Safe or Attack Style
Posted: Tue Dec 22, 2015 3:49 pm
Here is an interesting aberation that was seen during testing. Two virtually identical programs can produce the following results on average when matched (win-loss-draw).
(1) 30-30-40 prune turned on
(2) 40-40-20 prune turned off
The only difference is whether a quiescent futility SEE prune is used. The results show exactly the same theoretical strength, but the percentage of draws is much higher for the first example. This means the program plays safer. The second example shows the engine plays much riskier, but ends up with the same results. This is an old argument in chess. Is it better to play an attacking or a safe style in chess? Which style would you prefer to see a chess engine play?
(1) 30-30-40 prune turned on
(2) 40-40-20 prune turned off
The only difference is whether a quiescent futility SEE prune is used. The results show exactly the same theoretical strength, but the percentage of draws is much higher for the first example. This means the program plays safer. The second example shows the engine plays much riskier, but ends up with the same results. This is an old argument in chess. Is it better to play an attacking or a safe style in chess? Which style would you prefer to see a chess engine play?