Chessmaster GM Edition - personalities

dannygjk

I discovered that even when the Max depth is set to 1 I saw it sometimes go to 5 plies. I think if there is a check or big capture it keeps going until it reaches a 'quiet' position or it reaches the internal top limit of plies, (relative to the max search setting). That is common in chess programming to let the AI try to see if a move is better or worse than it seems with a shallower search. That explains why with Max search set at 3 it seems to be playing B class chess.

dannygjk

I used to use Arena is it hard to import CM the King64 to Arena? I also like to do experiments with different engines.

EscherehcsE
dannygjk wrote:

I used to use Arena is it hard to import CM the King64 to Arena? I also like to do experiments with different engines.

It takes a bit of effort to learn how to use the required WB2UCI adapter, although I think there's a procedure somewhere here that I could link to.

A larger issue is that there's no patch for the 64-bit "The King" engine. There IS a patch for the Chessmaster 10 32-bit engine, and there's a hex-edit fix for the Chessmaster Grandmaster Edition (CGE) 32-bit engine. The guy who made the hex-edit fix for the CGE 32-bit engine is on Chess.com, although he doesn't seem to post in the forums very often. Maybe we could ask him if he could do a hex-edit fix for the 64-bit version. (I tried to figure it out, but I just don't have the skilz...)

dannygjk

I'm familiar with hex editing files but of course I would need the file location and start location in the file and code and/or data. I'm just about to try it the easy way see what happens.

EscherehcsE
dannygjk wrote:

I'm familiar with hex editing files but of course I would need the file location and start location in the file and code and/or data. I'm just about to try it the easy way see what happens.

Here's the thread that discusses these issues. The guy who made the hex-edit fix is "Another-Life". The discussion starts at Post #11.

https://www.chess.com/forum/view/general/chessmaster-11-add-stockfish-or-other-uci-engines?page=1

 

dannygjk

ok ty right now my quick attempt is a semi-success it played but it moved instantly each move. I'll mess around see what i can do.

EscherehcsE
dannygjk wrote:

ok ty right now my quick attempt is a semi-success it played but it moved instantly each move. I'll mess around see what i can do.

Yeah, I'm sure that the instant moving means that the engine's operation is being blocked by the OPK code.

EscherehcsE

If you want, I could get you a copy of the CGE 32-bit hex-edit executable.

dannygjk

um not yet but ty I have a couple things i'll try first with the 64 bit version.

Nwap111

I really like chessmaster for a lot of reasons.  First it is more user friendly than Fritz.  Second, the personalities work about 1400.  So I can practice winning against say a personality that is down the exchange.  Once down material, it plays full strength, no matter the setting.  Also the Academy is really good.  Finally, I like to play it a game, lose and then see how I missed beating it.  It teaches me my weaknesses, first path to improvement.  Most of all, have fun with it.

EscherehcsE
dannygjk wrote:

I discovered that even when the Max depth is set to 1 I saw it sometimes go to 5 plies. I think if there is a check or big capture it keeps going until it reaches a 'quiet' position or it reaches the internal top limit of plies, (relative to the max search setting). That is common in chess programming to let the AI try to see if a move is better or worse than it seems with a shallower search. That explains why with Max search set at 3 it seems to be playing B class chess.

I reduced the max search depth to 2, and the engine seems to be playing around 1520 elo against Delfi 5.0 elo 1550. Then I reduced the max search depth to 1, and the engine is playing around 1170 elo. And yeah, even with max search depth set to 1, I sometimes see the engine going 4 plies deep on check extensions.

dannygjk
EscherehcsE wrote:

I ran your modified CM personality (max search depth 3, selective search 1) against Delfi 5.0 elo 1550 for several hundred games, and the CM personality had a winning expectancy of about 0.72 to 0.76, which translates to around 1720 to 1750 elo, at least based on the 1550 rating of the Delfi engine. (Bear in mind that the 1550 rating of the Delfi engine is strictly an arbitrary guess by the programmer.)

I actually ran two tournaments. One tournament was in the Arena GUI and produced the 0.76 winning expectancy; The second tournament was in the Chessmaster 10 GUI and produced the 0.72 winning expectancy. In both tournaments, both engines used the CMX opening book.

Hi, check out the TalSim personality I made:

Material/Positional      4

Selective Search          11

Own center  84

Opp center  84

Own mobility  128

Opp mobility  128

Own King safety  84

Opp King safety  120

Own passed pawns 110

Opp passed pawns 110

Own pawn weaknesses 90

Opp pawn weaknesses 90

Material (for both) Q 6.3  R 3.4  B 2.1  N 2.1  P 0.8

EscherehcsE
dannygjk wrote:
EscherehcsE wrote:

I ran your modified CM personality (max search depth 3, selective search 1) against Delfi 5.0 elo 1550 for several hundred games, and the CM personality had a winning expectancy of about 0.72 to 0.76, which translates to around 1720 to 1750 elo, at least based on the 1550 rating of the Delfi engine. (Bear in mind that the 1550 rating of the Delfi engine is strictly an arbitrary guess by the programmer.)

I actually ran two tournaments. One tournament was in the Arena GUI and produced the 0.76 winning expectancy; The second tournament was in the Chessmaster 10 GUI and produced the 0.72 winning expectancy. In both tournaments, both engines used the CMX opening book.

Hi, check out the TalSim personality I made:

Material/Positional      4

Selective Search          11

Own center  84

Opp center  84

Own mobility  128

Opp mobility  128

Own King safety  84

Opp King safety  120

Own passed pawns 110

Opp passed pawns 110

Own pawn weaknesses 90

Opp pawn weaknesses 90

Material (for both) Q 6.3  R 3.4  B 2.1  N 2.1  P 0.8

I just started an Arena tournament - Chessmaster 10 vs. Chessmaster 10 TalSim.

 

For anyone using TalSim in a non-Chessmaster GUI, here are the InitString settings (not including the opk value, which the patch should take care of) - Note that because the Material/Positional value isn't zero, some of the other InitString settings will change slightly.

opp=77 opn=68 opb=68 opr=66 opq=68 myp=77 myn=68 myb=68 myr=66 myq=68 mycc=86 mymob=131 myks=86 mypp=113 mypw=92 opcc=86 opmob=131 opks=123 oppp=113 oppw=92 cfd=0 sop=100 avd=0 rnd=0 sel=11 md=99

 

I will also note that for the Arena tournament, for both Chessmaster engines, I disabled pondering and changed the hashtable size to 128 MB. The Chessmaster 10 personality is using the CMX opening book, and the Chessmaster 10 TalSim personality is using the TalM opening book. Both books were converted to Arena format.

dannygjk

Will the patch work for CM11 for The Kingx64?

I saw that the numbers are a little different I guess converted from cm11 to cm10 due to internal setting differences?

EscherehcsE
dannygjk wrote:

Will the patch work for CM11 for The Kingx64?

I saw that the numbers are a little different I guess converted from cm11 to cm10 due to internal setting differences?

No, the patch won't work for CM11 at all, which is why we're having to approach the problem via the hex editor approach.

The different InitString numbers aren't due to CM11-to-CM10 conversion; It's because the Material/Positional value was changed from zero. When the Material/Positional value is changed from zero, it changes some of the other values for non-Chessmaster GUIs. If you're only working with a Chessmaster GUI, you don't have to worry about this problem.

dannygjk

RE: Settings

Ah, I see

RE: Hex editing

I forgot our earlier convo regarding hex editing CM11. So there is a solution enabling TheKingx64 to function properly in Arena? I'd like to be able to run test matches in Arena. I used Arena off and on years ago and I re-installed it 2-3 days ago.

Thanks,

Dan

EscherehcsE
dannygjk wrote:

RE: Settings

Ah, I see

RE: Hex editing

I forgot our earlier convo regarding hex editing CM11. So there is a solution enabling TheKingx64 to function properly in Arena? I'd like to be able to run test matches in Arena. I used Arena off and on years ago and I re-installed it 2-3 days ago.

Thanks,

Dan

Well, Another-Life made a hex-edited version of TheKingx32, and we have that. I don't know if he also tried the x64 version or not. For the x32 version, I think he said he disassembled the exe file, which I don't know how to do.

dannygjk
EscherehcsE wrote:
dannygjk wrote:

RE: Settings

Ah, I see

RE: Hex editing

I forgot our earlier convo regarding hex editing CM11. So there is a solution enabling TheKingx64 to function properly in Arena? I'd like to be able to run test matches in Arena. I used Arena off and on years ago and I re-installed it 2-3 days ago.

Thanks,

Dan

Well, Another-Life made a hex-edited version of TheKingx32, and we have that. I don't know if he also tried the x64 version or not. For the x32 version, I think he said he disassembled the exe file, which I don't know how to do.

Disassembling a program and analyzing it... been there, done that many times years ago. However you would need to learn assembly language programming first then get quite a bit of experience using it and practicing analyzing other people's code and data. That could take weeks or months. Myself I would have to get back into it and likely take weeks to get the hang of it again. Me and one of my brothers first got into BASIC and once we became pretty good at understanding programming we got into assembly. Altogether to get to the point of analyzing other people's programs understanding and modding them took quite awhile. Once we became decent at it we started cracking games to make copies, (that could take anywhere from minutes to weeks), but mostly modding them for fun to make them easier or harder once we became too good at the game.

Anyway it could be that they used the same method in CM11 but the relevant code would likely be in a different part of the program. Then the person would try searching for the group of numbers, (and or other methods), which does the thing they want to stop. As I said before maybe it would take a long time to do it.

The reason I want to do it is to be able to start a tournament then leave the computer while it runs all the games in the tourney or match. Someone told me in a thread that can't be done in CM10 and I can't find a way to do it in CM11 apparently the last version of CM that could be done was CM9000, (CM9).

I didn't mean to make my post tl;dr lol.

EscherehcsE
dannygjk wrote:

<snip>

The reason I want to do it is to be able to start a tournament then leave the computer while it runs all the games in the tourney or match. Someone told me in a thread that can't be done in CM10 and I can't find a way to do it in CM11 apparently the last version of CM that could be done was CM9000, (CM9).

 

Go for it! I'll be cheerleading from the sideline. happy.png  If you find it too difficult, you could always just use the 32-bit engine.

One more comment - Even without an OPK patch or hex-edit fix, there IS a way to have TheKing x64 run in another GUI, but it's a little labor intensive. Using the CM GUI, you have to jump through a few hoops to determine the OPK value, then you edit that value into the engine configuration for the non-CM GUI. Then it WILL run in that GUI. There's only one catch - The OPK value will change whenever you reboot your PC, so after rebooting, you'll have to determine the new OPK value if you want to run the engine again in the non-CM GUI. Once you get good at doing this, it shouldn't take more than about 5 minutes to determine a fresh OPK value.

EscherehcsE
EscherehcsE wrote:
dannygjk wrote:
EscherehcsE wrote:

I ran your modified CM personality (max search depth 3, selective search 1) against Delfi 5.0 elo 1550 for several hundred games, and the CM personality had a winning expectancy of about 0.72 to 0.76, which translates to around 1720 to 1750 elo, at least based on the 1550 rating of the Delfi engine. (Bear in mind that the 1550 rating of the Delfi engine is strictly an arbitrary guess by the programmer.)

I actually ran two tournaments. One tournament was in the Arena GUI and produced the 0.76 winning expectancy; The second tournament was in the Chessmaster 10 GUI and produced the 0.72 winning expectancy. In both tournaments, both engines used the CMX opening book.

Hi, check out the TalSim personality I made:

Material/Positional      4

Selective Search          11

Own center  84

Opp center  84

Own mobility  128

Opp mobility  128

Own King safety  84

Opp King safety  120

Own passed pawns 110

Opp passed pawns 110

Own pawn weaknesses 90

Opp pawn weaknesses 90

Material (for both) Q 6.3  R 3.4  B 2.1  N 2.1  P 0.8

I just started an Arena tournament - Chessmaster 10 vs. Chessmaster 10 TalSim.

 

For anyone using TalSim in a non-Chessmaster GUI, here are the InitString settings (not including the opk value, which the patch should take care of) - Note that because the Material/Positional value isn't zero, some of the other InitString settings will change slightly.

opp=77 opn=68 opb=68 opr=66 opq=68 myp=77 myn=68 myb=68 myr=66 myq=68 mycc=86 mymob=131 myks=86 mypp=113 mypw=92 opcc=86 opmob=131 opks=123 oppp=113 oppw=92 cfd=0 sop=100 avd=0 rnd=0 sel=11 md=99

 

I will also note that for the Arena tournament, for both Chessmaster engines, I disabled pondering and changed the hashtable size to 128 MB. The Chessmaster 10 personality is using the CMX opening book, and the Chessmaster 10 TalSim personality is using the TalM opening book. Both books were converted to Arena format.

I stopped the tournament after 500 games. The TalSim personality scored about -123 against the stock CM10 personality. So, a little less than 2500 elo (say, 2490) compared to the CM10 of 2615.