Stockfish: SF-zh matches vs humans

[Vinvin20]

JannLee (2805) 1st match : 6-0
JannLee (2805) 2nd match : 4-0
eekarf (2764) : 4-0
mastertan (2688) : 5-0
opperwezen (2687) 5-1! One lost because I let SF play a weak move in the opening : Be6??

note : for the 3 last opponents, SF got only black pieces.

Games are here :
https://fr.lichess.org/@/Vinvin/search?ratingMin=2100&perf=18&mode=0&sort.field=d&sort.order=desc

It's clear that Stockfish is way stronger than any human in rapid games with good CPU and equivalent time.
JannLee thinks humans would be able adapt to the new engine but I think he's wrong, human cannot catch up, the level of SF is too high for human mind in this very complex game.
On the other side, with 0.5 second (1 CPU), SF stills blunder sometimes some "simple mates" and human can use that to win sometimes. Compare to chess, it's no more possible to tricks SF like that.

This 25 games was a very interesting experience for me !
Any suggestion for further experiences ?

Vincent

[ianfab]

Thanks for arranging these matches and summing up the results.

Well, I think you and JannLee are both right. If SF would not be improved for years like it happened for other crazyhouse engines, humans might catch up again, but I do not think this will happen.

Even though opperwezen is very strong, I think there is a typo in his rating ;).

[Vinvin20]

Even though opperwezen is very strong, I think there is a typo in his rating ;).

3687 -> 2687, done ! :-)

[ddugovic]

I agree, at present humans still have chances to win!

Even after #131 I think games against "team" opponents (multiple strong players, or a player with engine(s) assistance) could be fun, as could time and/or material odds games!

[Vinvin20]

You make me think about correspondence Crazyhouse game ...

[Vinvin20]

JannLee and opperwezen declined :-(
3 games on going :
https://fr.lichess.org/cp2IRN72609k
https://fr.lichess.org/yxYif0MAy4ey
https://fr.lichess.org/D4YoYzDHmkWx

[ianfab]

3 games on going

How much time do you use per move for SF's analysis?

[Vinvin20]

I've not started the engine yet.
Until January 2nd, I'll be few at home.
I plan to let it run for around 8 hours by move.

[ianfab]

Wow, that should yield high-level games. It will be interesting to see whether black can hold.

[Vinvin20]

I play only black with SF.

[ianfab]

Yes, I know. Considering white's advantage in crazyhouse I think it might be more difficult for SF as black on long time controls.

[Vinvin20]

Both games vs mastertan was already >+5 for SF at move 10th.
https://fr.lichess.org/cp2IRN72609k
https://fr.lichess.org/yxYif0MAy4ey (3...d6 is a bad move but I would have like to diverge from the other game)

The 3rd game still only on move 3.

[Vinvin20]

The 3rd game will be finished soon too : https://fr.lichess.org/D4YoYzDHmkWx
The tactical shot 11...P@e2!! made the score rise to >+3 quickly.
After forcing 12.Kc1! Q@f1+! score rises to >+8.

Humans are too weak even at corr games :-(

[ddugovic]

In fairness to the opponent, 11...P@e2 isn't that surprising, although it's difficult to calculate several moves in advance.

[Vinvin20]

P@e2! is not that easy for SF.

BTW, SF entered in the line 5...Bxf2+ to make a forced draw (because score is a slight white advantage after 5.Bg5? ) : 5... Bxf2+ 6. Kxf2 Ng4+ 7. Ke1 Qxg5 8. Bxf7+ Kd8 9. Nxg5 P@f2+ 10. Kd2 B@e3+ 11. Kc3 Bxg5 12. Qxg4 N@b5+ 13. Kb3 Nbd4+ 14. Kc3 Nb5+ 15. Kb3 Nbd4+
But eekarf played the incorrect 7.Kf1?

[ddugovic]

While I don't yet see a Black advantage after 12. Kc1, it's clear that 12. Kxe2?? gets mated and other moves lose the knight on g5. I hope it is interesting how this game plays out.

[Vinvin20]

A new won game vs eekarf : https://fr.lichess.org/PntdcE8e
Some comments from eekarf about 10.P@e5?

i saw the entire sequence that happened in the game, i just didnt think i would be worse, what should i have played
ah i see now. p@e5 seemed logical to me. its really hard playing the computer. it sees so many moves ahead