Not logged inRybka Chess Community Forum
Up Topic Correspondence Chess / Correspondence Chess / Machine Vs. Centaur Challenge (I/X)
1 2 Previous Next  
- - By Uly (Gold) Date 2012-01-17 06:23
I am the Centaur, Razor is the Machine, this is an experiment to see how far away are we from the point at which interaction in correspondence chess isn't necessary because engines can come up with the best moves by themselves, given enough time.

For conditions on the match check:

http://rybkaforum.net/cgi-bin/rybkaforum/topic_show.pl?pid=393117

Good luck Razor :smile:
Parent - - By Kappatoo (*****) Date 2012-01-17 09:47
If your opponent doesn't want to use an opening book, what about playing Fischerandom then? Or starting from something like 1.f3 f6?
Your advantage - if you have one - should be bigger than with opening books on both sides.
Also, will Razor say how much time he gave to the engine?
Parent - - By Uly (Gold) Date 2012-01-17 18:22

> Your advantage - if you have one - should be bigger than with opening books on both sides.


I have the black pieces.
Parent - - By Kappatoo (*****) Date 2012-01-17 18:26
I don't mean the opening advantage. I was talking about the advantage from being a better player: I.e., this advantage is more likely to manifest itself if both players are out of book on move 1 (or 2).
Parent - By Uly (Gold) Date 2012-01-17 18:41
Well, we're going to play 10 games, all sort of things can be tested.
Parent - - By Razor (****) Date 2012-01-17 17:39
Game 1: Single Core Engine vs Uly

1.e4

Is this sufficient information - I guess it must be as anything else would give the game away!  :smile:
Parent - - By Uly (Gold) Date 2012-01-17 18:26

> Is this sufficient information


Well, I guess you're not using Critter :smile:

Anyway, I don't think an engine can follow the subtleties of the Sicilian, chances are e5 would lead to Spanish exchange which is drawish, so this is my move:

1...c5
Parent - - By Razor (****) Date 2012-01-17 18:44
Game 1: Single Core Engine vs Uly

1.e4 c5 2.Nf3

I'm not going to make it easy for you by telling you what I'm not using!  :smile:
Parent - - By Uly (Gold) Date 2012-01-17 18:48
So, are you using book on or off? Nf3 came quite fast :lol:
Parent - - By Razor (****) Date 2012-01-17 18:51
Book off - the engine is in my preferred ShredderClassic GUI and I let it run in 'infinite Analysis' mode for just over 15mins - if I say anymore I might give the game away - is this OK?

BTW, I'm happy to let the engine think for longer if you would prefer?
Parent - - By Uly (Gold) Date 2012-01-17 18:56
Well, I thought you were going to take 24h/move, but I think you can decide on that.

2...Nc6
Parent - - By Razor (****) Date 2012-01-17 19:04
Well, I'm happy to take 24h but also really happy to get this game underway.  I certainly won't take anymore than a day unless I'm away from home.  If the engine were coming up with rubbish {so far it hasn't! :smile:} then I would certainly make it take the whole day.
Parent - By Uly (Gold) Date 2012-01-17 19:12
Indeed, if the engine has already found the best move there's no need to make it think longer.
Parent - - By Vempele (Silver) Date 2012-01-17 19:46

> If the engine were coming up with rubbish {so far it hasn't! :smile:} then I would certainly make it take the whole day.


I have a better engine for you: it changes its move selection every minute, cycling through all legal moves. Its CPU usage is insignificant so you can perform interactive analysis with your favorite engine while it's running.
Parent - - By Uly (Gold) Date 2012-01-17 20:06
Wouldn't that play at the strength of the human (since the human can choose at which time it stops.)
Parent - - By Vempele (Silver) Date 2012-01-17 23:02
It would play at the strength of the centaur.
Parent - - By Uly (Gold) Date 2012-01-18 02:18
So, Centaur Vs. Centaur? Then Razor could just have challenged me to a normal match.
Parent - - By Vempele (Silver) Date 2012-01-18 07:14
By choosing when to stop analysis while aware of the current move choice, he's already centauring.
Parent - - By Uly (Gold) Date 2012-01-18 08:09
Since the moves of the machine have already been optimal, I don't think waiting to stop the engine at a set time would have made any difference in move choice, and it would have only increased the game length unnecessarily.
Parent - - By Vempele (Silver) Date 2012-01-18 08:54

> Since the moves of the machine have already been optimal,


Exactly! Suppose you gave the machine enough time to see that 19 of the 20 possible starting moves are drawn and it plays Nh3. Was that more optimal than 1. e4?
Parent - - By Uly (Gold) Date 2012-01-18 09:01
I didn't say "perfect", optimal are moves that would get better results against a Centaur. I think 3.Bc4 would have been a non-optimal move, even if it was perfect.

1.e4 is optimal while 1.Nh3 (and 1.Nc3 which some engines want to play) isn't.
Parent - - By Vempele (Silver) Date 2012-01-18 09:05
So you accept that he possibly gained an advantage by not letting the engine almost solve chess?
Parent - - By Razor (****) Date 2012-01-18 18:39
Hi Uly,

I have just got home from work and have read the thread above and would be happy to stop if that's what you want to do.

Some points for you to note before you let me know:
- I was of the opinion I could let the engine run for up to 24h and said to you that it should usually be less than that
- the engine decides what to play - not me - the engine is on 'Infinite Analysis' and I just take the move recommended - if you don't trust me then happy to stop
- I believe this test is about proving to me/and those following that you are able to steer the game to expose weaknesses - ultimately win of course.  I believe you need to do at least the following to prove this to me/others:
     1) When you are ready, let me know what engine I am using - I will confirm this after the game or some moves later if (2) below is not working
     2) Tell us what you are doing, i.e., what is your plan.  I realise you cannot do this until you have stated (1) but unless you do this, then how can you prove you can do this!

If you want to continue then the move on the screen is 6.Bg5

I will pick up your response later - about to go out for a few hours - club night!  :smile:
Parent - - By Uly (Gold) Date 2012-01-18 21:13
Let's continue, Vempele says that if you let the engine think for longer it may pick up a worse move, and so that by stopping it early you're playing a better move. I claim that the spirit of the match is that you play the best move the engine can come up within 24 hours, if the engine already comes with the best move in one minute there's no reason to waste 23 hours, 59 minutes to see if it would come with a worse one. This is specially true in obvious recaptures.

Anyway, my move is d6, and there's not much to comment since I'm still in book :wink:
Parent - - By Razor (****) Date 2012-01-18 21:45
OK Uly,

Yep, I saw your last move was 5... d6 hence I said the move on the screen was 6.Bg5 - I assume you didn't see this bit!  :smile:

Off to bed now - I will see if you have posted a move in the morning before I leave for work - sadly 06:30!  :sad:
Parent - By Uly (Gold) Date 2012-01-18 22:08
Oh, I say your move but made a typo! I meant to post 6...e6
Parent - By Uly (Gold) Date 2012-01-18 21:14

> So you accept that he possibly gained an advantage by not letting the engine almost solve chess?


I'm certain the single core engine couldn't have solved chess in 24 hours, and wouldn't have went close.
Parent - - By Uly (Gold) Date 2012-01-23 20:56
Hey Vempele, are you done fixing your repertoire yet? We, like, need some more players for WBCCC 2 :lol:
Parent - By Vempele (Silver) Date 2012-01-24 12:32

> Hey Vempele, are you done fixing your repertoire yet?


Not even close. :( It's like chess is a complex game.
Parent - By Razor (****) Date 2012-01-18 18:41
Hi Uly,

I have just got home from work and have read the thread above and would be happy to stop if that's what you want to do.

Some points for you to note before you let me know:
- I was of the opinion I could let the engine run for up to 24h and said to you that it should usually be less than that
- the engine decides what to play - not me - the engine is on 'Infinite Analysis' and I just take the move recommended - if you don't trust me then happy to stop
- I believe this test is about proving to me/and those following that you are able to steer the game to expose weaknesses - ultimately win of course.  I believe you need to do at least the following to prove this to me/others:
     1) When you are ready, let me know what engine I am using - I will confirm this after the game or some moves later if (2) below is not working
     2) Tell us what you are doing, i.e., what is your plan.  I realise you cannot do this until you have stated (1) but unless you do this, then how can you prove you can do this!

If you want to continue then the move on the screen is 6.Bg5

I will pick up your response later - about to go out for a few hours - club night!  :smile:
Parent - - By Razor (****) Date 2012-01-17 19:35
Game 1: Single Core Engine vs Uly

1.e4 c5 2.Nf3 Nc6 3.d4
Parent - By Uly (Gold) Date 2012-01-17 20:06
cxd4
Parent - - By Razor (****) Date 2012-01-17 20:08
Game 1: Single Core Engine vs Uly

1.e4 c5 2.Nf3 Nc6 3.d4 cxd4 4.Nxd4

My engine is happy!  :smile:
Parent - By Uly (Gold) Date 2012-01-17 20:12
Yes, engines overestimate white's chances on the Sicilian.

4...Nf6
Parent - - By Razor (****) Date 2012-01-17 20:37
Game 1: Single Core Engine vs Uly

1.e4 c5 2.Nf3 Nc6 3.d4 cxd4 4.Nxd4 Nf6 5.Nc3

I have never run an engine without an opening book before and as this engine appears to be playing moves I would play began to have doubts - just checked again and can see both book options in the ShredderClassic GUI are unchecked so believe this engine is figuring this out.  i hope this doesn't give you too many clues.

Engine still looks happy!  :smile:
Parent - By Uly (Gold) Date 2012-01-17 22:18

> i hope this doesn't give you too many clues.


Well, unless you're using the Shredder engine who has an internal opening module that helps it navigate the opening as if it had a book, I have no clue :smile:

Anyway:

5...d6
Parent - - By Razor (****) Date 2012-01-19 17:35
Game 1: Single Core Engine vs Uly

1.e4 c5 2.Nf3 Nc6 3.d4 cxd4 4.Nxd4 Nf6 5.Nc3 d6 6.Bg5 e6 7.Bb5

Hi,

This is certainly not a line I would have played as white - how about you?
Parent - By Uly (Gold) Date 2012-01-19 21:47
I would have gone with the plan of castling queenside there. But, anyway, enough book, I'm going to take a look and see what I can find about your engine.

I have played all my moves instantly so I can spend in this move 7-14 hours :lol:
Parent - - By Uly (Gold) Date 2012-01-20 05:45
Well, I have very high suspicions about the Shredder engine, as it matches all your moves so far, I just hope it's not Shredder 12 as I don't have it :lol:, but anyway, even if it isn't it's an engine with very close style to that, at least on the opening, so I have something to work with.

Anyway, of course you can't confirm if I nailed it, otherwise I could guess a different engine every move until you tell me it is :smile:
Parent - - By Razor (****) Date 2012-01-20 05:59
Yes, I can't tell you!  :smile:
Parent - By Uly (Gold) Date 2012-01-20 07:19
Anyway, just in case, I'm playing this move as Shredder has some ideas about playing tempo losing moves like Be2 or Be3 (or in some lines, both :eek:) at low depth, with a happy face of 0.30 something, so I think it's worth a shot.

1. e4 c5 2. Nf3 Nc6 3. d4 cxd4 4. Nxd4 Nf6 5. Nc3 d6 6. Bg5 e6 7. Bb5 Bd7

r2qkb1r/pp1b1ppp/2nppn2/1B4B1/3NP3/2N5/PPP2PPP/R2QK2R w KQkq -
Parent - - By Razor (****) Date 2012-01-20 22:12
Game 1: Single Core Engine vs Uly

1.e4 c5 2.Nf3 Nc6 3.d4 cxd4 4.Nxd4 Nf6 5.Nc3 d6 6.Bg5 e6 7.Bb5 Bd7 8.Nb3

Looks to be fairly stable on the move selected so won't leave the engine running until the morning.
Parent - - By Uly (Gold) Date 2012-01-21 09:42
Suspicion changed to Stockfish, though, there are a zillion versions and I don't currently have the latest one, I'll download it if the suspicion continues. In these positions Stockfish has great trouble with King Safety and material evaluation (in some lines it's very happy after I sacrifice a pawn, and takes ages to discover Black has a small edge.) Anyway, 9. Be2 is comming, which makes me happy as I would have played a6 in this line anyway.

1. e4 c5 2. Nf3 Nc6 3. d4 cxd4 4. Nxd4 Nf6 5. Nc3 d6 6. Bg5 e6 7. Bb5 Bd7 8. Nb3 a6

r2qkb1r/1p1b1ppp/p1nppn2/1B4B1/4P3/1NN5/PPP2PPP/R2QK2R w KQkq -
Parent - - By Razor (****) Date 2012-01-22 08:23
Hi Uly,

Whilst the above is useful, I believe you need to start responding with a 'what you are trying to do' statement?  When you do, perhaps something like . . .
- I believe I am now playing/or continuing to play against X engine
- I believe X engine will play Y move next
- Based on the above my plan remains the same, i.e., I believe the imbalances in the position are Z and that I will continue to exploit this on the kingside, centre, queenside {which?} and still believe the best squares for my Bishops, Knights, etc., are . . . perhaps some diagram to show what the position will look like some moves ahead.
- My thoughts on the outcome of this game is a win for me in < Y moves

Without the above {or something like the above} you will not be able to convince me/perhaps others? that you are able to predict the engine you are playing against or indeed that you are playing against the SP engine I am using.  Make sense?

I suppose the other point to remember on this is that I am playing with the book options disabled so the engine I am using has calculated all the moves played - I seem to remember you saying {this may be wrong so please help me out on this} that you can steer an opening towards positions that suit you - I assume you meant with books on and therefore assume with books off this challenge becomes easier for you - is this correct?

Looking forward to seeing how you are going to move this game to favour black and what your plan is to achieve this.
Parent - - By Uly (Gold) Date 2012-01-22 08:47
I can't have certainty on anything, and this is not about knowing what the engine will play and have a single, infallible plan against it, it's about having working plans against the most likely moves that are coming from the engine.

As I said, I currently think that your engine is Stockfish, but can't say with which certainty, as it's also possible you're using an engine I didn't check or one that I don't even have, you can't expect me to check all possible engines (which is impossible). Also, I'm not leaving the engines analyzing the same position for hours like you do, I'm interacting with them at low depth, which means, that even if I nail the engine that you're using, I can't be certain of what move would it play at high depth. But that's not what the challenge is about, it's that, even if the engine plays a move that surprises me, one that I didn't check at all, I'll be on positions that I can still steer to my advantage (something unlikely if I was playing a Centaur because the centaur would surprise by having a long term plan I can't see, while the engine would only have a single line with positive score at a tail of the variation I can outprepare for).

What I'm trying to pull off is a Xanatos Gambit NOT a Batman Gambit, the latter to which you'd expect my explanations of the game to be like you said, in where I could be wrong and fail (having a simple line predicted and focusing on it would be like putting all my eggs on one basket).

I don't think predictions are useful this soon in the game, I think I can win in the end but eventually it's going to be a lot of work, right now my move was a simple one in where I'd expect to win the Bishop pair or get an attack on the Queenside against the castled King, while other white moves don't cause black problems. However, my plans will change as the game advances, and I can't predict what I'll do (say, by move 40), that would require trying to plan the whole game in advance which I don't think it's necessary.

I'm still going to predict that from this position I can make your engine show a negative score by move 20 or 24 (though you won't confirm that), and if not, engines at high depth play stronger than my estimations.

>I assume you meant with books on and therefore assume with books off this challenge becomes easier for you - is this correct?


I turned off my books at move 7, and just had my analysis from there, which I think will make Kappatoo happy as the position after your 7th move was one in where I didn't have any analysis at all, so that's the position where the game started. My only "book off" advantage was your engine's Bb5!? Be2!? plan (the latter which is best but can be played straight), so not much impact unless in future games your strategy would be to play full book until a drawn position or something. From the current position I think I have full equality.
Parent - - By Razor (****) Date 2012-01-22 11:46
OK Uly, in that case I have misunderstood you and suspect this may only be me as opposed to you/Kappatoo {not sure about spelling!}/Others.  I thought what I was signing up to was you playing {not a chess engine/engines of your choice} against any engine I choose and to show that you could steer a game to your advantage over an engine that does not have any influence from the engine user other than to take the best move it recommends within a 24h period.  What it looks like you are saying below is that you cannot tell us what your current plan is against the engine you believe you are playing against, i.e., by making plans based on who you believe you are playing alongside the current view you hold on the position in front of you.  Without this 'prediction' {NB: I am saying current prediction; I am not saying what is your plan and by the way you cannot change this plan} then I cannot see how you are going to convince me/others that you can in fact steer the game towards your current forecast.  I'm happy for your forecast to change as you discover more about your opponent but there should be some indication that you have followed your current forecast.

When thinking about the above alongside the point you make around using an engine/engines for a small amount of time then I believe this just makes it even harder to prove/disprove.  Some of your moves so far have taken longer to land than the ones I have taken from the engine and this, together with the lack of comparing actual against your forecast, makes the task of convincing me/others even more challenging.  For example, say I was using a SP version of Stockfish {I'm not BTW} you just need to use the strongest engine out there {lets say for now this is Houdini 2c} on some decent hardware {I guess you will have i7 technology at least} and even if I run my SP engine for say 10h {most I have used so far} your quad {or greater beast! :sad:} will just crush my SP in the end due to the horizon issues with engines and you can clearly hide behind this and claim it was all your own work - do you see this?

Assuming you believe I have understood you correctly {I will wait for your feedback on this} and agree with the inherent problems I have shown, then actually I am going to stop as this test is not going to prove/disprove anything i signed up for.
Parent - - By Uly (Gold) Date 2012-01-22 22:20

> Some of your moves so far have taken longer to land than the ones I have taken from the engine


What are you counting on this? I played the first 7 moves on the game instantly, took about 4 hours in total for 7...Bd7 8...a6 , and played Be7 instantly, from the time I read your message, to the time I posted my move (I actually spent much less than this since I also have idle times like eating or typing messages in the middle of analysis while I leave my CPU doing nothing). If you are getting the feeling that after you post a move, I'm taking ALL that time to analyze my move, then, yes, you have a wrong understanding about what is going on.

As for your chickening out from the match, I don't get what your problems with it are, I'm going to repost our conditions:

I use my usual Interactive Analysis methods for 1 or 2 hours to decide on a move.
You use an unknown engine to me, leaving it analyzing the position for some 24 hours, and play the chosen move at time limit.
You have to use the same engine for the whole game.
You can switch to a different engine each time a new game starts.
You have to disclose what engines you used at the end of the match.
No time forfeit (but we're expected to play within 24 or 48 hours - Don't worry, I have never abandoned a match).

Nowhere does it says that I have to explain my plans, my methods, or post my analysis per move (which would take unreasonable time to me, anyway, organizing all my lines in a readily manner), if you wanted to negotiate that, you had to do it before the match started.

If you want to cancel this, resign, I refuse to abort it since I've so far complied with the conditions we agreed for the match.

>I run my SP engine for say 10h {most I have used so far}


That's your decision, you could have used up to 216 hours on all previous moves with your engine, if you chose to stop at 90 that's not my problem.

> you can clearly hide behind this and claim it was all your own work - do you see this?


I thought we were going with the honor system in this? You could also hide that you're really using a 12core, or many, or switching engines every move, or highly interacting with the position, or receiving moves from WBCCC winner ppipper instead of what we agreed. This is a matter of trust, I trust you're using the same SP engine for all the moves in Infinite Analysis, and you have to trust that I'm doing what we agreed I'd do. This claim that I may be leaving Houdini analyzing for 10 hours on big hardware to pick my moves is ridiculous (at least as ridiculous of my claims about your possible actions in the start of this paragraph), since I usually don't let an engine analyze the same position for more than one minute, instead, I interact.

I find your posture unfair considering I'm not demanding any proof that you're doing what you claim, I'm not even checking how much time you take to post a move.

You proposed 10 games and I accepted, so stick to them.
Parent - - By ppipper (*****) Date 2012-01-22 22:40
Hey :smile:
Parent - By Uly (Gold) Date 2012-01-22 22:46
Heh, just wanted to put in there "or receiving moves from a strong corr chess player instead of what we agreed", but thought of you, though, I know you'd never play like Bf1-Bb5-Be2 :wink:
Parent - - By Razor (****) Date 2012-01-23 05:39
Thanks Uly for your feedback, it looks like my understanding was correct.

This arrangement is not what I thought and we certainly won't be able to prove you can steer a game how you want against any chess engine thinking for up to 24h without first telling us what you are trying to do and why.  For me its all about evidence; being able to walk the walk is everything.

I wish you all the best in your future chess endeavours.
Up Topic Correspondence Chess / Correspondence Chess / Machine Vs. Centaur Challenge (I/X)
1 2 Previous Next  

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill