Engine Personality that plays like Kasparov (and Karpov)

[BrendanJNorman]

Hey guys,

Some of you guys know that I'm more an engine tweaker than ELO eeker, and I like to make personalities and test engines to be/that are more human-like in playing style.

I made some personalities for Rodent II, Gambit Fruit and others, usually just naming them after my website (chessncognac).

Recently, I came to the conclusion that a personality that plays like say Tal, isn't really a Tal personality if it is using something like "Perfect 2015" as the opening book because you're going to get "Tal" playing Slav Defences and Caro Kanns and other openings that weren't part of this players repertoire.

This really causes the games to lack realism.

This led me to think how interesting it would be to try to create EXACT (or as close as possible) clones of a human player (playing style, evaluation of imbalances and primary opening repertoire) and see how useful it is for analysis/training.

Tools needed are:

1. A "skeleton" engine with as many tunable features under the "hood" as possible (sadly, I'm not a programmer ).

2.An opening book tuned to play in the given players exact opening repertoire.

3. A pretty logo file for use in Arena

All I do then is create the settings, rename the "new" engine "Kasparov" for example, assign the logo file and assign the opening book.

Pretty basic, but has been consuming a lot of my time lately.

Thus far, I have fairly accurate personalities (What I'll call "GM in a box") lol, for Karpov and Kasparov, with Tal, Fischer, Morphy and Kramnik on the way.

Here are some sample games by "Kasparov" (look at the loss to Karpov too! So life-like the playing style of BOTH sides...assuming you know Kasparov and Karpov's styles well. )



[pgn]
[Event "Kasparov Tests"]
[Site "DESKTOP-FVSLDI3"]
[Date "2016.07.30"]
[Round "1"]
[White "Kasparov, Gary"]
[Black "Karpov, Anatoly"]
[Result "0-1"]
[ECO "B18"]
[EventDate "2016.??.??"]
1.e4 c6 2.d4 d5 3.Nd2 dxe4 4.Nxe4 Bf5 5.Ng3 Bg6 6.Nf3 Nd7 7.Bd3 Ngf6 8.O-O e6 9.Bxg6 hxg6 10.Bf4 Qb6 11.c4 Qxb2 12.a4 Qc3 13.Qe2 Qa5 14.Rfb1 Qa6 15.Ra2 Be7 16.Rab2 b6 17.a5 bxa5 18.h3 a4 19.Nf1 c5 20.Rd2 a3 21.Ra1 Rc8 22.d5 exd5 23.Ne3 d4 24.Rd3 Qe6 25.Ng5 Qb6 26.Nf1 a2 27.Rdd1 Kf8 28.Re1 Bd8 29.Nd2 Kg8 30.Ndf3 a5 31.Qxa2 Qc6 32.Bh2 Rh5 33.Qd2 Ra8 34.Re2 a4 35.Qf4 Ba5 36.Re7 Bc3 37.Ra3 Rb8 38.Qc1 Rb3 39.Nxf7 Bb2 40.Qc2 Rxa3 41.Ne1 Ra1 42.Qxg6 Rxe1+ 43.Rxe1 Nf8 44.Qc2 Bc3 45.Ne5 Qe6 46.Re2 N8d7 47.Nxd7 Qxd7 48.Qd3 Qf5 49.Qd1 a3 50.f3 Qg6 51.Kh1 Bb4 52.Rf2 Rh4 53.Rf1 d3 54.Qb1 Kf7 55.Qb3 d2 56.Rd1 Nh5 57.Qa4 Qe6 58.Qa7+ Kf6 59.Qa4 Rxc4 0-1
[/pgn]

I must say, I thought both players could get a pass for the "realism" although Kasparov's attack never really got off the ground.

Let's see the rematch.

[pgn]
[Event "Kasparov Tests"]
[Site "DESKTOP-FVSLDI3"]
[Date "2016.07.30"]
[Round "2"]
[White "Karpov, Anatoly"]
[Black "Kasparov, Gary"]
[Result "0-1"]
[ECO "D82"]
[EventDate "2016.??.??"]

1.d4 Nf6 2.c4 g6 3.Nc3 d5 4.Bf4 Bg7 5.e3 c5 6.dxc5 Qa5 7.Rc1 dxc4 8.Bxc4 O-O 9.Nf3 Qxc5 10.Qe2 Bg4 11.O-O Bxf3 12.gxf3 Nc6 13.Ne4 Qf5 14.Bd3 Nd5 15.Bg3 Ne5 16.Bxe5 Qxe5 17.f4 Qe6 18.Nc5 Qd6 19.Be4 Nc3 20.bxc3 Qxc5 21.Bxb7 Rab8 22.Be4 a5 23.c4 e6 24.Rc2 a4 25.Rd1 Bf6 26.Bf3 a3 27.Qd3 Rb2 28.Rdc1 Kg7 29.Qe2 Rfb8 30.Be4 Bd8 31.Qd3 Ba5 32.Qe2 Kg8 33.Kg2 Kf8 34.Bf3 Kg7 35.Kg1 R2b4 36.h3 Rb2 37.Be4 R8b6 38.Kf1 Kf8 39.Bf3 Kg8 40.Kg2 Rb8 41.Be4 Kf8 42.Kg1 Kg7 43.Kh1 R8b4 44.Kh2 Rb6 45.h4 h5 46.Bd3 Rb8 47.Kh3 Bd8 48.f3 Qe7 49.Qf2 Qd6 50.Rd1 Bf6 51.Qe2 Qc5 52.Rdc1 Be7 53.Be4 Qa7 54.Rd2 Bc5 55.Rc3 Rxd2 56.Qxd2 Qa5 57.Qc1 Rb2 58.Rc2 Rxc2 59.Qxc2 Bxe3 60.Qh2 Qe1 61.f5 exf5 62.Bd5 Bb6 63.Qg3 Bf2 64.Qc7 Qh1+ 65.Qh2 Qf1+ 66.Qg2 Qe1 67.Qh2 Bxh4 68.Qc7 Be7 69.Qg3 Qf1+ 70.Qg2 Qc1 71.Qf2 Qb2 72.Qa7 Qe2 73.f4 h4 74.Qd4+ Bf6 75.Qa7 Qg4+ 76.Kh2 Qxf4+ 77.Kg2 h3+ 0-1
[/pgn]

The following game really impressed me from Kasparov.

[pgn]
[Event "Kasparov Tests"]
[Site "DESKTOP-FVSLDI3"]
[Date "2016.07.30"]
[Round "2"]
[White "Karpov 1, Anatoly"]
[Black "Kasparov, Gary"]
[Result "1/2-1/2"]
[ECO "D34"]
[EventDate "2016.??.??"]
1.d4 d5 2.c4 e6 3.Nf3 c5 4.cxd5 exd5 5.g3 Nc6 6.Nc3 Nf6 7.Bg2 Be7 8.O-O O-O 9.dxc5 d4 10.Na4 b5 11.cxb6 axb6 12.Nxd4 Nxd4 13.Bxa8 Ba6 14.Nc3 Nxe2+ 15.Nxe2 Qxa8 16.Re1 Rd8 17.Qa4 Nd7 18.Nc3 Nc5 19.Qg4 Bc8 20.Qh5 Bf8 21.Bg5 f6 22.Be3 Qc6 23.Red1 Bb7 24.f3 Re8 25.Bxc5 Re5 26.Qg4 Qxc5+ 27.Kg2 Re3 28.Rf1 h5 29.Qg6 Qb4 30.Qc2 Qg4 31.Qb3+ Kh7 32.Qd1 Qf5 33.Rc1 Bc5 34.h4 Ba6 35.a4 Bc8 36.Qc2 Rd3 37.Kh1 Qh3+ 38.Qh2 Qf5 39.Qg2 Bb7 40.Ne4 Rxf3 41.Rxf3 Qxe4 42.Rcc3 Bd4 43.Rcd3 Bxb2 44.a5 Be5 45.axb6 Bxg3 46.Kg1 Be5 47.Qf2 Bc6 48.Rh3 Qb4 49.Kf1 Bb5 50.Qf3 Kh6 51.Qf2 g6 52.Kg2 Kg7 53.Rb3 Qe4+ 54.Rhf3 Bc6 55.Kf1 Bd5 56.Rfe3 Bc4+ 57.Ke1 Qh1+ 58.Kd2 Qd5+ 59.Rbd3 Qb5 60.Rc3 Qb2+ 61.Rc2 Qd4+ 62.Ke1 Qa1+ 63.Kd2 Qd4+ 64.Ke1 Qa1+ 65.Kd2 Qd4+ 1/2-1/2[/pgn]

Two engines that I LOVE (Thinker and Wasp) were absolutely owned by Kasparov.

[pgn]
[Event "Kasparov Tests"]
[Site "DESKTOP-FVSLDI3"]
[Date "2016.07.30"]
[Round "1"]
[White "Kasparov, Gary"]
[Black "Wasp 1.02"]
[Result "1-0"]
[BlackElo "2800"]
[ECO "B90"]
[EventDate "2016.??.??"]

1.e4 c5 2.Nf3 d6 3.d4 cxd4 4.Nxd4 Nf6 5.Nc3 a6 6.Be3 Ng4 7.Bg5 h6 8.Bh4 g5 9.Bg3 Bg7 10.h3 Ne5 11.Nf5 Bxf5 12.exf5 Nbc6 13.Bd3 Nxd3+ 14.Qxd3 Qa5 15.O-O Rc8 16.Nd5 Bxb2 17.Rab1 Bd4 18.c4 Rb8 19.Rfe1 Qc5 20.Re4 Be5 21.Bxe5 Nxe5 22.Qe2 f6 23.Rb6 Kf8 24.Qb2 Qc8 25.Rxe5 fxe5 26.f6 e6 27.Qb4 Kf7 28.Qxd6 Re8 29.Nc7 Kxf6 30.Rb3 Kg6 31.Nxe8 Qxe8 32.Qxe5 Rd8 33.Rxb7 Rd1+ 34.Kh2 Rd7 35.Qe4+ Kf6 36.Rb6 Re7 37.c5 Qd7 38.Rd6 Qc7 39.Qe3 a5 40.f4 gxf4 41.Qxf4+ Kg7 42.c6 Qb8 43.Qd4+ Kf7 44.Qd3 Kf6 45.Kh1 Qb2 46.Qf3+ Kg7 47.Qg3+ Kf7 48.Rd1 Qg7 49.Qxg7+ Kxg7 50.Rd7 Kf6 51.c7 Rxd7 52.c8=Q 1-0[/pgn]

[pgn]
[Event "Kasparov Tests"]
[Site "DESKTOP-FVSLDI3"]
[Date "2016.07.30"]
[Round "1"]
[White "Kasparov, Gary"]
[Black "Thinker 5.3B Inert"]
[Result "1-0"]
[BlackElo "2960"]
[ECO "C10"]
[EventDate "2016.??.??"]

1.e4 e6 2.d4 d5 3.Nc3 dxe4 4.Nxe4 Nd7 5.Nf3 Ngf6 6.Bd3 Be7 7.Qe2 c5 8.Nxf6+ Bxf6 9.c3 cxd4 10.cxd4 Nb8 11.O-O Nc6 12.Rd1 O-O 13.Qe4 g6 14.Bh6 Bg7 15.Bxg7 Kxg7 16.Rac1 Qb6 17.Qf4 Qxb2 18.h4 Nb4 19.h5 Kg8 20.Ne5 Nd5 21.Qh6 Nf6 22.Rc7 Bd7 23.Nxd7 Nxh5 ( 23...Nxd7 24.hxg6 hxg6 25.Bxg6 fxg6 26.Qxg6+ Kh8 27.Qh6+ Kg8 28.Qg5+ Kh8 29.Qh4+ Kg8 30.Rxd7 $18 ) 24.Rc5 Rfd8 25.Rxh5 Rxd7 26.Qxh7+ Kf8 27.Qh6+ Ke8 28.Bxg6 Qe2 29.Rc1 Qg4 30.Qh8+ Ke7 31.Qxa8 Qf4 32.Qc8 Qf6 33.Bxf7 a5 34.Qe8+ Kd6 35.Bxe6 Qe7 36.Qg8 Qxe6 37.Qf8+ 1-0[/pgn]

I think this Kasparov personality (with the book telling him to play his "real" opening repertoire ) is really on the right track, although it seems MUCH too strong for ME to train against right now.

Might have to dumb it down a bit haha.

What do you guys think?

I was even more impressed by the Karpov personality beating Kaspy with his beloved Caro Kann and making Kasparov's attack look so ridiculous.

This has been really fun.

Once all of the wrinkles are out of these "GMs in a box" I'll let you guys have a try.

Lots of fun ahead.

Brendan

P.S Kaspy also lost to Toga Checkov and Hiarcs 10, but beat ALL the Thinkers, Deep Junior 10, Ktulu 9, Frenzee, Prodeo 2.0, Bright 0.5c, King 3.50 and others.

[Ferdy]

Hey guys,

Some of you guys know that I'm more an engine tweaker than ELO eeker, and I like to make personalities and test engines to be/that are more human-like in playing style.

I made some personalities for Rodent II, Gambit Fruit and others, usually just naming them after my website (chessncognac).

Recently, I came to the conclusion that a personality that plays like say Tal, isn't really a Tal personality if it is using something like "Perfect 2015" as the opening book because you're going to get "Tal" playing Slav Defences and Caro Kanns and other openings that weren't part of this players repertoire.


This really causes the games to lack realism.

This led me to think how interesting it would be to try to create EXACT (or as close as possible) clones of a human player (playing style, evaluation of imbalances and primary opening repertoire) and see how useful it is for analysis/training.

Tools needed are:

1. A "skeleton" engine with as many tunable features under the "hood" as possible (sadly, I'm not a programmer ).

2.An opening book tuned to play in the given players exact opening repertoire.

3. A pretty logo file for use in Arena

All I do then is create the settings, rename the "new" engine "Kasparov" for example, assign the logo file and assign the opening book.

Pretty basic, but has been consuming a lot of my time lately.

Thus far, I have fairly accurate personalities (What I'll call "GM in a box") lol, for Karpov and Kasparov, with Tal, Fischer, Morphy and Kramnik on the way.

In the past I tried creating GM personalities, by tuning the engine eval and search parameters to the move of the player you are trying to create. For example, to create Fischer personality I take some of the games of Fischer and extract positions where he is to move and also save his move in that position. Say you have 800 games x 20 positions/game = 16000 positions. Adjust the Engine Parameters call it EP1 then let EP1 solve the 16000 positions, It will score 1 if its bestmove is the move in that given position, if not it will get 0 points. Record the score of EP1. Then create a new engine setting call it EP2, then let it solve the 16000 positions, record the score and compare it with the score of EP1, if EP1 has more score points, the bestEP is EP1, create again a new engine setting call it EP3, do the same and compare its score with the bestEP, if EP3 has more score points, then the bestEP is EP3. Do it more times depending on the number of parameter combination you can create, the engine setting that will get more points take the Fisher personality.
I remember the engine with personality loses to my default engine setting. There are settings that gets more points where in that setting the null move pruning technique was disabled.

There are positions where Fischer blundered, if the blunder is big I remove that position from the test set. So part of this exercise is to blunder check all those positions, but this is fast, analyzing those positions at 1sec per move could be enough because you are only after the big blunders like more than 1 pawn blunder will be removed.

All of this are done automatic, the script will create engine settings, run the test sets, save the score of that particular setting, then create new setting again and so on.

If you are interested, I will create a personality with this method and send it to you to test against your created personality.

My engine is deuterium and plan to tune the following parameters. Perhaps I will try Anand preferably with more games or depending on what personality you want me to create.

Code: Select all

option name KingSafetyPercent type spin default 100 min 0 max 1000
option name MobilityPercentMg type spin default 100 min 0 max 1000
option name MobilityPercentEg type spin default 100 min 0 max 1000
option name OffensivePercent type spin default 100 min 100 max 10000
option name DefensivePercent type spin default 100 min 100 max 10000
option name PawnValueMg type spin default 89 min 70 max 110
option name PawnValueEg type spin default 125 min 90 max 150
option name KnightValueMg type spin default 325 min 300 max 350
option name KnightValueEg type spin default 315 min 300 max 350
option name BishopValueMg type spin default 328 min 300 max 350
option name BishopValueEg type spin default 327 min 300 max 350
option name RookValueMg type spin default 504 min 480 max 530
option name RookValueEg type spin default 498 min 480 max 530
option name QueenValueMg type spin default 996 min 950 max 1200
option name QueenValueEg type spin default 999 min 950 max 1200
option name PawnThreat type spin default 100 min 0 max 1000
option name KnightThreat type spin default 100 min 0 max 1000
option name BishopThreat type spin default 100 min 0 max 1000
option name RookThreat type spin default 100 min 0 max 1000
option name QueenThreat type spin default 100 min 0 max 1000
option name KingThreat type spin default 100 min 0 max 1000
option name ThreatOnUndefended type spin default 100 min 0 max 1000
option name ThreatOnDefended type spin default 28 min 0 max 1000
option name Rook7thRankMg type spin default 15 min 0 max 200
option name Rook7thRankEg type spin default 50 min 0 max 200
option name Rook8thRankMg type spin default 15 min 0 max 200
option name Rook8thRankEg type spin default 50 min 0 max 200
option name KnightOutpostPercent type spin default 100 min 0 max 1000
option name TwoBishopAdv type spin default 50 min 20 max 80
option name DoubledPawnPenalty type spin default 10 min 0 max 100
option name IsolatedDoubledPawnPenalty type spin default 12 min 0 max 100
option name IsolatedPawnPenaltyPercent type spin default 100 min 0 max 1000
option name PawnDuo type spin default 1 min 0 max 8
option name ConnectedPasserPercent type spin default 16 min 0 max 100
option name BlockedPasserPenaltyPercent type spin default 70 min 0 max 100

I will not tune the search parameters, just the evaluation, perhaps I will add more.

[velmarin]

Perhaps you already know Szint, if you like to investigate...
http://chessprograms.260mb.com/?ckattempt=1

Immortal Forum with thread Szint.:
http://immortalchess.net/forum/showthread.php?t=15642

[Ferdy]

Perhaps you already know Szint, if you like to investigate...
http://chessprograms.260mb.com/?ckattempt=1

Immortal Forum with thread Szint.:
http://immortalchess.net/forum/showthread.php?t=15642

I don't know about it. Thanks for pointing this out.

[PK]

It would be good to determine what parameters are useful for style tuning. I have good experiences with asymmetric mobility and asymmetric king attack bonus. On the other hand, Rodent does not separate midgame and endgame piece values, but on the other hand it uses little bonuses for keeping a given type of pieces on the board. What do You think is useful and what is not?

[BrendanJNorman]

Ferdy, that's a very scientific method you use haha.

Definitely send me your Fischer settings and I'll throw him in a match with Kasparov. I'l' even have him use the Fischer book I made (Najdorf and KID/Grunfeld as black, 1.e4 and Fischer's main 1.e4 repertoire)

When I make personalities it is generally more "intuitive" and from 13 odd years of studying the players themselves.

So I have a nice knowledge of most of the World Champions habits, styles and opening repertoires and try to apply this within the restrictions I have ("under the hood" UCI tuning options), then...

1. I make a "test" version and put it in a match.
2. The moment I see a move which doesn't "feel" right (like a "Kasparov move" for example), I go "under the hood" and tweak something.
3.Restart match.
4. I have reasonable feel for this and a reasonable playing strength myself so this can be a good process for me. It tends to work well without me having programming knowledge etc.


velmarin,

Yep I have used Szint a lot and particularly liked the Karpov personality (its use of "good knight vs bad bishop" imbalances is very Karpov like), but I find the Szint GM personalities to be too much alike in their evaluation (the same goes for the Prodeo GM personalities, except Tal...).

They might differ a little from each other, but they still leave traces behind and lack the uniqueness that the true GMs opinion might have.

For example, in positions where Kasparov and Karpov might be COMPLETELY in disagreement about who they prefer in real life (perhaps some English Opening where black sacrifices an exchange), but Szint Kasparov AND Szint Karpov will both show in the +0.34 range for example.

I try to make the personality almost an exaggeration of the GMs style (just like an actor in a movie might exaggerate Al Capone's personality for dramatic effect) and don't care too much if strength is affected (unless its ridiculous).

For me, priority is 1.STYLE>2.strength.

Pawel,

Here are my ideas about parameters for style tuning.

Please correct me if one of these things already exist whether directly or indirectly (a weight causes this idea to happen automatically).

What you already have in Rodent II are awesome and what I'd love to see (again as a non-programmer) are:

-Weight for Two Bishops in middlegame/ending

- Weight for opening the position when one HAS the two bishops (even if its via a pawn sac, or creating a pawn break somewhere. Remember the old Kasparov games in the Queens Indian with Bishops on b2 and d3? He'd play d5!, sacrificing the central pawns to open the bishop's diagonals and begin a kingside attack. This is only an obvious example though, there might be nice ideas an engine, especially Rodent could come up with, with this weight.)

-Weight for keeping the position CLOSED when the opponent has the bishops

-Weight for centralising the King early when queens are off the board (even in queenless middlegames with lots of pieces still around). This should lead to moves like Ke2 instead of 0-0 (something Russian GMs understand very well) and early Kf1-e2 as we see a lot in Karpov and Capablanca games. Not sure If I explained that one well, but Im sure you know what I mean.

-"Favour" exchange sacrifice (over move that is = in eval), this might be similar to one of Ed's Rebel settings or Junior's "favour sacrifices" setting. This would be good for creating Topalov/Petrosian personalities (Topalov's other settings would be aggressively tuned, Petrosian's defensively) and such types of play.

- Weight for having d+e pawns in centre stably
-Weight for having e+f pawns advanced two squares stably might be interesting for aggressive personalities.

I could think of ideas all day haha. I wish I could write an engine!

[Ferdy]

Ferdy, that's a very scientific method you use haha.

Definitely send me your Fischer settings and I'll throw him in a match with Kasparov. I'l' even have him use the Fischer book I made (Najdorf and KID/Grunfeld as black, 1.e4 and Fischer's main 1.e4 repertoire)

When I make personalities it is generally more "intuitive" and from 13 odd years of studying the players themselves.

So I have a nice knowledge of most of the World Champions habits, styles and opening repertoires and try to apply this within the restrictions I have ("under the hood" UCI tuning options), then...

1. I make a "test" version and put it in a match.
2. The moment I see a move which doesn't "feel" right (like a "Kasparov move" for example), I go "under the hood" and tweak something.
3.Restart match.
4. I have reasonable feel for this and a reasonable playing strength myself so this can be a good process for me. It tends to work well without me having programming knowledge etc.

I used to study and observed a players's style before mostly Kasparov and Karpov and some local GM's Torre and Antonio.
It is interesting to know your method of tuning. It would be nice to examine the games between these personalities.
All right I will prepare it, probably 2 days from I will send it.

[PeterO]

Hello!

I am very iterested in testing your engine too! Please send me engines (especially the FISCHER! )

Peter

[PK]

Main problem with adding more parameters is that Rodent already has 30 of them. It could easily be 50, but then they wouldn't show in some GUIs and their number would be discouraging for casual users. I have a grand plan of creating an online tool for defining Rodent personalities, but it will take some time.

What you already have in Rodent II are awesome and what I'd love to see (again as a non-programmer) are:

-Weight for Two Bishops in middlegame/ending

This will be added, Rodent already has it as an internal weight.

- Weight for opening the position when one HAS the two bishops (even if its via a pawn sac, or creating a pawn break somewhere. Remember the old Kasparov games in the Queens Indian with Bishops on b2 and d3? He'd play d5!, sacrificing the central pawns to open the bishop's diagonals and begin a kingside attack. This is only an obvious example though, there might be nice ideas an engine, especially Rodent could come up with, with this weight.)

I'm not sure what you mean and how it is different from the ordinary bishop pair bonus. But this sentence inspired me to add a micro bonus for two bishops standing next to each other.

-Weight for keeping the position CLOSED when the opponent has the bishops

Most of the time when the opponent has the two bishops, the engine has a bishop+knight or two knights. In that event side without a bishop pair gets a bonus defined by parameter KnightLikesClosed, which increases with the number of pawns.

-Weight for centralising the King early when queens are off the board (even in queenless middlegames with lots of pieces still around). This should lead to moves like Ke2 instead of 0-0 (something Russian GMs understand very well) and early Kf1-e2 as we see a lot in Karpov and Capablanca games. Not sure If I explained that one well, but Im sure you know what I mean.

This is interesting, but non-trivial.

-"Favour" exchange sacrifice (over move that is = in eval), this might be similar to one of Ed's Rebel settings or Junior's "favour sacrifices" setting. This would be good for creating Topalov/Petrosian personalities (Topalov's other settings would be aggressively tuned, Petrosian's defensively) and such types of play.

Again, Rodent has an internal parameter for discouraging exchange sacrifices, all I need is to expose it to the user.

- Weight for having d+e pawns in centre stably
-Weight for having e+f pawns advanced two squares stably might be interesting for aggressive personalities.

Perhaps this can be reduced to one weight called "Center control". Will think about it.

[S.Taylor]

Fisher is already in the grave, as everyone will be one day. What a person does when he is young, he does with great enthusiasm, but then he himself doesn't like what he did so much, OR he just loses energy and has hardly any connection with his original energy and interests.
So i don't see the great thing in making GM personalities.
If Tal were alive today, his chess would be worth almost nothing, because computers can do exactly what he could do and better.