Making dumb dumber

[abulmo2]

I just played 1000 games on 3+0.5 time controls and not a single game was lost on time or otherwise terminated. Also, considering the limited feature set of Dumber it's performance is very impressive! Well done!

+432 for Dumber. If the rating that was calculated for MMC in my gauntlets is anywhere near accurate, Dumber performs around 2000 Elo. What does it *exactly* have as a feature set? PVMove ordering, MVV-LVA, Killer, History heuristics, and a tuned (but not tapered) evaluation, with only PST's and material? (Or does it have other evaluation terms?) If that is all, I wonder if I can actually reach 2000 Elo with only that feature set in Rustic. I'd be happy if I can reach 2000 WITH a hash table.

Dumber seems to be very fast, being able to search to about 9 ply in the midgame. Must be the move sorting

Move ordering just uses MVV-LVA. At root, each best move is sorted first (with a stable sort) for the next depth iteration.
Evaluation is tuned and tapered and just uses material, PST and tempo. The Pesto's weights (from tscp pesto) are better than dumber's ones.

[Carlos777]

Hi Richard,

Could you compile a Windows binary compatible with old PC's? Mine is an AMD Phenom II x4. At least from Dumb 1.5 to 1.8 binaries are compatible with it.

Do you mean a compilation to a 32-bit executable? The 64-bit executable targeted core2 CPU which are pretty old already.

This is my PC:



It has W10 64bits. Dumb 32bits works fine, but Dumb 64bits one doesn't. Same happens with Dumber.

[mvanthoor]

Move ordering just uses MVV-LVA. At root, each best move is sorted first (with a stable sort) for the next depth iteration.
Evaluation is tuned and tapered and just uses material, PST and tempo. The Pesto's weights (from tscp pesto) are better than dumber's ones.

Just MVV-LVA and PVMove first, ordering at the root, and a tuned, tapered evaluation. Thanks.

Rustic Alpha 1 does not do PVMove first; only MVV-LVA, and has no tuned and no tapered evaluation. A 200 game match:

Score of Rustic 2021-02-15 vs Dumber 1.2: 24 - 158 - 18 [0.165]
... Rustic 2021-02-15 playing White: 14 - 79 - 7 [0.175] 100
... Rustic 2021-02-15 playing Black: 10 - 79 - 11 [0.155] 100
... White vs Black: 93 - 89 - 18 [0.510] 200
Elo difference: -281.7 +/- 60.7, LOS: 0.0 %, DrawRatio: 9.0 %
200 of 200 games finished.

It seems dumber sits around ~1950 Elo, +/- 60. ( Rustic 2021-02-15 is just Alpha 1, with a bit of refactoring/variable renaming.)

Rustic gets some extra ordering in Alpha 3, tuning in Rustic 4, and a tuned tapered evaluation in version 5. (If all goes well.)

Rustic 5 should have feature-parity with Dumber 1.2, if I disable Rustic's transposition table. I'll run another match then. If I can actually reach 1950, the transposition table addition should add another 100 points for a rating of about 2050.

That would be great. In that case, I could reach the same rating as VICE, with A LOT LESS features (Only one difference: Rustic's eval will be tapered and tuned, but VICE will have a lot of additional evaluation terms on top of PST's and material.)