepting it as an input move) takes 32 characters. It would than still be surprised by it, and never play one himself.
The fact that there also exists a Winboard-compatible version of uMax, is only a concession to user-friendliness. The WB versions of uMax have no official status, and their characters are not counted. The only guarantee is that, under certain time control settings (in particular fixed-time per move setting) it plays exactly the same moves the original stand-alone uMax would play, when playing on an otherwise unloaded computer. That ratings are in practice determined with other time controls(e.g. 40/40') and not at the type of time control the stand-alone uMax plays I don't really consider as a problem. As long as the opponents have to play the same time control, I do not expect the relative ratings to be dependent on such tiny details as playing 40/40' or 120/120'. So I assume the ratings as, e.g. CCRL determines them would equally apply to the stand-alone version of uMax.
So the 1433 characters of uMax 1.6 include the input line for under-promotions, and the Winboard version of it will not resign if one is played. OTOH, the 1961 characters of uMax 4.8 do not include an input line for under-promotions in its interface, so the WB version of it does resign on an underpromotion. And like I explained, it does it a bit too liberally, as even delaying the resign to give uMax the opportunity to capture it sometimes leads to resigns where the stand-alone uMax would have simply won the game (capturing the promoted piece after a check).