During testing for my custom built openings book, i find that my book stops playing its book moves and reverts to engine use despite there being book moves available, sometimes more than 1. Is there something i'm missing? perhaps in the settings?
![[gb]](/mwf/v12/flags/gb.png "United Kingdom")
Hi,
Not sure how experienced you are in book making so apologies if I am stating what you already know. There are 2 factors that could be influencing this.
One is the booksettings which if you use Fritz is Menu: Tools – Book settings (or press F4). On of the settings here is "Minimum games" tells the program how many games must be available for a move for it to be played. The default or optimised setting puts this at 3, so there must be at least 3 games in the book with that position in order for the book move to be chosen. You can change this setting to be 0 or 1 which means the move should be played even if zero or only 1 game. Another setting is "Variety of play" which is a scale you can move from left to right and will determines to what extent only the ones with the highest statistical values will be chosen ie move slider to the right means greater variety of play so it will play the seldom moves as well as the more common ones so you could try adjusting this.
The other main factor is the statistics in the book itself. If you open the book you'll see a whole set of values in each of the columns. "
Fact" = the factor or "weights" allocated to each move and control the probability with which it will choose the move in a game. The value can range between –125 to +125. The values are initially set to zero, but may change when the program actually plays games depending on the book settings ie. the weighting is changed in accordance with the results of the games achieved, so the book is "learning from experience". The weights can be manually modified by right-clicking a move and selecting “Change weight”. To reset all weights click Edit –Openings book – Reset weights. After that, the program will use only the purely statistical information to select its moves.
Prob and [%]: The first value is the purely statistical probability of the move, based on the number of times it was played and the results achieved in the games that went into the book. This is the probability of the move being played by the program as long as the weights have not been changed – either manually or by automatic learning. The value on the right [%] is the practical probability of the program playing the move. This is a product of the statistical evaluation and the weights. If you change the latter you can see how the [%] value (but not the “Prob”) changes. In Book options you can determine how much the weights influence the probability of play.
So all of these things can influence the extent to which a move is actually selected during play.
Hope this helps.
Mark
Not sure how experienced you are in book making so apologies if I am stating what you already know. There are 2 factors that could be influencing this.
One is the booksettings which if you use Fritz is Menu: Tools – Book settings (or press F4). On of the settings here is "Minimum games" tells the program how many games must be available for a move for it to be played. The default or optimised setting puts this at 3, so there must be at least 3 games in the book with that position in order for the book move to be chosen. You can change this setting to be 0 or 1 which means the move should be played even if zero or only 1 game. Another setting is "Variety of play" which is a scale you can move from left to right and will determines to what extent only the ones with the highest statistical values will be chosen ie move slider to the right means greater variety of play so it will play the seldom moves as well as the more common ones so you could try adjusting this.
The other main factor is the statistics in the book itself. If you open the book you'll see a whole set of values in each of the columns. "
Fact" = the factor or "weights" allocated to each move and control the probability with which it will choose the move in a game. The value can range between –125 to +125. The values are initially set to zero, but may change when the program actually plays games depending on the book settings ie. the weighting is changed in accordance with the results of the games achieved, so the book is "learning from experience". The weights can be manually modified by right-clicking a move and selecting “Change weight”. To reset all weights click Edit –Openings book – Reset weights. After that, the program will use only the purely statistical information to select its moves.
Prob and [%]: The first value is the purely statistical probability of the move, based on the number of times it was played and the results achieved in the games that went into the book. This is the probability of the move being played by the program as long as the weights have not been changed – either manually or by automatic learning. The value on the right [%] is the practical probability of the program playing the move. This is a product of the statistical evaluation and the weights. If you change the latter you can see how the [%] value (but not the “Prob”) changes. In Book options you can determine how much the weights influence the probability of play.
So all of these things can influence the extent to which a move is actually selected during play.
Hope this helps.
Mark
Thanks for that.
I entered my first book tourney just last month, and while i'm not a complete novice, there were a couple of beginner errors in my book.
I've rebuilt my book, and thought i had most things just about right, i have my minimum games set to 1. The one game in particular that i refer to, my book came out of book (so to speak) when it had 4 different moves available to it, line 1 had been played 8 times, line 2 had 4 games, and lines 3 and 4 each had 1. My thinking was that since my minimum games is set to 1, it could have played either one of these, but that it really should have played one of them.
Could it be because of having played these lines before, and having been unsuccessful with them the book has defaulted to the engine to look for an improvement?
I entered my first book tourney just last month, and while i'm not a complete novice, there were a couple of beginner errors in my book.
I've rebuilt my book, and thought i had most things just about right, i have my minimum games set to 1. The one game in particular that i refer to, my book came out of book (so to speak) when it had 4 different moves available to it, line 1 had been played 8 times, line 2 had 4 games, and lines 3 and 4 each had 1. My thinking was that since my minimum games is set to 1, it could have played either one of these, but that it really should have played one of them.
Could it be because of having played these lines before, and having been unsuccessful with them the book has defaulted to the engine to look for an improvement?
It would be interesting to look at far right column Prob [%] for those lines at that stage, to see what the probability was of the moves being selected. Also check that the moves are not marked in grey as that indicates that there is insufficient statistical evaluation for the move to be made available.
Many thanks for your input. I think deeper scrutiny is required...
OK, i can't recall now which game it was i originally referred to, but here is another game:-
After 1.c4 Nf6 2.d4 e6 3.Nc3 Bb4 4.Qc2 0-0 5.a3 Bxc3+ 6.Qxc3 b6 7.Bg5 Bb7 8.e3 d6 9.Ne2 Nbd7 10.Rd1, my openings book shows 5 different moves available to black:-
N % Av Perf Fact Prob %
11 59.1 2637 2565
10...h6 6 50.0 2643 2653 -60 100.0 0
10...Ne4 2 50.0 2629 2620 -10 0 0
10...Qc8 1 50.0 2635 2599 -6 0 0
10...Qe7 1 50.0 2581 2435 -19* 0 0
10...e5 1 50.0 2678 2377 -19* 0 0
None of the moves are marked in grey, nor are they marked in either red or green. All the figures in the "Fact" column are red.
It's at this point where the engine takes over, whereas i believe it should continue using the book, and most likely play 10...h6
After 1.c4 Nf6 2.d4 e6 3.Nc3 Bb4 4.Qc2 0-0 5.a3 Bxc3+ 6.Qxc3 b6 7.Bg5 Bb7 8.e3 d6 9.Ne2 Nbd7 10.Rd1, my openings book shows 5 different moves available to black:-
N % Av Perf Fact Prob %
11 59.1 2637 2565
10...h6 6 50.0 2643 2653 -60 100.0 0
10...Ne4 2 50.0 2629 2620 -10 0 0
10...Qc8 1 50.0 2635 2599 -6 0 0
10...Qe7 1 50.0 2581 2435 -19* 0 0
10...e5 1 50.0 2678 2377 -19* 0 0
None of the moves are marked in grey, nor are they marked in either red or green. All the figures in the "Fact" column are red.
It's at this point where the engine takes over, whereas i believe it should continue using the book, and most likely play 10...h6
Hi,
The reason it won't play any of those moves at that point in the book tree is that the % = 0. As I said in my post above the [%] is the practical probability of the program playing the move. It is a product of the statistical evaluation associated and the weighting ("Prob"). In your example all the weightings are negative (which is why they are marked red) which has resulted from games being lost by adopting those moves in the past (the program has "learned" they are not moves that result in a win) plus very low number of games so the combination of low statistics and negative weighting means 0 % chance of the the program choosing that book move. If you were to change one of the weightings to a positive value by right clicking on it and doing adjust weighting and try again you will probably see that the book move will be played.
The reason it won't play any of those moves at that point in the book tree is that the % = 0. As I said in my post above the [%] is the practical probability of the program playing the move. It is a product of the statistical evaluation associated and the weighting ("Prob"). In your example all the weightings are negative (which is why they are marked red) which has resulted from games being lost by adopting those moves in the past (the program has "learned" they are not moves that result in a win) plus very low number of games so the combination of low statistics and negative weighting means 0 % chance of the the program choosing that book move. If you were to change one of the weightings to a positive value by right clicking on it and doing adjust weighting and try again you will probably see that the book move will be played.
Many thanks for your help, much appreciated
Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill