Time management ideas

Code, algorithms, languages, construction...
Post Reply
lucasart
Posts: 201
Joined: Mon Dec 17, 2012 1:09 pm
Contact:

Time management ideas

Post by lucasart » Mon Apr 03, 2017 4:59 am

Time management is a surprisingly difficult topic.

Here are a couple of perfectly logical ideas that failed in testing. Not only they don't gain elo, but the results indicate a significant probability of regression:
  • Playing recaptures faster: use half the allocated time as long as the best move has a strictly positive SEE. Half should be safe enough, you would think…
  • Playing forced moves immediately: when there's only one legal move, play it immediately. Perhaps hash pressure is needed to make this work: i'll give it shot. Point is searching is not useless as it fills the HT, but if HT trashing becomes sufficient, perhaps we can measure sth…
Does anyone have ideas that actually work in testing ?

No hand waving / arguing on a theoretical leval about "should" work, please.
"Talk is cheap. Show me the code." -- Linus Torvalds.

H.G.Muller
Posts: 190
Joined: Sun Jul 14, 2013 10:00 am
Real Name: H.G. Muller

Re: Time management ideas

Post by H.G.Muller » Mon Apr 03, 2017 10:30 am

Have you tested this with ponder on? Filling the hash table might not be useless, but it might be smarter to do it while the opponent's clock is running than in your own time. (While allowing the opponent to fill his hash table on a guaranteed ponder hit.)

In HaQiKi D instant response in 'only-move' positions was helpful. But that was Xiangqi, where at some point the opponent starts to perpetually check you all the time (with only one legal evasion) at nearly every move. In Chess such positions seem too rare to expect any measurable impact.

In Shokidoki it had a positive effect to reduce the search time by a factor 10 for singular moves (i.e. when all moves lose a piece compared to the best).

I guess it would also depend on how you get your ponder move after an instant reply.

Post Reply