15

I'm working on a rating/ranking algorithm and I'm finally at a point where I can test some hypothesis but in order to do so I need real data. To be honest almost any board game could be used for this but I thought that chess is the board game with the highest possibility of having such a database.

I was thinking of posting it on other stackexchange pages but in the end I thought going to the root would be most efficient.

What the database should fulfill:

  • At least 50 players
  • At least 1000 matches
  • Downloadable (maybe not required if I can parse it from the website chronological)

Since I only play chess on rare occasions I'm not really too familiar with all the websites out there so I hope someone can help me.

qwr
  • 3,535
  • 11
  • 35
  • What is a chess "match?" All I know is that most of the major ones have more than 2 games and people alternate colors. – PartyLovingHermit Jun 19 '17 at 02:55
  • It seems that this kind of test would need some kind of correlation data. Do you need the match players to have a known and solid rating? – user30536 May 23 '23 at 05:22

3 Answers3

17

Million Base 2.5 (available from a link here) and ICOfY Base are two free options that more than meet your given criteria, each providing millions of games in PGN format to draw from as you please. In particular, ICOfY Base includes well over 4 million games.

sharcashmo
  • 1,201
  • 10
  • 15
ETD
  • 23,592
  • 4
  • 95
  • 169
2

Lichess has an open database of over 4 billion game PGNs: https://database.lichess.org/

Nicolas Noel also has a Lichess Elite database that only keeps games of 2500+ vs 2300+ players: https://database.nikonoel.fr/

qwr
  • 3,535
  • 11
  • 35
0

I would also like to suggest Chessbase, which is my personal favorite. It has 8 million games. It does require premium to see the move PGN though. You can replay the moves, similar to our move replayer, for free. It can be downloaded here.

Starship - On Strike
  • 1
  • 23