Twisted Logic: Pros and Cons

[Marc MP]

Since Edsel asked me about my overall evaluation/review of Twisted Logic, I decided to study its games. These are noticeable Pros and Cons and the frequency of similar mistakes/similar good ideas exhibited by the engine has caught my attention, so I decided to take notes of it.
This engine has good scope and will probably make its way to the top freewares.


This analysis is based on studies of atleast 70 games, frequency of usual mistakes happening that is commonly exhibited by the engine is outlined below.


Pros:

-Early attacker. Opens up king side pawns earlier in the game sometimes to launch an attack.
- Closed/Open Position: Plays excellent aggressive style, Better prefer open position chess.
- Excellent Piece Activity at any stage of the game.
- Good understanding of Space advantage, and Open files and Diagonals
- Good understanding of some pawn structures: Especially value of connected center pawns, Pawn Chains
- Good at making long range plans in closed position.
- Attack and Defense: Pretty good attacker, good defender as well.
- Outpost and Overprotection: No problem in there.
-Openings preferable: Audy's book is pretty good anyway.

Cons:

- Doesn't fully understand the pawn structures: Underestimates passer pawns, isolated pawns.
- Plays occasional computer like incomprehensible waiting moves, but rarely.
- Little understanding of Weak and Strong squares.
- Careful when opening up the king side pawns. Doesn't care king side pawns sometimes, In normal position, Instead of capturing the white bishop on f6 with Bishop on e7, it captured the piece with g7 pawn instead thus leading to lesser concern for kind side pawns.
- Little understanding of positional chess relative to other stronger engines.
- Bishop pawn endings need improvement.
- Better time management needed especially in blitz, slow at first and faster and bolder in the end.

Test Sets: Positional Chess Test Score: 50 out of 73: 68.4%

Very kind from you Swami,

I'd guess a lot of programmers of "lesser" engines like to have feed back (please don't take "lesser" as pejorative, only less rating points) like you did.

When the program is already commercial, it is a bit more difficult to point out weaknesses (because it might hurt the sales). For sure, Twisted Logic was a good choice because it is rapidly improving and Edsel is almost always available for comments. To me it goes together: the engine is free; then users give feedback to willing programmers to improve them. I'll try to follow your example in the future (forgive me if I'm too lazy to do it, but you showed the way to go).

Have a good day,

[swami]

Since Edsel asked me about my overall evaluation/review of Twisted Logic, I decided to study its games. These are noticeable Pros and Cons and the frequency of similar mistakes/similar good ideas exhibited by the engine has caught my attention, so I decided to take notes of it.
This engine has good scope and will probably make its way to the top freewares.


This analysis is based on studies of atleast 70 games, frequency of usual mistakes happening that is commonly exhibited by the engine is outlined below.


Pros:

-Early attacker. Opens up king side pawns earlier in the game sometimes to launch an attack.
- Closed/Open Position: Plays excellent aggressive style, Better prefer open position chess.
- Excellent Piece Activity at any stage of the game.
- Good understanding of Space advantage, and Open files and Diagonals
- Good understanding of some pawn structures: Especially value of connected center pawns, Pawn Chains
- Good at making long range plans in closed position.
- Attack and Defense: Pretty good attacker, good defender as well.
- Outpost and Overprotection: No problem in there.
-Openings preferable: Audy's book is pretty good anyway.

Cons:

- Doesn't fully understand the pawn structures: Underestimates passer pawns, isolated pawns.
- Plays occasional computer like incomprehensible waiting moves, but rarely.
- Little understanding of Weak and Strong squares.
- Careful when opening up the king side pawns. Doesn't care king side pawns sometimes, In normal position, Instead of capturing the white bishop on f6 with Bishop on e7, it captured the piece with g7 pawn instead thus leading to lesser concern for kind side pawns.
- Little understanding of positional chess relative to other stronger engines.
- Bishop pawn endings need improvement.
- Better time management needed especially in blitz, slow at first and faster and bolder in the end.

Test Sets: Positional Chess Test Score: 50 out of 73: 68.4%

Very kind from you Swami,

I'd guess a lot of programmers of "lesser" engines like to have feed back (please don't take "lesser" as pejorative, only less rating points) like you did.

When the program is already commercial, it is a bit more difficult to point out weaknesses (because it might hurt the sales). For sure, Twisted Logic was a good choice because it is rapidly improving and Edsel is almost always available for comments. To me it goes together: the engine is free; then users give feedback to willing programmers to improve them. I'll try to follow your example in the future (forgive me if I'm too lazy to do it, but you showed the way to go).

Have a good day,

Thanks Marc, I find this valuable. I plan to do more of this stuff for prospective engine authors, Ofcourse any authors are welcome to send me a message to look into their engine, I'll check their engine out in the weekends when I'll have free time. ZCT, TwistedLogic, Hamsters are already in my list. I've yet to get the Learninglemming else I'd go for the authors who update their engines fairly regularly.

Yes,I'd like to see you getting into this, I'm sure you will bring in your astute observation. First step is to take note of the weakness and check if the engines make the same weak moves number of times, then you could cross check it.

I have this saved in a folder. I could spot any engine's weakness using this. It'd help if you study these strategies thoroughly. If you are already a decent player, then no problem. All Strategic concepts are present in this.

Opening Strategy:

1. Struggle for the center
2.Fast Devlopment of pieces
3.Efficient deployment of pawns
4.King Safety

Middle Game Strategy:

1.Pawn Structures
(connected pawns, Isolated, backward, Doubled, Passed pawns)
Pawn chains, Pawn cover of the king.
2. Weak and Strong Squares
3.Outpost and Overprotection
4.Open files and Diagonals.
5.Piece Activity
6.Preference of Bishop/Knight in crucial stages.
7. Kinds of Centre
(Mobile centre, Closed, Fixated and isolated pawn in the centre)
8.Attack of the king
(Defense, Counter attack, prophylaxis)

General:

1.Material advantage vs positional advantage
2.Space advantage
3.Sacrificial - too much?
4.positional exchange
5.Limiting the moility of pieces
-Limiting the mobility of pawns - blockade.
6.Positional sacrifice.
(sacrifice of pawn, piece, exchange)
7.Maneuvering in an equal positin

Endgame Strategy:

1.Activity of King
2.Role of pawns
3.Principle of two weaknesses.
EGTB takes care of Bishop, Knight Rook endings and basic stuffs

[swami]

Hi Edsel,

These are the few more analysis based on the longer time control games of TwistedLogic.

I've compared the analysis lines of Rybka and TwistedLogic
to see if there's any big Eval jump in the former in critical positions. I've taken note of the mistakes that occured in Twisted Logic games frequently. Marked "Critical" are something you might look into it more and take steps correct it immediately. Then I can assure that you will notice elo jumps

Few More... Cons:

[Critical] 1. Maneuvering in an equal position - Tries to do that but invain.

For example: Makes sacrifice even when its known in later moves that sacrifice is of no value other than give the engine slight minus score.

2.Instead of connecting Rooks to gain battery in same file , it tries to place the rook in another open file.

3. Rook pawn endings needs improvement.

4. Rarely opens up the rook to bring it to the king side fight when there is a chance(+=eval:0.30). Instead it resorts to making a computer like move (eval:0.00) and later setting two rooks up in queen side fights.

5. Doesn't understand the outpost and overprotection in some games, the engine thinks it is ok to keep the knight in the area that gives an impression that the knight acts as outpost, but unfortunately the pawn that assists the knight is isolated and under attack.

[Critical] 6. Plays too much backward moves with pieces, IE takings it pieces to its own territory in order to set up an attack.
Doesn't often think of placing its own pieces in enemies territory and gaining advantage in evaluation when it is possible.

[Critical] 7. Doesn't push its pawn to give the opponent a tempo [-0.38] and instead plays some piece backward movement [Eval:-1.54]

8. Doesn't understand pawn structures especially pawns in the center fully. Doesn't understand over protection sometimes. Needs to learn how to preserve the pieces in the center. Instead resorts to making development and ending up messing up the pawn structures.

9. Gives opponent too much tempo threat. Apparently seems to think giving tempo is always an advantage. But the theory is giving tempo may not always be right but it sometimes backfires.

10. King safety needs improvement. The engine thinks that it is ok to gain material advantage in the center to compensate the poor pawn cover of the castled king.

11. Most of the games, it opens up the king side pawns but plays in the queen side... when the king side pawns are threatened, it gives up the king side pawns or sacrifices the pawns on the queen side in order to divert attention but evaluation went down for less material on the board.

[critical] 12. Doesn't attack well with pawns on the opponents king side, it doesn't understand the pawn structures fully. The engine could have opened up opponents king side by simply pushing up its pawn instead it locks and fixes up the position with pawns and ends up in closed position which makes it hard to attack even further.
The theory is when attempting to attack the king side, always keep it in open position and not fix and lock pawns and end up in closed position.

[BubbaTough]

I've yet to get the Learninglemming

I just emailed you a version. Sorry for the delay, I do not usually compile 32 bit, and it turned out I had to do some fixes to get a 32 bit version working. Let me know via email if there are any problems ASAP, as I am going on a business trip (New York, then Italy) starting Tuesday that will take me out of regular contact for a bit.

-Sam

[swami]

I've yet to get the Learninglemming

I just emailed you a version. Sorry for the delay, I do not usually compile 32 bit, and it turned out I had to do some fixes to get a 32 bit version working. Let me know via email if there are any problems ASAP, as I am going on a business trip (New York, then Italy) starting Tuesday that will take me out of regular contact for a bit.

-Sam

Thanks, Sam. I got it and its working just fine, I have started the gauntlet 40 moves in 40 minutes repeating - tournament. When the tournament ends, I could make a list of weaknesses and strengths. Also I'd send you the commented games.

[Olivier Deville]

I've yet to get the Learninglemming

I just emailed you a version. Sorry for the delay, I do not usually compile 32 bit, and it turned out I had to do some fixes to get a 32 bit version working. Let me know via email if there are any problems ASAP, as I am going on a business trip (New York, then Italy) starting Tuesday that will take me out of regular contact for a bit.

-Sam

Hello Sam

Waiting too

It would be nice to run a small test before ChessWar real games start.

Olivier

[BubbaTough]

Waiting too

It would be nice to run a small test before ChessWar real games start.

I sent you a version maybe an hour ago to an email address off of the chesswar website, so hopefully you have it. If not let me know if not and confirm the email address you want me to send it to via message (and that you can receive .zip files, which some email clients like to protect you from).

-Sam

[Olivier Deville]

Waiting too

It would be nice to run a small test before ChessWar real games start.

I sent you a version maybe an hour ago to an email address off of the chesswar website, so hopefully you have it. If not let me know if not and confirm the email address you want me to send it to via message (and that you can receive .zip files, which some email clients like to protect you from).

-Sam

Sure. The Lemming is running now and you are even watching it live

Olivier

[swami]

I have been getting pm's from 4 members wanting to know how I tried to find out both the strength and weaknesses of the certain engine using my chess skills. I'm flattered, Thanks!


I'm currently in the process of creating a test suite that covers all the features, so please be patient for a while

As for finding out the strength of the engines and the weakness:

1.Ferret out and get the longer time controls of the certain engine. Divide games into two category: let's say

-50 winning games of Twisted Logic called Strength.pgn. (should contain 30 long battle wins, that means the number of the moves should be >70)

-50 Losing games (should contain 30 short losses in longer time controls atleast, by 'short' losses I mean, the games where the engines lost within 40-50 moves) put them in a file called "Weakness.pgn"

Search for these and put them in a separate PGN if possible.

2. Simply run Rybka as a analysis sparring partner while going through the winning or losing games.

Closely watch the Eval flop or rise after each moves. Keep track of how low or high the eval (for Rybka) fluctuates for that move. Note the move and mark it as the either the strength of the weakness. Comment it in a Pgn, and write the weakness in your own words and save it in a notepad for future use.

Ofcourse determining the weakness is far easier than determining the strength.


3. I have some test suite which covers all the possible positional chess chapters, such that 10 problems get allocated to each chapter, sub chapter as well. I may release it one day. I also test the engine using that, to find the weakness/strength etc. It's still in beta stage. Ofcourse some chess knowledge may help.

4. Collect the PGNs. Comment on them. Also write in the notepad If you notice common mistakes made by the engine in several games, especially If you *sense* it using chess skills, then mark the frequent mistakes exhibited by the engine as 'critical' and write it in your own words about what that usual mistake was that needs correction, and point the move using the PGN game number 4,6,23 etc so the author may look into the weakness.pgn for that mistake that's marked as 'critical'.

(C) GPL 2008.

[Dr.Wael Deeb]

And who accused Swami of not having a scientific approach