Recaptture direction pic

g6 Recapture Direction – Measured Statistics

Pablo Lopez and I have been discussing recapturing on g6 (or g3). Is it better in general to recapture with the F pawn or the H pawn? Novices like myself like to recapture with the F pawn. More experienced players prefer the H pawn. In the comments to an earlier post, Pablo gave reasons why hxg6 is often the preferred move. In reply, I said that I could use my software – with a few tweaks added – to gather some statistics.

I thought with a “PieceOn” command added to my chess interpreter I could filter out some appropriate games with a position resembling Pablo’s and do some counting.

The illustration above shows a snippet of my script for game filtering and counting, the players in some of the games found, and the position Tal faced as black in his game against Botvinnik. I once met Botvinnik. A cruel English audience laughed at him for writing a computer chess program in Fortran. It also shows the table of results. The sample of games is smaller than it should be, due to limitation in my computer’s memory.

The results show that

  • hxg6 is by far the most popular reply: 191 games out of 228. (fxg6 played 37 times.)
  • fxg6 might give black a slightly higher chance of winning (14/37 is more than a third, )
  • But! fxg6 gives black more of a chance of losing and less chance or drawing.

fxg6: 12 losses to 11 draws
hxg6: 33 losses to 93 draws

So black is not really thinking about improving his chances of winning. The difference is the chance of black drawing versus the chance of black losing. Of course that is my interpretation of the numbers in the table and you are free to draw your own conclusions.

For the curious, Tal in the position illustrated chose hxg6 – just like Pablo.

Here are the numbers again.

result  all fxg6 hxg6
1-0   45   12   33
0-1   79   14   65
1/2-1/2  104   11   93
*    0    0    0
total  228   37  191

One thought on “g6 Recapture Direction – Measured Statistics

  1. Ivan Urwin Post author

    Pablo mentioned to me that certain openings give rise to positions where the Bxg6 Pxg6 exchange occurs naturally – regularly – simply as part of the opening. How difficult could it be to use my interpreters “opening” command and increment a counter for each opening, to find the these openings? A bit harder than I thought!

    21 : B13k Caro-Kann: Exchange, Rubinstein, 6…Bg4
    16 : C42j Russian Game: 3.Nxe5 d6 4.Nf3
    9 : D18c Slav: Dutch, 6…e6
    8 : B13e Caro-Kann: Exchange, 4.Bd3
    6 : C01p French: Exchange, 4.Nf3 Bd6 5.Bd3 Nf6
    5 : D43a Semi-Slav
    4 : D10r Slav: 3.Nc3 Nf6
    4 : A11 English: Caro-Kann Defence
    4 : C80r Spanish: Open, Bernstein, 9…Nc5

    The number is a count of how often the opening occurred in the 228 games I selected for having a Bishop swap on g6 somewhat like Pablo’s. I have not listed the openings that occurred less than 4 times.

Leave a Reply