Self-play at different time controls

[lkaufman]

Thanks! The graph really explains the results.

Explanation is a bit harder. 1+1 and 2+0 are in competition for all three top engines for testing framework as strength versus time used goes. One of two:
1/ Time management of 2+0 can be improved.
2/ Use more games allowed by 2+0 compared to 1+1 for same time used, with no much strength loss.

Well, 2 + 0 did show a 22 elo loss, so your second explanation is not quite accurate. Probably the time saved is worth more than the 22 elo, but it would be nice to see this verified by a direct match betwee 2'25" + 0 and 1' + 1" since those would take about the same amount of time based on the data shown. My guess is that 2'25" would win, but not by much. Of course I would also be interested in the same test on Komodo. In any event, what this shows is that a base to increment ratio of only 60 is too small for an efficient test. We use 100, but perhaps that is not ideal either.

[Adam Hair]

Here is the plot for Komodo 8: