Introduction to Centaur Chess


Back to forum

John Smith    (2009-11-13)
Introduction to Centaur Chess

Hi all,

While I have played allot of chess, so far I only used my computer for an occasional analysis and mostly for the database features.

I am assuming it takes to know engines quite well to become good at advanced/centaur chess, so any advice would be really helpful.

1) Which engines are better at what type of positions? Is Deep Junior best at unclear sacrifices?, Rybka best for positional play?, Schredder best for endgames?

2) Which engines understand different pawn structures better, e.g. which is the best engine to study a stonewall-structure game and which is best for a King's Indian Mar de Plata game?

3) How to interpret the engine value for the position? e.g. if I, say as White, sac a pawn and the evaluation is -0.1, that is less that 1 pawn, does this mean I have enough positional compensation for the pawn?

4) Which engines take long-term weaknesses into their evaluations, even if they can't see anything concrete within their horizon?

5) Which free engines are worth consulting? toga? stockfish? Glaurung? thinker? Which of these are good for complex positions, which for quiet ones?

6) Are there any engines which improve their play during time, that is they learn? e.g. if in a position guiding the engine by hand proves that another move than its preferred one is best, will it be able to spot the move again, if the same position is re-entered?

7) Which is the best interface for analysis?

8) Is there a page with statistics of how each engine performs in every opening?


Michel van der Kemp    (2009-11-13 10:38:55)

I think for pretty much all these questions you are on your own. On the web you will find many rankings on how certain engines perform, mainly whilst playing other engines. All engines are able to make nice sacrifices or play positionally, as long as it doesn't go beyond their calculation horizon. If we as humans want to make decisions with a little help from our computers, well that's what determines if you are a top player I guess. But if there was a recipe to become such a player, I wish I knew it :)

Experience, and good understanding yourself is the key I guess.

Thibault de Vassal    (2009-11-13 16:44:56)
Centaur chess

Hi John, I'm afraid there's no clear answer to these questions, in my experience it is not possible to classify chess engines so accurately, each position is differently understood by chess engines, and actually is differently not understood, in example many great moves were found in some way "by chance" by engines like Deep Junior... But this is old computer chess already, Junior has not been updated for a while and Rybka is probably best in all parts of the game for a few years (maybe that is to change).

As Michel said, experience is the key IMO.

John Smith    (2009-11-13 18:19:12)
thanks for your responses

I see, it is a quite unexplored area.

Have engines advanced really that much though? Surely, there has been progress, but I did an experiment, I annotated some of my games using Fritz6 and Fritz 11. What caught me off guard was the fact that their 1st recommendation was the same everywhere, and in fact, oddly, Fritz 6 converged first to the "correct" reply.

They still sometimes fail to find some critical moves made by positional masters like Karpov or Kramnik. Of course they have also found many moves of their own (Bxh2!? vs Kasparov) which were not even considered by human masters.

Since my initial questions are probably on too abstract grounds, I'd like to ask a purchase question. I will buy Rybka soonish, however I was wondering if it is worth purchasing other engines as well.

- Is Shredder 12 worth it? are there some parts of the game where it does better than Rybka?

- Are they planning to release a new Deep Junior?

- Is the old Deep Junior 11 worth it, or because it hasn't been updated, even in positions where its strengths lie it has been surpasses by other engines?

Wayne Lowrance    (2009-11-13 18:42:00)
Centaur Chess

Howdy John, I will try and give maybe a little more detail to the best my ability.
1) Rybka is probably best up to "end game" and in end game I like Naum and Zappa.
2) Stonewall: , do not know best engine, but probably not Rybka
3.) Not enough information and again it depends on Engine Tactics. But in general, probably is worth looking into deep P.V.
4) Have no idea
5) No idea.
6) Out of book move "repeat/improvement recognition" as far as I know is not a feature. You as the Centaur of course should recognize this.
7) Well, you will get big arguments here. Many will tell you Aquarium (if you can master it).
8) NO In CC chess, the book is probably the most important thing to study and improve.As mentioned, experience is number one. Dont know if this post is of any value Wayne

Michel van der Kemp    (2009-11-13 19:26:27)
If I really want to delve into it

If I really want to delve into a position, I let two different engines analyze a position. If I'm out to win, I let both engines look for closed lines that don't lead to quick exchanges, unless they lead to clear advantages.

If engines evaluate a certain position very different, then those moves will catch my interest very quick, because those are the lines that may be highly imbalanced. I let both engines descend into those lines playing them against each other, and when you see the evaluation of one of the engines drop or go up, then it's a good time to draw a conclusion.

I hope that was clear a bit.

So yes it's good to have multiple engines.

John Smith    (2009-11-14 12:18:56)
thanks for the update!

Therefore I should be looking into purchasing Rybka and another engine whose style is different, and the two put together could create some controversy.

Which (strong) engines have a style that is as far as possible to Rybka's?

Thibault de Vassal    (2009-11-14 22:32:27)
Fritz 11 vs. Fritz 6

IMO, it is useless to try to compare Fritz 11 and older versions (or even other engines) when analyzing a human game... simply because there are much less "critical points" for engines in such games than in engine games or centaur games, 'cause there are much less horizon effects (sorry, I can't remember if this is the correct term, particularly after 1000 km on the road :)) in human games.

John Smith    (2009-11-15 04:37:55)
by critical

by critical you mean points where engine choices deviate?

Thibault de Vassal    (2009-11-16 21:11:07)
critical points

That's it, sorry my english is still quite weak :)

John Smith    (2009-11-17 04:11:40)

Nothing wrong with your English!, it was just a matter of definition :) I am not familiar with advanced chess so it was the first time I came across the term "critical points" in chess.

John Smith    (2009-11-17 09:54:56)
engine styles

I would be thankful if I had a little more information on engine styles:

- Which is the most positional engine(s) (has positional knowledge+plays more positionally)?

- Which is the most solid engine(s)?

- Which engine(s) that goes for wild complications most often (and is strong in complex/unclear positions)?

Thibault de Vassal    (2009-11-17 10:10:55)

Most solid+positional+knowledge is most probably Rybka, as for the term of "complications", this is a complicated question :) IMO any good player is better than any engine to find it & go in, then a centaur with Rybka is probably best to play it.

John Smith    (2009-11-17 10:25:12)

So which engine is good to co-consult given Rybka?
In order to be useful it should have a different style, as an engine with style similar to Rybka's is unlikely to provide significantly additional information.

I am guessing Deep Junior is somewhat old (?), Shredder is also solid+positional so possibly not too useful to co-consult, Naum & Zapa have been discontinued and in Fritz the focus has shifted more to training features in the interface.

Is it only Rybka these days? What about the open source & free engines that exist?

Thibault de Vassal    (2009-11-17 12:15:57)
One engine vs. several engines

Naum discontinued ? Really ? That would be bad news...

In my experience, other engines do not really give "additional informations". One chess engine give informations, several chess engines give odds. Other (weaker) chess engines may give bad odds as well and actually quite often. Only your analysis will give you true informations, you just have to explore not only the best variant provided by your engine.

Finally most important is to NOT trust chess engines IMO. In most cases at correspondence chess against strong players, in some ways "I" won some games (by analysis and ideas) and "engines" made me lose some others, that's correspondence chess... (hear me, of course it was my entire responsibility in all my losses & I wouldn't have been able to get good results without the help of engines :))

John Smith    (2009-11-17 13:38:01)
regarding Naum

Regarding Naum retiring, I didn't find an official word on this, but

1) Naum's site
appears to be down

2) according to wikipedia " Naum tied for first with Rybka in the 2008 Internet Computer Chess Tournament [3], but has not yet competed in other public tournaments."

Thibault de Vassal    (2009-11-17 15:27:32)

All I can say is that the complete Geocities website closed, so it doesn't mean anything for Naum which appeared 2nd behind of Rybka in a recent CEGT rating list.