Thermopylay Marathon 2011 (live!)

[Daniel Shawul]

----

[Daniel Shawul]

A couple of mistakes I think. No GMS so anybody can analyze

8/5h2/2k5/2g2P2/5P2/3R3P/P3K3/8 w - - 3 48;avoidmove e7f6 fairy-max - sjaak
[D]8/5p2/2k5/2r2P2/5P2/3R3P/P3K3/8 w - - 3 48 [/D]

8/3l4/Pgkw1hk1/4Rh1N/8/8/Q4PPP/2R3K1 w - - 3 77;avoidmove c1c6 oberon - fairy-max
[D]8/3n4/Prkq1pk1/4Rp1N/8/8/Q4PPP/2R3K1 w - - 3 77[/D]

2k5/4c1hh/4g3/1R6/1PKB1h2/3P4/2P5/8 w - - 3 51;avoidmove b5b6 Nebiyu - ChessV
[D]2k5/4b1pp/4r3/1R6/1PKB1p2/3P4/2P5/8 w - - 3 51[/D]

4k3/3c4/4l3/PQ2NwP1/8/B2h4/5P1P/4K3 w - - 2 58; avoidmove b5b8 catalyst - spartacus
[D]4k3/3b4/4n3/PQ2NqP1/8/B2p4/5P1P/4K3 w - - 2 58[/D]

8/5lk1/4l3/P3RcPP/8/8/3R1P2/g2N2K1 w - - 13 58; avoidmove e5c5 chessV - spartacus
[D]8/5nk1/4n3/P3RbPP/8/8/3R1P2/r2N2K1 w - - 13 58[/D]

6k1/hkhhh1P1/4c2P/1g2l3/4hPB1/8/P7/B2RK2R b - f3 0 27; avoidmove e5g3 spartacus - chessv
[D]6k1/pkppp1P1/4b2P/1r2n3/4pPB1/8/P7/B2RK2R b - f3 0 27[/D]

8/3h4/3k1k2/PP2h3/6P1/2g5/5K2/R2R4 b - - 2 41; avoidmove e5f4 spartacus - fairymax
[D]8/3p4/3k1k2/PP2p3/6P1/2r5/5K2/R2R4 b - - 2 41[/D]

[Evert]

I noticed that Spartacus seems to give the Spartans more than +2 in the initial position. Is that deliberate?

By the way, the first game between Sjaak and Spartacus has a nice example of hoplite vs. wrong bishop.

[Daniel Shawul]

Probably due to very high mobility score. It likes to bring out the captains and leutenants very early +2. Catalyst also seems to have differing scores than the rest of the engines. My guess is different piece values. Almost all disagree with passed pawn evals.
I was of the opinion that the spartans were the favorites but after watching some games I now think white could be better if it tries to open up the position and push its pawns. The way the hoplites move is favorable for making passed pawns. The hoplites are dangerous when they are in groups of four at the centre and then march forwards by doubling up. But for that they leave a lot of room for the pawns on the wings to run away.

[Evert]

Probably due to very high mobility score. It likes to bring out the captains and leutenants very early +2. Catalyst also seems to have differing scores than the rest of the engines. My guess is different piece values. Almost all disagree with passed pawn evals.

Quite possibly. It doesn't really matter, of course, as long as the relative score reflects whether the position is better or not, but it looks a bit funny.
Using the recommended piece values for Spartans and Persians, Sjaak would also give an edge to the Spartans (but not +2, more like +.25), so I actually compensated for that by slightly tweaking the value of the Queen. Sjaak also likes to bring out its Captains and Lieutenants early. Probably makes sense, since those are the minor pieces.
Sjaak's handling of passed pawns is extremely simplistic, by the way, but I have something better in the works.

I was of the opinion that the spartans were the favorites but after watching some games I now think white could be better if it tries to open up the position and push its pawns. The way the hoplites move is favorable for making passed pawns. The hoplites are dangerous when they are in groups of four at the centre and then march forwards by doubling up. But for that they leave a lot of room for the pawns on the wings to run away.

Sounds reasonable to me. Almost all of the Spartan pieces are leapers, so they're not hindered so much by a closed position. I don't actually know whether you want to centralise hoplites or not. It makes intuitive sense, but maybe what you want is to push them to the sides and build up a counter attack to the centre when the Persians try to break through. Or simply make sure that you keep hoplites on all wings where there are still pawns present.
Having said that, I have the impression that getting an extra queen is not as much help if the Spartans have a strong defensive position, especially with two kings on the board. All of their pieces are less valuable than a queen after all...

[Richard Allbert]

Will do...

[hgm]

I noticed that Spartacus seems to give the Spartans more than +2 in the initial position. Is that deliberate?

No, this was not deliberate. I guess it must be because of way-too-large positional scores in the Piece-Square Tables. (Spartacus does not have mobility in its eval). The 'advantage' unavoidably disappears as the Persians develop. Compared to the Spartans the initial position for the Persians is much farther away from optimum.

Spartacus is very poorly tuned. It is based on the new engine I was writing, and I just slammed in some PST there from the top of my head, to be able to debug the search. I had not gotten to the stage of actually tuning anything. The next thing on my to-do list was to implement the increental attack table (which was sort of working in normal Chess, but disabled right now), and when that was properly validated, rewrite the move generator for captures to use that attack table. (Which should result in a speedup.)

Mobility would be available for nearly free (also incrementally updated) as a side effect of the attack table. And I wanted to postpone tuning until I had the mobility term in my eval.

[Richard Allbert]

I'm testing a version of Catalyst with Hash, Passed Pawn Eval, null move, and some small positional eval features.

Oh, and corrected Piece values (as per the website link you posted)

It play a little better

[hgm]

What was the download link again?

[Richard Allbert]

30 Games played without error vs FairyMax

http://www.bluefever.net/enginepage.html