Computer Olympiad 2015 Leiden

[Rein Halbersma]

BTW, is your engine Scan going to be open source? I'd love to learn your approach to the eval in full detail.

Scan will be open source but this will only include code directly related to game playing. The development version is more like a library of modules to allow me to run "lab" experiments.

Great, looking forward to it!

Furthermore the learning stuff is very old. Regression techniques for position evaluation date back to 1982 apparently (http://chessprogramming.wikispaces.com/ ... Regression).
In the case of 10x10 draughts, Dragon already used evaluation learning about 8 years ago IIUC. I have not found evidence of a public discussion though.

There was a paper by two strong 10x10 draughts players on applying the well-known TD-leaf (from back-gammon) algorithm. Their program was a little stronger than GWD but much weaker than Dam 2.2. See http://www.researchgate.net/publication ... _databases

The Chinook team also tried it: http://ps2pdf.com/tgetfile/td.pdf?key=1 ... ame=td.pdf but IIRC they never used the machine-tuned weights in solving the game.

I use Othello stuff in Scan, so you should re-read the papers by Michael Buro.
They are all good but you seem in a hurry so I suggest starting with "Statistical Feature Combination for the Evaluation of Game Positions" for an accessible overview and "Experiments with Multi-ProbCut and a New High-Quality Evaluation Function for Othello" for mathematical details.

Nah, not in a hurry, just very curious Thanks for the links, it's been a while since I read those.

[BertTuyt]

And also from my side congratulations to Fabien, well done and a great result.

I'm as many others really interested to learn more about the (new) techniques used and applied.
I analyzed with the current version of Damage (but with only 1 core active, and with a 7P Database) the game JDraughts - Scan (with my old computer, so not the hyper system used during the tournament).

Also Damage didn't like the move 22. .... 1-7 of Scan , and prefers 20-24.
However although the score increased to +0.50 (perspective white) , Damage didn't not see a win, or any really bad move from one of the 2 players, from that point onwards.
So in the end the Draw was logical result.

Although Scan was right, I'm not sure if I prefer the move played, as the draw was far beyond the search horizon.
See attached a picture of the Analysis.

Bert

[BertTuyt]

According to Damage the TD King move 50. 39-34 against Scan was wrong, whereas 50. 41-36 is a draw.
See Diagram.



Possible Move sequence.

50 41-36 16-21
51 50-45 21-26
52 45-40 20-25
53 29x20 25x14
54 40-34 14-20
55 35-30 20-25
56 30-24 9-14
57 28-23 27-32
58 23-18 14-20
59 33-29 32-27

On my 4 Ghz 8-core system the draw is found (score 0.00) after 12.7 seconds at ply 26.
After 7.4 seconds at ply 24 the score already dropped towards -0.06 (slight advantage for black), but already with move 41-36.
At ply 21 ( 3.0 seconds) the program switches from 39-34 towards 41-36.

I expect similar (or even faster) results with Dragon and Kingsrow who both have parallel search and 8P DB's.

I didn't check if TD King missed draw opportunities after move 50.

Bert

[BertTuyt]

Please write as someone knows what versions of the programs participated in the Olympics.

The Damage configuration was:

Hardware: Intel i7-5960X 8 core at 3 GHZ, which was overclocked towards 4 GHZ as I have water cooling.
64 GByte RAM, which enables me to pre-load the complete 7P Database
2 SSD's with 250 GByte and 1 TByte

Software: GUI Damage 2015, Engine Damage 12.6, with 7P Databases , no opening book.

Bert

[Fabien Letouzey]

According to Damage the TD King move 50. 39-34 against Scan was wrong, whereas 50. 41-36 is a draw.

Yes, 39-34 seems a mistake. Scan expected 50-44 instead which it thought was equal. I vaguely remember discussing it with Ton during the game. Scan believed TD King made positional mistakes earlier, but they didn't seem decisive.

[Krzysztof Grzelak]

Fabien and on what hardware you have played the tournament.

[Fabien Letouzey]

Fabien and on what hardware you have played the tournament.

It was a quad i7 2.5 GHz (actual clock speed with 4 threads). There was 16 GB RAM but Scan only used about 2GB.

[BertTuyt]

According to Damage the TD King move 50. 39-34 against Scan was wrong, whereas 50. 41-36 is a draw.

Yes, 39-34 seems a mistake. Scan expected 50-44 instead which it thought was equal. I vaguely remember discussing it with Ton during the game. Scan believed TD King made positional mistakes earlier, but they didn't seem decisive.

Fabien, based upon the evaluation score at that point, TD King made indeed some non decisive positional mistakes earlier in the game.
I checked the move 50-44 and it seems also to lead to a draw, with 20-25 as the next black move.

Bert

[Krzysztof Grzelak]

According to Damage the TD King move 50. 39-34 against Scan was wrong, whereas 50. 41-36 is a draw.

Yes, 39-34 seems a mistake. Scan expected 50-44 instead which it thought was equal. I vaguely remember discussing it with Ton during the game. Scan believed TD King made positional mistakes earlier, but they didn't seem decisive.

Fabien, based upon the evaluation score at that point, TD King made indeed some non decisive positional mistakes earlier in the game.
I checked the move 50-44 and it seems also to lead to a draw, with 20-25 as the next black move.

Bert

That's right Bert. After moving 39-34 and answers blacks 20-25 a draw. Bert and Fabien thank you for information about the hardware.

[Krzysztof Grzelak]

Bert if there are some photos from the Olympics.

[BertTuyt]

Just follow this link: https://icga.leidenuniv.nl/

In the menu choose Computer Olympiad Photos.
The Draughts event took place the 6th and 7th day.

Bert

[Krzysztof Grzelak]

Just follow this link: https://icga.leidenuniv.nl/

In the menu choose Computer Olympiad Photos.
The Draughts event took place the 6th and 7th day.

Bert

Thank You Bert. By the way I write that the Olympics are unfair.

[Fabien Letouzey]

Thank You Bert. By the way I write that the Olympics are unfair.

Well, whatever the reason, I'm keeping the medal

[Krzysztof Grzelak]

Well, whatever the reason, I'm keeping the medal

Medal deserved the most.

[BertTuyt]

It is indeed unfair that others write better programs then I do

Bert