Doch evaluation contest

[Don]

I have not set the submission date yet for the doch evaluation contest. I won't set that date until I know I have enough entries to make it worthwhile. (I want at least 3 or 4.) But I will give plenty of advance notice.

I recently submitted bug fixes to Jim Ablett for compiling. This will help with the loading of parameters and bugs that have been reported.

I hope some interesting weight sets are produced that are substantially different from the defaults and being different will be one of the primary criteria for making it a built in personality.

Don Dailey

[DomLeste]

I'm in if i can find an improvement.. Count me in. hopefully i can find a more dynamic style for Doch even if i dont win go down in a blaze of glory!

[Marc MP]

Thank you M. Dailey,

A wind of fresh air and a good challenge!

I'll try a few test positions in the next couple of weeks, and try to suggest some settings.

Good luck for the Doch project,

Marc

[Graham Banks]

Don - could you please give a brief explanation of the effect that each parameter has. Some are fairly self-explanatory, whereas others are unclear.
If you could do this, it would encourage myself and probably others also to participate.

Cheers,
Graham.

[Dr.Wael Deeb]

Don - could you please give a brief explanation of the effect that each parameter has. Some are fairly self-explanatory, whereas others are unclear.
If you could do this, it would encourage myself and probably others also to participate.

Cheers,
Graham.

Ditto to that
Dr.D

P.S.Hey Graham,why bother to participate,you know my personality will beat the hell out of yours

[Graham Banks]

Hey Graham,why bother to participate,you know my personality will beat the hell out of yours

You're on!

[Don]

Don - could you please give a brief explanation of the effect that each parameter has. Some are fairly self-explanatory, whereas others are unclear.
If you could do this, it would encourage myself and probably others also to participate.

Cheers,
Graham.

If you look in personality.README you will see a brief explanation of each feature. I don't go into any serious detail but most of it is self-explanatory.

There are piece square tables that are fixed and not possible to modify, but you can control how much weight is given to those tables. I tried to make those tables as general as possible since they are fixed - and they are only a part of the value of each piece.

[Ron Langeveld]

The readme leaves a lot of questions open. I am still not sure how to assess the difference between penalties for double pawns on half open files compared to open files. I presume a double pawn on an half open file has enemy pawn(s) on it ? (in front).

So far I am seeing way too big eval differences when changing a single parameter just slightly. A little bit more info about the parameters and how they work would be welcome. For now I am not sure I am waisting my time, so I stopped experimenting. Sorry!

To end on a positive note : have you thought about a bonus/penalty per pawn in opposite colored bishop endings ? I was looking for that parameter in vain after seeing a test game.

[Don]

The readme leaves a lot of questions open. I am still not sure how to assess the difference between penalties for double pawns on half open files compared to open files. I presume a double pawn on an half open file has enemy pawn(s) on it ? (in front).

So far I am seeing way too big eval differences when changing a single parameter just slightly. A little bit more info about the parameters and how they work would be welcome. For now I am not sure I am waisting my time, so I stopped experimenting. Sorry!

To end on a positive note : have you thought about a bonus/penalty per pawn in opposite colored bishop endings ? I was looking for that parameter in vain after seeing a test game.

We have something to cover bishop of opposite color although it is not working all that well. It's basically a coefficient, if the only pieces on the board is one white and one black bishop and they are on opposite colors, the evaluation is multiplied by something like 50% - so if you are pawn up the score becomes 0.50 instead of 1.0. I would prefer being more severe but it weakens the program to push this too far. I think the right formula should probably not be linear - it should be something much more severe when the advantage is small and progressively decline when it's large. Bishop of opposite color is not automatically a draw as many people seem to think.

This is one of those parameters that you cannot control and is built in to the program. I admit that Doch could use much improvement in this area - and one thing I think might help is to use 2 coefficients, one to be applied to the material advantage (which should be severe) and one that applies to the positional components - because the positional evaluation is much more important that the material evaluation in these kind of endings.

By the way, if there are other pieces on the board but only 1 bishop each of opposite color I use 0.90 coefficient, so the program will tend to either aim, or avoid moving towards this particular ending.

[Don]

The readme leaves a lot of questions open. I am still not sure how to assess the difference between penalties for double pawns on half open files compared to open files. I presume a double pawn on an half open file has enemy pawn(s) on it ? (in front).

So far I am seeing way too big eval differences when changing a single parameter just slightly. A little bit more info about the parameters and how they work would be welcome. For now I am not sure I am waisting my time, so I stopped experimenting. Sorry!

To end on a positive note : have you thought about a bonus/penalty per pawn in opposite colored bishop endings ? I was looking for that parameter in vain after seeing a test game.

It's a lot of work for me to go through and document in great detail how the entire evaluation works. I don't mind expanding on it a bit if you have a specific question.

If there are 2 pawns on the same file, they each get a doubled pawn penalty. If there are 3, then of course all 3 get the penality and it's likely that there will also be isolated pawn penalties applied to each.

There are only 2 cases of doubled pawns, they are on open or half open files. A half open file just means there are enemy pawns in front of the leading pawn, so I guess that's really a "closed" file. A bit confusing. An open file means there are no enemy pawns in front of the leading pawns.

Something else not documented is candidate passed pawns. Even though there is no specific weight for this a candidate pawn gets 3/4 of the value it would get if it were actually passed. I don't specifically remember the definition for a candidate passed pawn but it much be on an open file of course and I think it must be "duofended."

I use the term "duofended" to mean that a pawn is either part of a duo or it's defended by a friendly pawn. In either or both cases it is "duofended" which seems to be a useful concept that I use a lot.

A pawn is backward if it is the least advanced pawn considering the 2 neighboring files and it's not duofended. It gets more penalties if it's restrained or blocked. Restrained means it cannot move forward without being subject to capture by another pawn or it cannot move forward because another pawn is sitting in front of it.

I hope that helps. If you have other specific questions please ask.

Don