Data-mining on the game of chess

I wondered what I would find if I examined a lot of chess-games. Not using conventional chess analysis but using data-mining techniques.
For that I retrieved some pgn-files from the web containing 4226520 games of chess played by grandmasters and rookies.
Total number of moves: 243580341
Total number of distinct moves: 7628
Average number of moves per distinct moves: 31932.399187205
Average number of moves per game: 57.63 (white+black!)
27.37% of all games end up as a draw.
Most games are played on a saturday (23.00%) and are mostly won by white on that day.
White wins most games on tuesday, black wins most games on saturday.
White wins most games in September, black wins most games in April.
The games analyzed were played between 98 and 2501.
Most of them were played in 2004 (5.088%).
The games were played at 44855 events at 86398 sites by approx. 346513 players.
Most games (11.21%) were played in August.
The most idle month (6.092%) was February.

Most used openings

4224001 games used a standard opening, in total 5093 distinct openings were used.

Most used openings (ECO codes) with their usage percentage

Eco%Most used on which dayFactor
1B221.829%saturday1.09150991
2B071.563%saturday1.23868398
3B011.507%saturday1.23315054
4D021.332%saturday1.14097035
5B331.190%saturday1.05051602
6B061.151%saturday1.13204159
7C021.069%saturday1.00572394
8D000.974%saturday1.14044145
9A000.962%saturday0.98491304
10C010.926%saturday0.93438863
The 'factor' in the table above is how much more often white won then black.

Opening with which white wins and looses most

I ignored the openings that were played less then 1% of those 1 million. I'll list the 10 with the biggest number of wins for white and the 10 lowest number of wins for white. n is the number of games played to determine that percentage.
+------+---------+-------+
| eco  | %       | n     |
+------+---------+-------+
| D36  | 67.5265 | 13826 | 
| C10  | 67.3591 | 14697 | 
| C41  | 62.9212 | 21363 | 
| B09  | 61.1395 | 15130 | 
| C06  | 60.5199 | 14741 | 
| B31  | 60.1284 | 12083 | 
| D35  | 59.9141 | 14020 | 
| E11  | 59.7699 | 12384 | 
| C42  | 59.5760 | 23622 | 
| D85  | 58.4710 | 11938 | 
+------+---------+-------+
+------+---------+-------+
| eco  | %       | n     |
+------+---------+-------+
| B20  | 45.2232 | 17137 | 
| B50  | 45.4277 | 19940 | 
| B25  | 45.6665 | 14064 | 
| A48  | 47.1792 | 19400 | 
| C01  | 47.4710 | 26337 | 
| B21  | 48.5237 | 20070 | 
| B02  | 48.5298 | 15478 | 
| A03  | 48.5911 | 14912 | 
| A46  | 49.0279 | 16619 | 
| A04  | 49.4418 | 15785 | 
+------+---------+-------+

Which color wins most often?

White wins 1.2215 more often then black.

Which moves have the most wins regardless of the piece-type

White

'Factor' is how more often white won then black
MoveFactor
1f6-f73.4651
2h8-f63.2662
3f7-f83.2448
4e5-h83.2163
5e7-g73.1881
6g6-g73.0875
7e6-e73.0365
8f6-h82.9871
9f5-f72.9569
10c6-c72.9459

Black

'Factor' is how more often black won then white
MoveFactor
1f3-f22.6734
2h1-f32.5719
3f2-f12.4563
4g3-g22.3948
5e3-e22.3821
6e2-g22.3188
7d3-f32.2889
8d3-d22.2648
9e3-f32.2591
10f3-g32.2100

High move-frequency for a certain piece: which one gives a better chance of winning?

White

'Factor' is how more often white won then black
PieceFactor
1Q1.3017
2P1.2999
3N1.2637
4B1.2581
5R1.2171
6K0.8925

Black

'Factor' is how more often black won then white
PieceFactor
1Q0.8973
2R0.8790
3P0.8686
4B0.8471
5N0.8335
6K0.6446

Which moves with what piece gives the most wins

White

MovePieceFactor
1f7-g8P6.3797
2a7-b8P6.0968
3b7-c8P5.8899
4f7-f8P5.8325
5f7-e8P5.6647
6g7-g8P5.6357
7c7-c8P5.4406
8e7-e8P5.4158
9a7-a8P5.3556
10b7-b8P5.2106

Black

MovePieceFactor
1a2-b1P5.1277
2g2-g1P4.8847
3f2-f1P4.7445
4a2-a1P4.5302
5e2-e1P4.5246
6f2-e1P4.5169
7d2-d1P4.3709
8b2-b1P4.3610
9c2-c1P4.3344
10h2-h1P4.2792
A conclusion might be that getting a pawn to get promoted gives you a big chance for winning the game.

What object is moved most?

White

Piece# moves
1P32843601
2N21841280
3R19858128
4B19551292
5Q15120406
6K13405660

Black

Piece# moves
1P32966178
2N21798577
3R19173198
4B18390028
5Q14737494
6K13894499

Which move was performed most?

White

PieceMove# moves
Bc1-e3953690
Ke1-g12374090
Ng1-f32703855
Pd2-d42530838
Qd1-d2596590
Rf1-e1919784

Black

PieceMove# moves
Bf8-e71100486
Ke8-g82390768
Ng8-f62716605
Pe7-e61723075
Qd8-c7621168
Rf8-e8860767



Interested in datamining in general? Then have a look at MiningCO.



Slashdot It!
Back to the menu
Last updated: September 28, 2008Please send feedback to: folkert@vanheusden.com