Forums

Lucas Chess engine settings are very weak

Sort:
lkmk16
lkmk16
Oct 25, 2017
0
#21

The program's website has changed, are the problems of creating a free software, now is:

http://lucaschess.pythonanywhere.com/

 

This website I don't know how long it will work, maybe years, I don't know, it's more stable the blog:

http://lucaschess.blogspot.com.es/

EscherehcsE
EscherehcsE
Closed: Inactive
Oct 25, 2017
0
#22
lkmk16 wrote:

Engines with 2100 of fixed elo have a handicap set by the author of the engine, normally the number of positions analyzed per second is limited.
Lucas Chess only sends the command setoption name UCIElo value 2100 to the engine.
When depth 3 is set, another limitation is added, the engine is instructed to analyze only to depth 3 in each movement.
There are two limitations, and only the most restrictive has effect.

To play against a 2100, it is preferable to remove the depth option (-- as indicated by EscherehcsE), and play for time (there is a tab to indicate it).

And the biggest problem with engines is finding out with less than 2000 of it and running properly, almost all engines have an elo superior to 2100, of course it is playing without limitations.
In the next version 11, there will be 33 engines with an elo greater than 2500, and 9 with an elo greater than 3000. I don't think it's necessary any more, perhaps they are more interesting those that have a minor elo, for the majority of the users of the program.

And thank you for your interest in the program.

Hi Lucas, thanks for responding. However, I'm still trying to figure out this issue. I selected an "Engine with fixed elo", specifically McBrain 2.1a (elo 2100). I set the depth to "--", and I left the time as "0.0". (The time control for the game was 6/2.) After doing this, the engine appears to be running at full strength. It beat engines rated in the mid 2500s. I even tried the same thing with McBrain 2.1a (1200 elo), and I swear, it still seemed to be running at full strength. Is there something I'm not understanding about these settings?

EscherehcsE
EscherehcsE
Closed: Inactive
Oct 25, 2017
0
#23

One quick observation - I tried low elo settings for Rybka and Rodent, and they seemed to work OK. Maybe the problem is only with the McBrain engine?

lkmk16
lkmk16
Oct 25, 2017
0
#24

Version 11 of LC works with McBrain 2.7, and I believe that this appart is updated.

The fixed elo is controlled exclusively by the engine, so the problem may be in McBrain 2.1.

EscherehcsE
EscherehcsE
Closed: Inactive
Oct 25, 2017
0
#25
lkmk16 wrote:

Version 11 of LC works with McBrain 2.7, and I believe that this appart is updated.

The fixed elo is controlled exclusively by the engine, so the problem may be in McBrain 2.1.

OK, thanks. I'll wait for LC ver 11 to come out, then I'll check McBrain 2.7.

Debistro
Debistro
Oct 26, 2017
0
#26
lkmk16 wrote:

Engines with 2100 of fixed elo have a handicap set by the author of the engine, normally the number of positions analyzed per second is limited.
Lucas Chess only sends the command setoption name UCIElo value 2100 to the engine.
When depth 3 is set, another limitation is added, the engine is instructed to analyze only to depth 3 in each movement.
There are two limitations, and only the most restrictive has effect.

To play against a 2100, it is preferable to remove the depth option (-- as indicated by EscherehcsE), and play for time (there is a tab to indicate it).

And the biggest problem with engines is finding out with less than 2000 of it and running properly, almost all engines have an elo superior to 2100, of course it is playing without limitations.
In the next version 11, there will be 33 engines with an elo greater than 2500, and 9 with an elo greater than 3000. I don't think it's necessary any more, perhaps they are more interesting those that have a minor elo, for the majority of the users of the program.

And thank you for your interest in the program.

Hi Lucas,

Thanks for responding here. And I do like Lucas Chess, I think I will be using it more from now on.

Regarding setting the Depth, I still don't see how to remove the depth option?

Anyway, all the engines there are great with their varying levels...my suggestion is to set their default settings so that they perform according to their stated Elos. Right now, the default of 3 ply is too weak, according to all my testing. 

And how do I get the engines not to move instantly? In spite of increasing the depth, they still move instantly.

EscherehcsE
EscherehcsE
Closed: Inactive
Oct 26, 2017
0
#27
Debistro wrote:

Regarding setting the Depth, I still don't see how to remove the depth option?

Just click your mouse cursor inside the depth box. You should get a pick list to pop up with values from 1 to 30, and the "--" choice at the very top of the list.

EscherehcsE
EscherehcsE
Closed: Inactive
Oct 26, 2017
0
#28
Debistro wrote:

And how do I get the engines not to move instantly? In spite of increasing the depth, they still move instantly.

Maybe lkmk16 will have a better answer, but that's not right - The engines shouldn't be moving instantly if you have the ply level set to high values. At least I'm not having that problem.

EscherehcsE
EscherehcsE
Closed: Inactive
Oct 26, 2017
0
#29
EscherehcsE wrote:
Debistro wrote:

And how do I get the engines not to move instantly? In spite of increasing the depth, they still move instantly.

Maybe lkmk16 will have a better answer, but that's not right - The engines shouldn't be moving instantly if you have the ply level set to high values. At least I'm not having that problem.

Ya know, I think I might have the answer. I just tried setting Stockfish set to a depth of 15. I think that Stockfish is just so fast and powerful that it doesn't take very long to search 15 plies deep. It's only a fraction of a second, and it looks almost instantaneous. However, the analysis window shows that Stockfish is indeed looking 15 plies deep.

Edit - Try increasing the depth to 20 and you should start to notice a definite delay in Stockfish's moves.

SeniorPatzer
SeniorPatzer
Oct 26, 2017
0
#30

If you beat a computer engine that's rated at 2100 USCF, is there a high correlation that you are playing at a USCF Expert level?  

EscherehcsE
EscherehcsE
Closed: Inactive
Oct 26, 2017
0
#31
SeniorPatzer wrote:

If you beat a computer engine that's rated at 2100 USCF, is there a high correlation that you are playing at a USCF Expert level?  

I'm not sure if there's a consistent method of comparing computer ratings with USCF ratings. I know years ago, the USCF allowed some dedicated units to enter USCF OTB tournaments, but that was long ago, and that hasn't been done in recent years.

I know that many of the dedicated units would advertise a rating for the machine, but the number was often wildly optimistic, if not an outright lie.

I think the only way to get any semi-accurate USCF rating for an engine would be for a number of USCF-rated players to play a large number of games against the engine under controlled tournament conditions.

Debistro
Debistro
Oct 26, 2017
0
#32
EscherehcsE wrote:
EscherehcsE wrote:
Debistro wrote:

And how do I get the engines not to move instantly? In spite of increasing the depth, they still move instantly.

Maybe lkmk16 will have a better answer, but that's not right - The engines shouldn't be moving instantly if you have the ply level set to high values. At least I'm not having that problem.

Ya know, I think I might have the answer. I just tried setting Stockfish set to a depth of 15. I think that Stockfish is just so fast and powerful that it doesn't take very long to search 15 plies deep. It's only a fraction of a second, and it looks almost instantaneous. However, the analysis window shows that Stockfish is indeed looking 15 plies deep.

Edit - Try increasing the depth to 20 and you should start to notice a definite delay in Stockfish's moves.

Ok, now I am getting the opposite effect though. I fiddled with the Time and Set the Depth to ---, and also changed the Depth to 15. Now setting the Time to something specific while Depth is removed, means the computer engines do start thinking and playing normally.

But setting the Depth to 15 ply now....and they lose on time; no longer moving instantly.

I just played a game between Simplex at almost 2400 vs Vanessa 2300, and although Simplex was winning despite playing the opening like a dufus (minus any books), but eventually lost on time.

And setting the Depth at 20 or more, means they will start thinking for a very long time for each move, and ignore their own time control.

I think the optimal Depth should perhaps be 7-9? The conclusion is that the default settings of the engines straight out of the box (after downloading and installing Lucas Chess) are still far from optimal.

Hope this can be improved in later versions, with perhaps, more control options over the engines, including a mode where they just analyze while we make moves on the board for both colors - My suggestion.

EscherehcsE
EscherehcsE
Closed: Inactive
Oct 26, 2017
0
#33
Debistro wrote:
EscherehcsE wrote:
EscherehcsE wrote:
Debistro wrote:

And how do I get the engines not to move instantly? In spite of increasing the depth, they still move instantly.

Maybe lkmk16 will have a better answer, but that's not right - The engines shouldn't be moving instantly if you have the ply level set to high values. At least I'm not having that problem.

Ya know, I think I might have the answer. I just tried setting Stockfish set to a depth of 15. I think that Stockfish is just so fast and powerful that it doesn't take very long to search 15 plies deep. It's only a fraction of a second, and it looks almost instantaneous. However, the analysis window shows that Stockfish is indeed looking 15 plies deep.

Edit - Try increasing the depth to 20 and you should start to notice a definite delay in Stockfish's moves.

Ok, now I am getting the opposite effect though. I fiddled with the Time and Set the Depth to ---, and also changed the Depth to 15. Now setting the Time to something specific while Depth is removed, means the computer engines do start thinking and playing normally.

But setting the Depth to 15 ply now....and they lose on time; no longer moving instantly.

I just played a game between Simplex at almost 2400 vs Vanessa 2300, and although Simplex was winning despite playing the opening like a dufus (minus any books), but eventually lost on time.

And setting the Depth at 20 or more, means they will start thinking for a very long time for each move, and ignore their own time control.

I think the optimal Depth should perhaps be 7-9? The conclusion is that the default settings of the engines straight out of the box (after downloading and installing Lucas Chess) are still far from optimal.

Hope this can be improved in later versions, with perhaps, more control options over the engines, including a mode where they just analyze while we make moves on the board for both colors - My suggestion.

OK, I still don't think you're doing it right. Assuming that you're using the "Play against an engine of your choice" mode, and assuming you've selected a time control for the game, let's say, 6/2:

Let's say you want to play against the full-strength Stockfish. You would pick the "Stockfish 8 32-bit (3300)" from the internal engines list, set the engine depth to "--", and leave the time set to "0.0".

Because you left the time set to "0.0", Stockfish will allocate the time spent for each move as needed to not lose on time. I think if you were to specify a cetain time, like 60 seconds, it would on average use 60 seconds per move, even if the result is that it loses on time.

EscherehcsE
EscherehcsE
Closed: Inactive
Oct 26, 2017
0
#34

@Debistro, since you like the CM Vanessa personality so much, I worked up a WB2UCI version of CM10 Vanessa (complete with CMX polyglot opening book) that I imported into Lucas Chess. I've started a match between Simplex 0.9.8 (with GM opening book) and Vanessa to see how strong Vanessa really is. It's 200 games, time control of 6 min/2 sec. I'll let you know the results later. Smile

Debistro
Debistro
Oct 26, 2017
0
#35
EscherehcsE wrote:
Debistro wrote:
EscherehcsE wrote:
EscherehcsE wrote:
Debistro wrote:

And how do I get the engines not to move instantly? In spite of increasing the depth, they still move instantly.

Maybe lkmk16 will have a better answer, but that's not right - The engines shouldn't be moving instantly if you have the ply level set to high values. At least I'm not having that problem.

Ya know, I think I might have the answer. I just tried setting Stockfish set to a depth of 15. I think that Stockfish is just so fast and powerful that it doesn't take very long to search 15 plies deep. It's only a fraction of a second, and it looks almost instantaneous. However, the analysis window shows that Stockfish is indeed looking 15 plies deep.

Edit - Try increasing the depth to 20 and you should start to notice a definite delay in Stockfish's moves.

Ok, now I am getting the opposite effect though. I fiddled with the Time and Set the Depth to ---, and also changed the Depth to 15. Now setting the Time to something specific while Depth is removed, means the computer engines do start thinking and playing normally.

But setting the Depth to 15 ply now....and they lose on time; no longer moving instantly.

I just played a game between Simplex at almost 2400 vs Vanessa 2300, and although Simplex was winning despite playing the opening like a dufus (minus any books), but eventually lost on time.

And setting the Depth at 20 or more, means they will start thinking for a very long time for each move, and ignore their own time control.

I think the optimal Depth should perhaps be 7-9? The conclusion is that the default settings of the engines straight out of the box (after downloading and installing Lucas Chess) are still far from optimal.

Hope this can be improved in later versions, with perhaps, more control options over the engines, including a mode where they just analyze while we make moves on the board for both colors - My suggestion.

OK, I still don't think you're doing it right. Assuming that you're using the "Play against an engine of your choice" mode, and assuming you've selected a time control for the game, let's say, 6/2:

Let's say you want to play against the full-strength Stockfish. You would pick the "Stockfish 8 32-bit (3300)" from the internal engines list, set the engine depth to "--", and leave the time set to "0.0".

Because you left the time set to "0.0", Stockfish will allocate the time spent for each move as needed to not lose on time. I think if you were to specify a cetain time, like 60 seconds, it would on average use 60 seconds per move, even if the result is that it loses on time.

But my set up is/was like that...

Under the "Time" tab, is where I set a certain time frame for the game, which I thought would apply to the engine as well. But it does not adhere to the time.

Debistro
Debistro
Oct 26, 2017
0
#36
EscherehcsE wrote:

@Debistro, since you like the CM Vanessa personality so much, I worked up a WB2UCI version of CM10 Vanessa (complete with CMX polyglot opening book) that I imported into Lucas Chess. I've started a match between Simplex 0.9.8 (with GM opening book) and Vanessa to see how strong Vanessa really is. It's 200 games, time control of 6 min/2 sec. I'll let you know the results later.

Haha, Simplex should win most of them. It was winning in my latest game test, but failed to stick to the clock. Maybe try the test with Gaia 3.5, which is a bit lower in Elo and closer to Vanessa. Smile

EscherehcsE
EscherehcsE
Closed: Inactive
Oct 26, 2017
0
#37
Debistro wrote:
EscherehcsE wrote:
Debistro wrote:
EscherehcsE wrote:
EscherehcsE wrote:
Debistro wrote:

And how do I get the engines not to move instantly? In spite of increasing the depth, they still move instantly.

Maybe lkmk16 will have a better answer, but that's not right - The engines shouldn't be moving instantly if you have the ply level set to high values. At least I'm not having that problem.

Ya know, I think I might have the answer. I just tried setting Stockfish set to a depth of 15. I think that Stockfish is just so fast and powerful that it doesn't take very long to search 15 plies deep. It's only a fraction of a second, and it looks almost instantaneous. However, the analysis window shows that Stockfish is indeed looking 15 plies deep.

Edit - Try increasing the depth to 20 and you should start to notice a definite delay in Stockfish's moves.

Ok, now I am getting the opposite effect though. I fiddled with the Time and Set the Depth to ---, and also changed the Depth to 15. Now setting the Time to something specific while Depth is removed, means the computer engines do start thinking and playing normally.

But setting the Depth to 15 ply now....and they lose on time; no longer moving instantly.

I just played a game between Simplex at almost 2400 vs Vanessa 2300, and although Simplex was winning despite playing the opening like a dufus (minus any books), but eventually lost on time.

And setting the Depth at 20 or more, means they will start thinking for a very long time for each move, and ignore their own time control.

I think the optimal Depth should perhaps be 7-9? The conclusion is that the default settings of the engines straight out of the box (after downloading and installing Lucas Chess) are still far from optimal.

Hope this can be improved in later versions, with perhaps, more control options over the engines, including a mode where they just analyze while we make moves on the board for both colors - My suggestion.

OK, I still don't think you're doing it right. Assuming that you're using the "Play against an engine of your choice" mode, and assuming you've selected a time control for the game, let's say, 6/2:

Let's say you want to play against the full-strength Stockfish. You would pick the "Stockfish 8 32-bit (3300)" from the internal engines list, set the engine depth to "--", and leave the time set to "0.0".

Because you left the time set to "0.0", Stockfish will allocate the time spent for each move as needed to not lose on time. I think if you were to specify a cetain time, like 60 seconds, it would on average use 60 seconds per move, even if the result is that it loses on time.

But my set up is/was like that...

Under the "Time" tab, is where I set a certain time frame for the game, which I thought would apply to the engine as well. But it does not adhere to the time.

 

Yes, the "Time" tab is where you set the time control for the game. But you specified a depth of 12 ply for Simplex. I think you should have specified "--" for the depth. Forcing Simplex to go 12 plies deep in a short time control like 6/2 might be your problem.

EscherehcsE
EscherehcsE
Closed: Inactive
Oct 28, 2017
0
#38
Debistro wrote:
EscherehcsE wrote:

@Debistro, since you like the CM Vanessa personality so much, I worked up a WB2UCI version of CM10 Vanessa (complete with CMX polyglot opening book) that I imported into Lucas Chess. I've started a match between Simplex 0.9.8 (with GM opening book) and Vanessa to see how strong Vanessa really is. It's 200 games, time control of 6 min/2 sec. I'll let you know the results later.

Haha, Simplex should win most of them. It was winning in my latest game test, but failed to stick to the clock. Maybe try the test with Gaia 3.5, which is a bit lower in Elo and closer to Vanessa.

Yeah, Simplex was putting the beatdown on Vanessa (roughly 85%). I set my sights lower, but I had to switch to my Arena GUI. The problem I was having with Lucas Chess was, when using the WB2UCI adapter with Vanessa, the LC GUI wasn't "hard terminating" Vanessa's engine process after each game. So every time a new game was started, LC would create a new engine process but fail to delete the old one. The engine processes would build up and consume RAM resources, so I would have to go into my task manager and manually delete the extra engine processes. (I guess I could have paused the tournament and rebooted the PC, but that's no fun, either.) I tried adding a "TerminateHard = true" command to the WB2UCI adapter, but LC ignored it.

 

Anyway, Vanessa is playing against both Clarabit 1.00 32-bit (approx 2058 elo?) and Gibbon 2.01b (2015 elo) in Arena. Not quite 60 total games played yet (20 games between any two engines) with a time control of "40 moves in 4 minutes, repeating". Vanessa is playing the engines "even, or just below even". I'll need lots more games to get any semi-accurate result.

 

Edit - Eh, I decided to start a new tournament and dump Gibbon 2.01b, since it didn't have very many games played on CCRL, and I'm not sure how accurate the posted rating is. I'll add Matheus 2.3 in its place. (I kind of wanted to try out Matheus anyway.)

Debistro
Debistro
Oct 29, 2017
0
#39
EscherehcsE wrote:
Debistro wrote:
EscherehcsE wrote:

@Debistro, since you like the CM Vanessa personality so much, I worked up a WB2UCI version of CM10 Vanessa (complete with CMX polyglot opening book) that I imported into Lucas Chess. I've started a match between Simplex 0.9.8 (with GM opening book) and Vanessa to see how strong Vanessa really is. It's 200 games, time control of 6 min/2 sec. I'll let you know the results later.

Haha, Simplex should win most of them. It was winning in my latest game test, but failed to stick to the clock. Maybe try the test with Gaia 3.5, which is a bit lower in Elo and closer to Vanessa.

Yeah, Simplex was putting the beatdown on Vanessa (roughly 85%). I set my sights lower, but I had to switch to my Arena GUI. The problem I was having with Lucas Chess was, when using the WB2UCI adapter with Vanessa, the LC GUI wasn't "hard terminating" Vanessa's engine process after each game. So every time a new game was started, LC would create a new engine process but fail to delete the old one. The engine processes would build up and consume RAM resources, so I would have to go into my task manager and manually delete the extra engine processes. (I guess I could have paused the tournament and rebooted the PC, but that's no fun, either.) I tried adding a "TerminateHard = true" command to the WB2UCI adapter, but LC ignored it.

 

Anyway, Vanessa is playing against both Clarabit 1.00 32-bit (approx 2058 elo?) and Gibbon 2.01b (2015 elo) in Arena. Not quite 60 total games played yet (20 games between any two engines) with a time control of "40 moves in 4 minutes, repeating". Vanessa is playing the engines "even, or just below even". I'll need lots more games to get any semi-accurate result.

 

Edit - Eh, I decided to start a new tournament and dump Gibbon 2.01b, since it didn't have very many games played on CCRL, and I'm not sure how accurate the posted rating is. I'll add Matheus 2.3 in its place. (I kind of wanted to try out Matheus anyway.)

Thanks mate. I'm really curious to get some idea of Vanessa's strength, because I think in CM Grandmaster Edition, she is even stronger than previous versions. I don't know enough of how to mod or edit the engines but it must be quite cool to do what you're doing.

EscherehcsE
EscherehcsE
Closed: Inactive
Oct 29, 2017
0
#40
Debistro wrote:
 

Thanks mate. I'm really curious to get some idea of Vanessa's strength, because I think in CM Grandmaster Edition, she is even stronger than previous versions. I don't know enough of how to mod or edit the engines but it must be quite cool to do what you're doing.

I don't have CM GE, but I hear the full-strength engine for CM GE is maybe about 50 elo stronger than CM10th. (64-bit vs 32-bit, plus maybe other minor coding improvements?).

So, assuming the handicap settings for Vanessa haven't changed between the two versions, the newer Vanessa can't be more than 50 elo stronger than the CM10 Vanessa, and maybe somewhat less.