opening statistics

I have some questions about the opening tool where it tells me how many times a certain move has been made and the percentage for white/draw/black .

1) where do these stats come from? GM games only? Over the board or online? Is it just a chess.com thing or for all of chess worldwide. Would this be the same on other platforms
2) i do find it useful to use on daily games especially when someone plays an opening im not familiar with but i think at times i read too much into it and puts me off playing moves because they have a low score.
For instance i am playing a game now with a Alapin and i am also following the lesson from Ben finegold on this in his “beating the anti sicillians” , however my predicted outcome is very low. I have faith in Ben however 😄🤞
I’d be interested in others experiences using this tool.

They are from the master games database.  They would be from over the board games by players generally rated 2200 and up. They are not just GM games, which you can see if you still down to the game level.

There are millions of games. I am not sure how often they are updated.

In general, if there is a line with hundreds of games, where one side wins a preponderance of the games, you can have some confidence that the results reflect some advantage or disadvantage from the opening.

There are of course blunders made by masters. You need to evaluate each position on your own instead of blindly repeating what has been played.

 

---

---

 

The statistics come from a massive database of OTB games. Most of the games are FIDE rated, but the author of the database may decide to include games that are unrated yet verified (the games of Morphy and Andersson, for example). Because chess games are not patented and publicly available, there are multiple versions of the database, compiled by their respective authors.

A slight difference in game collection and a different rating floor (the minimum rating to be considered in the statistics), mean that you'll see different numbers on other platforms. It's not a big issue, though, since good moves are always good moves.

Also check out 365chess which has a similar database which includes non-master games.  

There are a lot of factors which can influence win % in weird ways.  1. e4 e6 2. d4 d5 3. pxp pxp is symmetrical.  White must has a slight edge.  Yet every database will show black with a higher win %.  Various reason.  One, usually its the lower rated player that wants to escape book (not that the exchange variation really does that, but its simple at least).  Two, usually the player with less confidence will play something like the exchange variation.  And three, the reality is that the black player of the French defense will be comfortable in the exchange variation since they will see it all the time.  

Another reason that the percentages can be misleading is this:

When a line is decisively refuted, people stop playing it. So the previous winning percentage, before the refutation was discovered, gets "frozen in" because no new games are being played to update the statistics.

My favorite example is a line from the Petroff:

If you look this line up in a database, you'll find that Black enjoyed great success with it. In the 365chess database, for instance, the figures are:

White wins 35.7% / Draws 21.4% / Black wins 42.9%

Advantage: Black... right?

... or maybe not.

The reason that this line is still shown in databases as better for Black is that nobody plays it anymore... not since 14. Bh6!! was found... so the statistics still reflect its previous popularity and cannot update without new games.

The database percentages can also mislead in cases where a given opening is not well known or poorly studied.  Even masters can blunder an opening if they misunderstand the principles behind it.  Many of the flank openings are nowhere near as bad as their stats might suggest, they often just require a different stylistic approach to the game in order to be played soundly.