Will my shatranj engine surpass Fairy-Stockfish 14 classical by 2025?
➕
Plus
15
Ṁ1438
Jan 2
92%
chance

Shatranj (https://en.wikipedia.org/wiki/Shatranj) is an ancient form of chess.

Fairy-Stockfish 14 is a fork of Stockfish designed to support a wide variety of chess variants. It is superhuman in nearly all variants, and is undisputed top engine in all variants that do not have dedicated engines (of which shatranj is one). Fairy-Stockfish 14 is approximately as strong as Stockfish 8 in regular chess.

Reasons to vote yes:
My engine is a dedicated shatranj engine, while Fairy-Stockfish 14 is not particularly optimized for shatranj
I can use NNUE (I'd have to learn NNUE first though)

Reasons to vote no:
Fairy-Stockfish 14 has Stockfish search and (classical) evaluation, both of which are much more sophisticated than what I can achieve (even if not optimized for shatranj).

Currently my engine is 500-800 elo worse than Fairy-Stockfish 14 classical.

In January 2025, I will run a 10000 game match using cutechess between my engine and Fairy-Stockfish 14. I will use shatranj_new.epd from https://github.com/ianfab/books. The time control will either be 6+0.1 or equivalent fixed nodes (currently, this is my engine 600k nodes/move vs Fairy Stockfish 100k nodes/move).
Cutechess will provide an a +- b elo difference estimate between my engine and Fairy-Stockfish 14.
Resolves YES if the lower bound of the elo difference between my engine and Fairy-Stockfish 14 is positive (i. e. a-b > 0).
Resolves NO if the upper bound of the elo difference between my engine and Fairy-Stockfish 14 is negative (i.e. a+b < 0).
Resolves 50% if the test is inconclusive (i.e. a-b < 0 < a+b).

Additional information:
My engine can be found at https://github.com/sscg13/sscg13-chess-engine/tree/shatranj
Fairy-Stockfish 14 can be found at https://github.com/fairy-stockfish/Fairy-Stockfish

Get
Ṁ1,000
and
S3.00
Sort by:

+117+-93
So actually we are ahead right now according to SSS
(20 wins 7 losses 13 draws)

130 +- 30 right now

Update, should be somewhere between 150 and 200 elo away, after adding NNUE.
It is possible that I get compute help from other engine developers (if so, this would greatly assist my progress).

When I first started this market, I would have been approximately 650 elo away from my goal. Now, I would be approximately 400 elo away (tested ~1 week ago)

© Manifold Markets, Inc.Terms + Mana-only TermsPrivacyRules