Stockfish 2.2.2 JA update available

PawnStormZ · Post by **PawnStormZ** » Fri Jan 20, 2012 5:15 pm

Hi Jim.

Thanks for all your work with the compiles! I am testing the "Intel sse 4.2 build" version of Stockfish in a game-in 10 secods match. So far no time losses, so it looks like the problem has been solved.

Take care

pohl4711 · Post by **pohl4711** » Sat Jan 21, 2012 10:16 am

pohl4711 wrote:
The Intel compiles (native Windows threads) include the time fix from 2.2.1.
OK, found it. Will try it this weekend and report it here.

Stefan

I can report, that the 2.2.2. JA 64bit Intel-Compile (no SSE) works fine on my system (only 2 time losses in now 570 bulletgames (60 seconds+500ms Fischerbonus)).

It seems to be that this version is about 5-10 Elo stronger than the 2.2.1-compile. I will test it for my NEBB-Rankinglists and will replace the older 2.2.1 with the nice new 2.2.2 Version, because it is only a bugfix-version (but a good one!)

Stefan

PawnStormZ · Post by **PawnStormZ** » Sun Jan 22, 2012 7:07 am

In an 800 game match: game in 10 seconds plus 1 second per move, the Intel sse 4.2 Stockfish compile lost 8 of its 536 games (1.5%) on time. Neither of the other two engines (Ivanhoe, RobboLito) lost a game on time. I guess there is still some issue in Stockfish.

Games here... http://www.datafilehost.com/download-dccf8772.html

geots · Post by **geots** » Sun Jan 22, 2012 10:47 am

PawnStormZ wrote: In an 800 game match: game in 10 seconds plus 1 second per move, the Intel sse 4.2 Stockfish compile lost 8 of its 536 games (1.5%) on time. Neither of the other two engines (Ivanhoe, RobboLito) lost a game on time. I guess there is still some issue in Stockfish.

Games here... http://www.datafilehost.com/download-dccf8772.html

It does not matter which incremental control you use- you can always take a calculator and come up with a repeating time control, ie 40/3 , 40/4 , 40/8 40/10 repeating, etc. that would basically give the engine the same amount of seconds on average for each move.

I know of 3 Ivanhoe versions that hate it, don't understand it, and will lose on time in a high percentage of games. And they are strong versions you really don't want to discard. So I now play a repeating control comparable to an incremental one. I will NEVER again play even one game at those controls. Fine if you aren't getting losses, but do you expect someone to fix the problem in the Ivanhoe's I mention? Won't happen-ever. And some programmers are not going to spend an inordinate amount of time on a problem you can easily circumvent.

Everyone has different ideas about it, but I don't call it a bug or an issue if I can get the same results a different way with not more than 1 time loss in 1000 games- if even 1.

Best,

george