I naively thought that if I picked the biggest winning move (for example using white) in the database and kept picking moves that white would always win, but at some point the advantage for white suddenly will often turn negative or even to 100% losing.

It seems that the move before that must have some bad data if it had a winning edge but one move later loses.

If the tree bears down to a total loss then how could the move before have had a winning edge?

Is this whole concept of winning % by move erroneous?

Am I missing something?

What other explanation can you provide?

2 Answers 2


You have to keep in mind that a tree of variations isn't actually a tree. You can reach the same position in many different ways, i.e. the branches can grow back together. So when you go through a variations the number of games doesn't necessarily dwindle until there is just one game left, occasionally games from another move order will join the fray and possibly change the statistics a lot.

So if you are following ten games with 90% for white, but reach a position that was also reached by a different move order, suddenly you might be looking at 50 games. And if for some reason white did much worse in the additional 40 games, the statistics will flip.

This is not as unlikely as it seems, because employing a different move order can indicate different overall knowledge of an opening. For example the 40-games move order may have been all the rage in a time when a very good later continuation for white wasn't known.


Opening trees are based on large databases with games played by high-level players. If a move has a 100% winning statistic for White, that means that in all games where that move was played, White won in the end. You should also look for a number which indicates how many times that move was actually played. If it is a small number, it is very unreliable. There might be only one game in the database, where Black blundered a piece later on even if the '100% winning move' was giving Black a good position. Or there might be a large rating difference between the players which explains the result.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.