Re: New engine releases 2020
Posted: Tue Mar 24, 2020 6:21 pm
Many thanks for the new engine Volodymyr.
Thanksvoffka wrote: ↑Sun Mar 22, 2020 3:11 am
Igel 2.4.0 at https://github.com/vshcherbyna/igel/releases/tag/2.4.0 (binaries for Raspberry Pi, Linux, macOS and Windows).
What's new:
- Implement singlular extensions
- Detect drop of score and using time more aggressively
- Remove IID and simplify search
- Do not tt cutoff on pv moves
- Simplification for nullmove pruning
- Simplification for connected pawn evalution
- Basic endgame knowledge to detect draw in eval/search
- Retune everything + decrease number of tunnable parameters from 2600 to 1500
- Apply LMR when at least one legal move has been played
- Decrease reductions in LMR for pv moves
Relative strength compared to previous versions of Igel in 60s+0.6s time control:
Relative strength compared to Winter 0.7 BMI2 in CCRL 40/4 time control:Code: Select all
os=linux hash=256 tc=all/60+0.6 Score of Igel 2.4.0 64 POPCNT vs Igel 2.3.1 64 POPCNT: 266 - 170 - 564 [0.548] 1000 Elo difference: 33.46 +/- 14.19
Relative strength compared to RubiChess 1.6 in CCRL 40/4 time control:Code: Select all
os=windows hash=256 tc=40/102 Score of Igel 2.4.0 64 POPCNT vs Winter 0.7 BMI2: 7 - 0 - 3 [0.850] 10
Code: Select all
os=windows hash=256 tc=40/102 Score of Igel 2.4.0 64 POPCNT vs RubiChess 1.6: 5 - 2 - 4 [0.636] 11
I remember xr_a_y telling me that ponder was not working well in Igel, in his test Igel lost all games to Minic but one, so I decided to disable the ponder support until I figure out what is wrong.
http://www.cegt.net/5Plus3Rating/BestVe ... EW/51.html
I think CEGT is one of the "major competitions" resp. rating lists and we have two ponder lists.Taking into account that no major competition uses ponder and development/testing time is of the essence my main goal in Igel is to increase elo, hence now ponder is permanently disabled.
I use ponder in my tournaments and never had any trouble with Igel.voffka wrote: ↑Thu Mar 26, 2020 10:20 pm Hello Wolfgang,
I remember xr_a_y telling me that ponder was not working well in Igel, in his test Igel lost all games to Minic but one, so I decided to disable the ponder support until I figure out what is wrong.
Taking into account that no major competition uses ponder and development/testing time is of the essence my main goal in Igel is to increase elo, hence now ponder is permanently disabled.
Once Igel reaches top 10 CCRL I will think of features like ponder or chess 960, but we are not quite there yet
Sorry about that. I was actually not aware that CEGT has a ponder=on list, my fault. I will check what was wrong with ponder in Igel and try to fix it in upcoming versions.
I think it was igel-220 under cutechess. When end-game is near, Igel starts to use too much time, and eventually lost on time.Wolfgang wrote: ↑Thu Mar 26, 2020 11:24 pm I don't know, what xr_a_y tested...
As you can see from the link above, we played 800 games including 100 vs. Minic 1.35 (!). 13 wins, 56 draws, 31 losses. Nothing to see like "Igel lost all games to Minic but one". Sorry to say, but the statement of xr_a_y is nonsense. No problems here with Shredder Classic and Arena 3.5.1
Code: Select all
[Event "My Tournament"]
[Site "?"]
[Date "2020.03.27"]
[Round "4"]
[White "minic_1.35"]
[Black "igel-220"]
[Result "1-0"]
[ECO "A05"]
[GameDuration "00:01:42"]
[GameEndTime "2020-03-27T08:50:20.047 CET"]
[GameStartTime "2020-03-27T08:48:37.643 CET"]
[Opening "Reti Opening"]
[PlyCount "63"]
[Termination "time forfeit"]
[TimeControl "60+0.5"]
1. Nf3 {book} Nf6 {book} 2. d4 {book} e6 {book} 3. e3 {book} c5 {book}
4. g3 {book} Nc6 {book} 5. Nc3 {book} d5 {book} 6. Bd2 {book} a6 {book}
7. Bd3 {book} c4 {book} 8. Be2 {book} Bd6 {book} 9. b3 {book} b5 {book}
10. bxc4 {book} bxc4 {book} 11. h4 {book} O-O {book} 12. h5 {book} h6 {book}
13. Rb1 {-1.15/17 2.8s} Qc7 {+1.15/15 1.6s} 14. O-O {-1.11/16 2.7s}
Rb8 {+1.12/15 1.3s} 15. Rxb8 {-1.12/18 1.6s} Nxb8 {+0.76/16 1.9s}
16. Nh2 {-0.99/18 2.6s} Bb7 {+0.82/16 1.2s} 17. Bf3 {-0.99/20 1.4s}
Ne4 {+0.80/16 1.3s} 18. Bxe4 {-1.04/18 1.4s} dxe4 {+1.06/16 1.6s}
19. Qe2 {-0.91/19 2.0s} Nd7 {+1.12/17 2.2s} 20. Rb1 {-0.89/18 0.34s}
Nf6 {+1.01/17 1.7s} 21. Nf1 {-0.92/19 2.3s} Nh7 {+0.88/16 1.2s}
22. Bc1 {-0.72/17 1.6s} Ng5 {+1.26/19 5.2s} 23. Nd2 {-0.77/17 2.2s}
Rc8 {+1.06/16 1.0s} 24. a3 {-0.75/16 1.4s} Ba8 {+1.29/17 1.1s}
25. Kg2 {-0.91/16 2.1s} Kh7 {+1.40/19 1.6s} 26. a4 {-1.09/14 1.7s}
Qa5 {+3.87/18 1.3s} 27. Bb2 {-3.70/16 2.0s} Bb4 {+4.23/20 1.2s}
28. Rh1 {-4.45/19 1.3s} Bxc3 {+5.22/18 1.1s} 29. Nxc4 {-4.42/22 1.2s}
Rxc4 {+5.54/19 1.2s} 30. Qxc4 {-4.60/20 1.9s} Bxb2 {+6.07/20 4.0s} 31. Qc5
Qxa4 {+6.34/24 37s} 32. c4 {Black loses on time} 1-0
First alternative would be fine as I just started a new tournament for our 5+3 list with Ponder=ON. My plan was to integrate 2.4.0 in this tournament but this did not work for known reason. I could do the same with 2.4.1 and share my experience with you. Arena 3.5.1 creates good log files btw.I am currently considering either to publish a 2.4.1 (soon) with ponder as it is or wait a couple of weeks and add some elo and add ponder.