World Draughts Forum

Hi Michel, I understand but I do not see clearly how the tempo_difference can be estimated by the patterns. Let's take a simple example. Suppose you are dealing only with classical game. With a separate feature and by dividing the tempo_difference in for example 20 possible values from -10 to +9 yo...

Hi Michel, It seems I have a different view (but I have not your experience in reinforcement learning !) I agree with you that the tempo_difference can be estimated by the patterns alone but does that mean that we should avoid to use such fixed feature? My view is the following : because the variou...

Hi Michel, Concerning tempo difference do you also consider the patterns themselves can do the count or do you think that this feature is irrelevant? I think the question is whether you can estimate the tempo difference by using patterns alone. The answer is a definite yes here. In fact, you can cr...

Hi Michel, I never tried that experiment, but it is not intuitive to me that it would only be a minor disadvantage. Without an eval term for man difference, I would expect the engine might toss a piece just to gain some minor positional advantage. Do you understand why this doesn't seem to be a pro...

Oh, I thought there were better index functions. Better = less memory :P I am using this for checkers as well and initial resullts were quite promising. May I ask which optimization algorithm u use for the logistic regression ? I tried stochastic graidient descent which works but is still rather sl...

In order to solve the problem you mentionned I implemented two points: 1) I initialised the value of my 20 features to a value near from +2000 2) As soon as I want to modify a value I take into account a handwritten constraint saying that these values have to be ordered. Concerning the fixed featur...

Hi Michel, Maybe I am wrong but it appears to me that by calculating your f function you consider that the piece-difference is associated only one feature. Is it true? FYI in my implementation each piece-difference value (from 1 to 20) is associated to one feature among 20. That way the learning pr...

I don't think there is any 'human' meaning to the function. It just maps -infinite to +infinite to the [0,1] range, and does so with some particular nice properties. One thing to consider is that when for example you are 5 pieces ahead and say f=2.5, winning another piece doesn't increase the winrat...

Hi Gérard, The learning that i implemented, uses the logistic function to map the sum of features to a winning rate: winrate = 1/(1+exp(-f)) The winrate is the number of points you can expect to gain from the position (loss=0, draw=0.5, win=1.0), f is the sum of the feature weights. The learner find...

A small update from my part. I have added the breakthrough rules to dragon. First i played with the original evaluation function (from international draughts) against scan 3.0. This scores about 10%. Currently i am learning a new set of evaluation patterns for breakthrough. The initial results are e...

hi Michel I tried to generate the bases of dragon with 32 gig of ram 4 tera of hd system windows 10 i7 processor and gave a fatal error Generally after out of memory condition, you can just restart and it will continue where it left of. However, if that doesn't help it along, you probably need to a...

I guess multiple. In my case, I am back to computer chess; it's been official for a few minutes: http://www.talkchess.com/forum/viewtopic.php?t=65680 I am still available for discussions, but am not working on draughts at the moment. Hopefully, Scan 3 will be strong enough for a couple of years. Fa...

Hi Michel, Good news to see your interesting by this beautitul draughts variant. As with Bert I have the same difficulty with you : I did not generate the capture positions. Bert managed to add statitistics on non capture position in order to compare our figures but I do not know if it is easy for ...

It's nice to see a draughts-like game not ending in a almost-certain draw. I'll do some experiments with it as well. I created a BT endgame database for 10x10. Can any of you confirm the numbers if you have 10x10 available? Solved 1 - 1: wins= 1.148, size= 1.985, rate=0.0 Mpos/sec Solved 2 - 1: wins...

Hi Luzimar, This is actually a very hard position. While the position is a win for black, it takes 30 moves to actually win it. The program would have to search 59 ply to find the actual moves. The last move black played (7-1) was the optimal move, i am not sure it is unable to win. The only sure wa...