Is there a data file available (e.g. a .csv) containing every single rated game?

Question

Question

Is there a data file somewhere containing every single (FIDE) rated game?

What I found out so far

I see many websites with some limited database search capability, or summary statistics, but I couldn't find anywhere to simply download a complete dataset of FIDE games.

Such a dataset would have each row being a new game, and would hopefully also contain who played, who won, when it was played, and other info.

(Very) Rough size estimate

I estimate such a file would be quite large, which I base of some extremely rough estimates

• How many FIDE players? (probably) tens of thousands
• How many games on average per year? Probably hundreds
• Therefore 10000 * 500 = 5m games per year
• 5m games per year * about 100 years comes to about 500m rows

If each row is approximately 1kb, the total file size (uncompressed) would be: half a terabyte!

But still, tools like spark or bigquery could easily handle this, and it would be fun to analyse if it is available somewhere in its entirety.

Answer 1

Is there a data file somewhere containing every single (FIDE) rated game?

As an arbiter I can tell you definitively that the answer is "No". However you don't need to be an arbiter to know that. You just need to have played in a FIDE rated blitz tournament. Nobody is writing the moves down in those tournaments.

In FIDE rapid tournaments you do get a few players who start out recording but many of them stop when they get short of time and those games don't end up in databases.

Even when it comes to standard time control FIDE rated games only those games played in norm events have to be sent to FIDE (to make sure there is no cheating to get titles like GM and IM). So, all games played in norm events should be recorded and sent to FIDE. For each such tournament the games should be available for download from the FIDE site and those games will be picked up by the spiders / web-crawlers used by the big database providers.

However the large majority of FIDE rated standard games are not played in norm events and so there is no requirement for arbiters and organizers to send those games to FIDE. Doing so is hard (unpaid) work and so most arbiters and organizers are not going to do it.

Answer 2

It would be fun to create a decission tree based on all those games, however, I don´t think there´s anything like that, at least publicly. Besides, you´re probably making an understimation, as right now there are 360,000 fide rated players.

The biggest public chessdatabases are

• Chess Microbase
• Chess Tempo
• Chess-DB (and I think this latter was closed)