I've played nearly 300 chess960 games so far under Cutechess GUI, no problems at all
It is easily over 3000 Elo on that list. It is hard to compare the Elo numbers on the standard 40/4 list and the chess960 one. Different database and less engines on chess960.
Slow Chess Blitz Classic
Moderators: hgm, Rebel, chrisw
-
- Posts: 11
- Joined: Sun Aug 04, 2019 6:44 am
- Full name: Jonathan Kreuzer
Re: Slow Chess Blitz Classic
Good to hear that it is all functioning properly and playing better than expected even, thanks for testing!
I checked the FRC list, and a rating staying steady in the mid 3000-3100 range is indeed a pleasant surprise. As you mentioned must be at least partially an artifact of the list, I wouldn't expect it to be better at FRC than standard, but It should play FRC just as well which might be enough to place bit better. (Eval mostly symmetric terms designed and tuned for general case, ie. what seems logical, no special tuning for standard chess or matching for standard chess positions from actual games.)
I checked the FRC list, and a rating staying steady in the mid 3000-3100 range is indeed a pleasant surprise. As you mentioned must be at least partially an artifact of the list, I wouldn't expect it to be better at FRC than standard, but It should play FRC just as well which might be enough to place bit better. (Eval mostly symmetric terms designed and tuned for general case, ie. what seems logical, no special tuning for standard chess or matching for standard chess positions from actual games.)
-
- Posts: 3550
- Joined: Thu Jun 07, 2012 11:02 pm
Re: Slow Chess Blitz Classic
Yes I'm delighted it ran flawlessly. Chess960 testing has now finished - rating 3069. On our standard chess list I'm sure it ill be above 3000 as well, let's see.
-
- Posts: 41455
- Joined: Sun Feb 26, 2006 10:52 am
- Location: Auckland, NZ
Re: Slow Chess Blitz Classic
Probably more around 2920 on the 40/40 list.Modern Times wrote: ↑Tue Sep 24, 2019 6:23 am Yes I'm delighted it ran flawlessly. Chess960 testing has now finished - rating 3069. On our standard chess list I'm sure it ill be above 3000 as well, let's see.
gbanksnz at gmail.com
-
- Posts: 3550
- Joined: Thu Jun 07, 2012 11:02 pm
Re: Slow Chess Blitz Classic
Probably less on 40/40 yes. But should be over 3000 Elo on the blitz list. All 3 databases are different and not comparables.Graham Banks wrote: ↑Tue Sep 24, 2019 7:16 amProbably more around 2920 on the 40/40 list.Modern Times wrote: ↑Tue Sep 24, 2019 6:23 am Yes I'm delighted it ran flawlessly. Chess960 testing has now finished - rating 3069. On our standard chess list I'm sure it ill be above 3000 as well, let's see.
-
- Posts: 3550
- Joined: Thu Jun 07, 2012 11:02 pm
Re: Slow Chess Blitz Classic
Well, 2991 on the standard chess 40/4 list, not quite 3000 but tantalisingly close. And 2930 so far on 40/40 but fewer games there.
-
- Posts: 1364
- Joined: Sat Jul 21, 2018 7:43 am
- Location: Szentendre, Hungary
- Full name: Gabor Szots
Re: Slow Chess Blitz Classic
Hi Jonathan,
This refers to version 1.6. It seems to me that when evaluation display is set to white's POV (in the GUI) then it is displayed from the POV of the to-move side and vice versa under Arena, just the opposite to what I expect.
This refers to version 1.6. It seems to me that when evaluation display is set to white's POV (in the GUI) then it is displayed from the POV of the to-move side and vice versa under Arena, just the opposite to what I expect.
Gabor Szots
CCRL testing group
CCRL testing group
-
- Posts: 178
- Joined: Wed Nov 13, 2019 1:36 am
- Full name: Jonathan Kreuzer
Re: Slow Chess Blitz Classic
Gabor : Yeah looks like "Eval from White's POV" setting was reversed. I made it a separate setting for winboard and unreversed it. (engineWB.dft file or Options->Winboard Settings)
Note if anyone is wondering : I lost the password to my jonk talkchess account in a hard-drive crash, and for some reason I don't receive emails from this forum so couldn't reset it.
Slow Chess Blitz Classic 1.7 is now downloadable at https://www.3dkingdoms.com/chess/slow.htm
It scored +60 Elo in 500 self-play games against 1.6 so should be stronger. (Against the older 1.5 in FRC scored +164, not sure if that was a fluke.)
- Optimized code to be about 15% faster. (Then immediately started slowing it down again working on eval.)
- Attempted to refine eval. More work on passed pawns and king safety and center control / central space.
Some increase in threat values (hanging pieces, pawns near promotion, checks, king safety in general.)
Overall tuning / selectivity still seems hit or miss, but it's getting a bit better.
If I get time I still want to try improving the eval & trying to increase aggressiveness more, then if that works maybe a more selective search would work better. Also I want to try out Multi-threading. So I'm still hoping to make another few versions, but will only release something if +50 Elo, or add Multi-threaded search or something else of interest.
Note if anyone is wondering : I lost the password to my jonk talkchess account in a hard-drive crash, and for some reason I don't receive emails from this forum so couldn't reset it.
Slow Chess Blitz Classic 1.7 is now downloadable at https://www.3dkingdoms.com/chess/slow.htm
It scored +60 Elo in 500 self-play games against 1.6 so should be stronger. (Against the older 1.5 in FRC scored +164, not sure if that was a fluke.)
- Optimized code to be about 15% faster. (Then immediately started slowing it down again working on eval.)
- Attempted to refine eval. More work on passed pawns and king safety and center control / central space.
Some increase in threat values (hanging pieces, pawns near promotion, checks, king safety in general.)
Overall tuning / selectivity still seems hit or miss, but it's getting a bit better.
If I get time I still want to try improving the eval & trying to increase aggressiveness more, then if that works maybe a more selective search would work better. Also I want to try out Multi-threading. So I'm still hoping to make another few versions, but will only release something if +50 Elo, or add Multi-threaded search or something else of interest.
-
- Posts: 41455
- Joined: Sun Feb 26, 2006 10:52 am
- Location: Auckland, NZ
-
- Posts: 178
- Joined: Wed Nov 13, 2019 1:36 am
- Full name: Jonathan Kreuzer
Re: Slow Chess Blitz Classic
Slow Chess Blitz Classic 1.9 is released https://www.3dkingdoms.com/chess/slow.htm
I think with the addition of SMP support for the multi-core CPUs of today, this completes the modernization of Slow Chess Blitz, so it's time to take a break. (If there are obvious bugs might need a minor update release.)
CHANGES for 1.9
- Support for multi-threaded search using lazy SMP.
Search->Threads option in own GUI, threads option in UCI, plus supports winboard cores command.
Tested threads at 2, 4, 8 and gives obvious gain in playing strength (though perhaps not as much as it should.)
- Tuned evaluation using an error utility function for suites of w/l/d scored positions, so eval tuning halfway between hand-tuned and automatic.
I think it was successful and very useful, increase in strength was measurable but not large. If I had done it earlier could have been a significant improvement in less time.
And there were more positional values that increased than decreased, so if anything style is more active not less. I mainly used the ZuriChess quiet-labeled.epd, which seemed to work well.
- King Safety in general is probably even higher, probably too high at this point, but also tried to work defenders into the score more.
- Scored +40 elo to Classic 1.8 in 2000 games.
FUTURE TODO SOMEDAY
There is still a lot I didn't get done... so I'm hoping someday will be a 2.0 version that might include some of the below items and more.
- Syzygy endgame tablebase support : Noticed they're smaller than my custom bitbases even, so only possible bitbase advantage is near free probing. Plus they would be useful as oracle for tuning endgame knowledge.
- Linux Engine Version : This is in realm of possibility now, but I'm sure there are still portability issues.
- More UCI functionality : Been getting better but I still notice stuff missing that's not hard to add, eg. multiPV, searchMoves, node limit, hashfull. Or even not as easy stuff like UCI_Elo.
- Personalities support : They used to be supported but was little effort in making/testing them. Still customizeable under the hood but I was hoping to get strength closer to current-gen programs before trying to make interesting variety of play styles.
- Better Learning / Tuning : Now that I know how promising it is, build up larger suite of scored positions from own games and use this to refine evaluation values and figure out what new terms are needed (and if any terms aren't needed.)
- Improved GUI : Better move display options, including displaying Analysis values for moves on board, better piece dragging, custom chess sets, ability to show piece evaluations on board, etc.
- And of course playing strength improvement which would include better testing resources.
I think with the addition of SMP support for the multi-core CPUs of today, this completes the modernization of Slow Chess Blitz, so it's time to take a break. (If there are obvious bugs might need a minor update release.)
CHANGES for 1.9
- Support for multi-threaded search using lazy SMP.
Search->Threads option in own GUI, threads option in UCI, plus supports winboard cores command.
Tested threads at 2, 4, 8 and gives obvious gain in playing strength (though perhaps not as much as it should.)
- Tuned evaluation using an error utility function for suites of w/l/d scored positions, so eval tuning halfway between hand-tuned and automatic.
I think it was successful and very useful, increase in strength was measurable but not large. If I had done it earlier could have been a significant improvement in less time.
And there were more positional values that increased than decreased, so if anything style is more active not less. I mainly used the ZuriChess quiet-labeled.epd, which seemed to work well.
- King Safety in general is probably even higher, probably too high at this point, but also tried to work defenders into the score more.
- Scored +40 elo to Classic 1.8 in 2000 games.
FUTURE TODO SOMEDAY
There is still a lot I didn't get done... so I'm hoping someday will be a 2.0 version that might include some of the below items and more.
- Syzygy endgame tablebase support : Noticed they're smaller than my custom bitbases even, so only possible bitbase advantage is near free probing. Plus they would be useful as oracle for tuning endgame knowledge.
- Linux Engine Version : This is in realm of possibility now, but I'm sure there are still portability issues.
- More UCI functionality : Been getting better but I still notice stuff missing that's not hard to add, eg. multiPV, searchMoves, node limit, hashfull. Or even not as easy stuff like UCI_Elo.
- Personalities support : They used to be supported but was little effort in making/testing them. Still customizeable under the hood but I was hoping to get strength closer to current-gen programs before trying to make interesting variety of play styles.
- Better Learning / Tuning : Now that I know how promising it is, build up larger suite of scored positions from own games and use this to refine evaluation values and figure out what new terms are needed (and if any terms aren't needed.)
- Improved GUI : Better move display options, including displaying Analysis values for moves on board, better piece dragging, custom chess sets, ability to show piece evaluations on board, etc.
- And of course playing strength improvement which would include better testing resources.