I did not play many games but found that it may be an interesting match for n=1000 or n=2000 and I cannot say by a small number of games who is supposed to be better.
I think it may be interesting to test with many games
10000 nodes against 10,000,000 nodes or 20000 nodes against 40,000,000 nodes or 30000 nodes against 90,000,000 nodes to see if more nodes still help the better engine
new stockfish with 10n nodes per move against stockfish1.9.1 with 10*(n^2) nodes per move
Moderator: Ras
-
Uri Blass
- Posts: 11113
- Joined: Thu Mar 09, 2006 12:37 am
- Location: Tel-Aviv Israel
-
jkominek
- Posts: 98
- Joined: Tue Sep 04, 2018 5:33 am
- Full name: John Kominek
Re: new stockfish with 10n nodes per move against stockfish1.9.1 with 10*(n^2) nodes per move
For whatever reason no one bothered to comment on your proposed match sequence. Myself, I waited to collect enough data before opining a reasonable forecast.
I've added dots and drawn correspondence lines for n=1000, 2000, 3000, 4000, 5000, and 10000. The first match of 10K vs 10M is projected to be dead even. As n increases the balance increasingly shifts in the direction of SF 17.1.
The new tech on O(n) improves faster than the old tech on O(n^2)! As we learned in school, exponential beats polynomial.
I've added dots and drawn correspondence lines for n=1000, 2000, 3000, 4000, 5000, and 10000. The first match of 10K vs 10M is projected to be dead even. As n increases the balance increasingly shifts in the direction of SF 17.1.
The new tech on O(n) improves faster than the old tech on O(n^2)! As we learned in school, exponential beats polynomial.