sacrificing

[smcracraft]

I am sure this has been beaten to death before, but...

Hardware is getting so fast that one can just code computer chess correctly/traditionally and let "safe play" rake in all the material and notch the win resulting in boring and safe algorithmic play by the machine.

What about another style?

In the positions reached after

e2e4 e7e5
g1f3 g8f6
d2d3 b8c6
c1g5 f8c5
f1e2 e8g8
e1g1 d7d6
b1c3 c8e6
c3a4 h7h6
a4c5 d6c5
g5e3 c5c4
d1d2 e6g4
a1d1 d8d7
f1e1 g4f3
e2f3 c4d3
d2d3 d7c8

The future sacrificial Bxh6 is speculative.

Tal did this kind of thing many times betting on just throwing the opponent into chaos and due to his greater tactical calculating skills the outcome was exciting to him, the audience and to the victim and the greater game.

Given the "new world order" of fast iron, why not just assume that, let the other side suffer and sac against the king in a greater manner than is currently done.

--Stuart

[mkchan]

Sort of like guiding the search to prefer maximizing a particular type of eval (king attack in this case)

[jdart]

The top programs at least are optimized for computer-computer play, and computers are very good defenders. So tuning to prefer speculative sacrifices is not a good idea, if you are after rating points. Playing against humans is a different story. Complicating the game and making the opponent find the narrow path to avoid loss can be a good strategy there, although again, at the very top level, this is risky because GMs are good at defending.

--Jon

[smcracraft]

I believe excitement should trump rating points as we already have what we have with the rating points.

Not seeking a defender or counter-attacker. Seeking a "sweep you off the board" style (if you're a human opponent or extremely strong tactically.)

This would be an option given to the program at the game's start, similar to Crafty's contempt but much, much more sacrificial in certain given pattern areas directed against the enemy king.

[PK]

Stuart, You dream about something that is already being done As of now, I am running some tests with different Rodent's personalities, trying to quantify the effect of style changes. Observations so far:

- material values set to 90% (this means piece values, piece pairs and imbalance adjustmnts lumped together, otherwise penalty for getting eval out of sync is huge) lose 16 Elo points
- material values set to 75% lose 45 Elo points This is a lot, but still not ridiculously low.
- assorted tweaks that make the engine lunge forward and place pieces near enemy king (on the top of normal king safety) lose 27 Elo points
- personality overemphasizing both attack and mobility (150%), further burdened with a sick affection to a knight, and with slightly decreased material weight, loses 32 Elo points
- one of the most insane personalities by Brendan J. Norman, with attack/defence ratio 4 to 1, loses 210 Elo and this is the biggest loss obtained by honest tweaking (by honest I mean not using Rodent's weakening mechanism and not trying to make material values useless)

All of that is substracted from the base strength of roughly 2900 on the CCRL scale. I guess that with careful tuning I'd be able to tune Rodent as an attacking only 100 Elo lower. Even if CCRL ratings are inflated at the top by 200 points, it would be still a decent GM range.

[Stan Arts]

Just halve or quarter your material values and marvel at the crazy and usually losing moves your engine makes. But against carbon lifeforms it can get interesting for a casual blitz game.

My engines usually had an option for it and Nemeton is quite aggressive about it at it's easier levels.
At the 2200 and 1900 level material is halved.
At the 1600 level material is quartered and at the 1200 level divided by eight.

But on average engines are much more aggressive than say 20 years ago anyway.
"Correctly" sacrificing pawns, qualities (rook vs minor) and whole pieces for enough play.
Problem is they are all doing it so you don't notice it much as such lines are usually avoided.
It's against humans and weaker engines where you'll see stronger engines go to town on their speculative sacrifices.

[op12no2]

Just for fun.

Somebody asked for a much-easier-to-beat version of Lozza. I made one that *only* uses mobility for eval - no material values other than those implicit in mobility - no other terms.

You can try it here:-

http://op12no2.me/toys/lozzamystery/play.htm

[Henk]

Just for fun.

Somebody asked for a much-easier-to-beat version of Lozza. I made one that *only* uses mobility for eval - no material values other than those implicit in mobility - no other terms.

You can try it here:-

http://op12no2.me/toys/lozzamystery/play.htm

Just run your engine on chess.com. If you don't get caught after some time you passed the test and you can give it to humans.

Maybe better if we create a website for testing human-like engines. Users should guess if they are playing against an engine or not.

[smcracraft]

Great!!!! Can I play Rodent somewhere?

[smcracraft]

By the way, I am not talking about sacrifices in ANY position. I am talking about absolute attacks against the king, wherever he might be, depriving him of pawns, pieces, etc. regardless of the material negation to the computer attacker. My point in this is to return the romanticism to chess which has been lost with the ascent of the machine, correct play, typical master praxis and all that has come with dehumanization due to algorithmization. Get emotional people!!!!