Working out the alternates

[angrypirate]

I think if you make an 'alternate' move, that Chesstempo should then continue and make you play it out.....dunno if this is possible or would create too many headaches...but if there are alternate lines, I think it would be cool to be able to get a 'correct' score by solving an alternate, and not be forced to go back and find another winning move. Great site, keep up the good work-- I love this place.

[revenant]

That would definitely increase the training value of the site.  Even better would be if the computer would randomly play one of the 2 or 3 humanlike responses that are usually available after any given move, approximating tournament play more closely since even when you're presented with a problem you've seen before, you can't count on being able to recite a rote winning sequence to get your rating points.

Maybe in the future (like, next year?), Chess Tempo will have evolved to the point where it stores a whole tree of moves for both white and black for each problem.  That would probably entail converting the database and code from a problem-centered approach to a position-centered approach, an idea that has received some tentative support from Richard.  For now, I'm guessing he wants to keep a single linear sequence the user is required to follow for each problem as it keeps the user interface simple and avoids the coding headaches associated with "exploding tree branches".  Richard?

[richard]

The exploding tree is the main issue in allowing alternative lines to be followed. Alternative lines also have their own alternative further along , which have their own alternatives etc etc.  Given a finite amount of CPU resources I have to decide what is the best choice on CPU time.  Allowing alternatives would cause a very large increase in CPU required for every problem produced.  For now I think CPU time is probably better spend on producing new problems (hopefully some of which will be helping to fill the relative void of very hard problems at the top end of the rating range).  At the moment aside from generating new problems I'd prefer to be spending CPU on providing evaluations for a) the common mistake moves where the mistake was not in the best 4 moves and b) evaluations for the opponents move so users can see what the different lines and evaluations were for the opponent.  Allowing users to contribute CPU time to problem generator is a longer term goal that might allow alternative lines to be played out, but for now there isn't enough CPU cycles to go around.

Regards,
Richard.