I wrote a little program mostly as a visual to see what squares are "hot". Meaning if you watched a game or series of games, what places are commonly considered strong positions to control.
The images pretty much summed up common knowledge (center square). But personally was a little intrigued to find some variations. I pushed, Fischer's pgn database against it and found it the hottest spot was d4, and while a lot of GM's had the center 4 squares as their hottest, Kasparov for example had d4 e4 even moreso than the rest.
I personally find it useful for seeing trends, but can also be used on a per game instance. Just feed it t a PGN file.
http://olympuschess.com/strength/
Curious what you all think, and any input is appreciated.
-Josh
Positional Strength/Most Controlled territory
Moderators: hgm, Rebel, chrisw
-
- Posts: 1342
- Joined: Wed Mar 08, 2006 9:41 pm
- Location: Morgantown, WV, USA
-
- Posts: 12541
- Joined: Wed Mar 08, 2006 8:57 pm
- Location: Redmond, WA USA
Re: Positional Strength/Most Controlled territory
The majority of FICS players are not strong. I suspect that the dominance of squares could possibly be different if only very strong players were involved.jshriver wrote:I wrote a little program mostly as a visual to see what squares are "hot". Meaning if you watched a game or series of games, what places are commonly considered strong positions to control.
The images pretty much summed up common knowledge (center square). But personally was a little intrigued to find some variations. I pushed, Fischer's pgn database against it and found it the hottest spot was d4, and while a lot of GM's had the center 4 squares as their hottest, Kasparov for example had d4 e4 even moreso than the rest.
I personally find it useful for seeing trends, but can also be used on a per game instance. Just feed it t a PGN file.
http://olympuschess.com/strength/
Curious what you all think, and any input is appreciated.
-Josh
The thing that intrigues me the most is the coloration of c3, c6, f3, and f6.
I would be very curious to see what the image looks like if you filter Marcel's database for both players at or over 2300 Elo in strength, and also what it looks like if you use this database:
http://cap.connx.com/chess-engines/new- ... db.pgn.bz2
-
- Posts: 1342
- Joined: Wed Mar 08, 2006 9:41 pm
- Location: Morgantown, WV, USA
Re: Positional Strength/Most Controlled territory
Actually I forgot to mention that the bottom image is actually filtered using only 2100 or higher for either B/W.
I have the full nearly 100gig pgn dataset here Nov 99 till Dec 2010. The original program is a perl script so I can run it locally if you'd like to see other variations. 2500 > 2300 > fics set.
Downloading hqdb.pgn.bz2 now to run it against my local version. Will add it to the page if you'd like.
I have the full nearly 100gig pgn dataset here Nov 99 till Dec 2010. The original program is a perl script so I can run it locally if you'd like to see other variations. 2500 > 2300 > fics set.
Downloading hqdb.pgn.bz2 now to run it against my local version. Will add it to the page if you'd like.
-
- Posts: 12541
- Joined: Wed Mar 08, 2006 8:57 pm
- Location: Redmond, WA USA
Re: Positional Strength/Most Controlled territory
I would like to see 2300+, 2400+, 2500+ and also the PGN file I posted a link to processed. It is 3.4 GB of high quality chess games.
I would also like to see the numerical data for the squares, along with the colors.
I would also like to see the numerical data for the squares, along with the colors.
-
- Posts: 1342
- Joined: Wed Mar 08, 2006 9:41 pm
- Location: Morgantown, WV, USA
Re: Positional Strength/Most Controlled territory
Will do, not sure if I can get them done tonight a little late but definitely this weekend.
-Josh
-Josh
-
- Posts: 12541
- Joined: Wed Mar 08, 2006 8:57 pm
- Location: Redmond, WA USA
Re: Positional Strength/Most Controlled territory
Thanks! I am looking forward to the results.jshriver wrote:Will do, not sure if I can get them done tonight a little late but definitely this weekend.
-Josh
Are you willing to share your processing script? The idea is simple enough to write it from scratch, but I am lazy as a bag of hay.
-
- Posts: 6401
- Joined: Thu Mar 09, 2006 8:30 pm
- Location: Chicago, Illinois, USA
Re: Positional Strength/Most Controlled territory
I think it is because of the Knights. d4,e4,d5,e5 are obvious, but f3 and the like are the only squares that attack two of the center squares. I think it is interesting the difference between h2 and h7. They are the only clearly "asymmetric" colored squares.Dann Corbit wrote:The majority of FICS players are not strong. I suspect that the dominance of squares could possibly be different if only very strong players were involved.jshriver wrote:I wrote a little program mostly as a visual to see what squares are "hot". Meaning if you watched a game or series of games, what places are commonly considered strong positions to control.
The images pretty much summed up common knowledge (center square). But personally was a little intrigued to find some variations. I pushed, Fischer's pgn database against it and found it the hottest spot was d4, and while a lot of GM's had the center 4 squares as their hottest, Kasparov for example had d4 e4 even moreso than the rest.
I personally find it useful for seeing trends, but can also be used on a per game instance. Just feed it t a PGN file.
http://olympuschess.com/strength/
Curious what you all think, and any input is appreciated.
-Josh
The thing that intrigues me the most is the coloration of c3, c6, f3, and f6.
I wonder what we would see if we remove the knights, or we include only knights, bishops etc. etc. Ok Josh, users are always greedy
Cool tool Josh!
Miguel
I would be very curious to see what the image looks like if you filter Marcel's database for both players at or over 2300 Elo in strength, and also what it looks like if you use this database:
http://cap.connx.com/chess-engines/new- ... db.pgn.bz2
-
- Posts: 12541
- Joined: Wed Mar 08, 2006 8:57 pm
- Location: Redmond, WA USA
Re: Positional Strength/Most Controlled territory (WOW)
This idea might be as revolutionary as the material imbalance idea.
I think everyone should give it some thought.
Center squares = good, edge squares = bad, corner squares = horrid. Everyone knows this. But to numerically quantify it precisely is something we should all want to know about. And I suspect that there may be some surprises.
I also suspect that the slight edge to towards the black side of the board for square intensity is probably due to white's slightly winning edge overall (e.g. we see more power on the 7th rank than the 2nd rank because white will have the upper hand by some small margin and get his rooks or queen developed there first).
IMO-YMMV
I think everyone should give it some thought.
Center squares = good, edge squares = bad, corner squares = horrid. Everyone knows this. But to numerically quantify it precisely is something we should all want to know about. And I suspect that there may be some surprises.
I also suspect that the slight edge to towards the black side of the board for square intensity is probably due to white's slightly winning edge overall (e.g. we see more power on the 7th rank than the 2nd rank because white will have the upper hand by some small margin and get his rooks or queen developed there first).
IMO-YMMV
-
- Posts: 12541
- Joined: Wed Mar 08, 2006 8:57 pm
- Location: Redmond, WA USA
Re: Positional Strength/Most Controlled territory
+1 for Miguel's suggestion about exploring divided out by chessman.michiguel wrote:I think it is because of the Knights. d4,e4,d5,e5 are obvious, but f3 and the like are the only squares that attack two of the center squares. I think it is interesting the difference between h2 and h7. They are the only clearly "asymmetric" colored squares.Dann Corbit wrote:The majority of FICS players are not strong. I suspect that the dominance of squares could possibly be different if only very strong players were involved.jshriver wrote:I wrote a little program mostly as a visual to see what squares are "hot". Meaning if you watched a game or series of games, what places are commonly considered strong positions to control.
The images pretty much summed up common knowledge (center square). But personally was a little intrigued to find some variations. I pushed, Fischer's pgn database against it and found it the hottest spot was d4, and while a lot of GM's had the center 4 squares as their hottest, Kasparov for example had d4 e4 even moreso than the rest.
I personally find it useful for seeing trends, but can also be used on a per game instance. Just feed it t a PGN file.
http://olympuschess.com/strength/
Curious what you all think, and any input is appreciated.
-Josh
The thing that intrigues me the most is the coloration of c3, c6, f3, and f6.
I wonder what we would see if we remove the knights, or we include only knights, bishops etc. etc. Ok Josh, users are always greedy
Cool tool Josh!
Miguel
I would be very curious to see what the image looks like if you filter Marcel's database for both players at or over 2300 Elo in strength, and also what it looks like if you use this database:
http://cap.connx.com/chess-engines/new- ... db.pgn.bz2
I think you will probably have to share the script or the requests will be far more than you can ever process. But even if you don't I suspect that an hour of code will make a clever count routine (but the pretty colors will take someone who likes to do GUI stuff -- personally I hate to write that stuff but I love to use it).
-
- Posts: 1342
- Joined: Wed Mar 08, 2006 9:41 pm
- Location: Morgantown, WV, USA
Re: Positional Strength/Most Controlled territory (WOW)
Most of the time on my own engine has been dealing with the eval and positional influence. But not only that I believe positional influence can also change depending on different times of the game.
I also think this can be useful for mimic'ing player behavor. That's why I was pumping different datasets from various GM's to see what their "hotspots" are.
I also think this can be useful for mimic'ing player behavor. That's why I was pumping different datasets from various GM's to see what their "hotspots" are.