Originally posted by ZygalskiYou have no evidence. I know enough of statistics to understand your numbers well. They are NOT a clear proove!
No amount of evidence I can produce will convince you, so I'm not playing your game any more.
They are an indication, yes. But circumstancial evidence is missing...
Sorry for repeating, but please think about this again:
This may sound trivial, but it shouldn't be forgotten. Our law-systems on most places of this planet are living with the thought 'if in doubt, give freedom'. This has taken humanity thousands of years to establish and it is a valuable good.
This eternal right - if in doubt, let go free - is actually so valuable, that even IF Skeeter was a cheat for real and it would be proven at some point, there is no reason to now behave less humane.
This value stands much higher then any cheap, dirty cheater of cc games. They are not worth it, to brake those values
07 Jan 12
Originally posted by tharkeshI keep seeing claims for this, from top GMs in casual conversations, but I've never seen it manifest itself in matchup rates myself. magnus carlsen has lover matchup rates than fischer or capablanca.
Good to know... how about the points I posted in the post on top of this page?
[b]Is this a tribute to Skeeter?
Isn't it true, that todays GMs/IMs also analyze and practice with engine help and that their game has become (naturally) more engine-like because of this?
Isn't Skeeter playing correspondence chess and (let's assume) an awesome chess ...[text shortened]... yer OTB, who apparently excels in cc?
Wouldn't that lead to very high match-up rates?[/b]
I think they're talking about different aspect of 'engine-like playing', namely the willingness for breaking classical rules more. (which happens maybe once in three games, as the position allows it.) instead of statistical matchup-rates like we talk about here.
Originally posted by wormwoodIt is interesting to know, that certain GMs believe this could play a role. And I would be extremely surprised, that it wouldn't affect a players style, if learning from childhood on with a computer program as support.
I keep seeing claims for this, from top GMs in casual conversations, but I've never seen it manifest itself in matchup rates myself. magnus carlsen has lover matchup rates than fischer or capablanca.
I think they're talking about different aspect of 'engine-like playing', namely the willingness for breaking classical rules more. (which happens maybe once ...[text shortened]... s, as the position allows it.) instead of statistical matchup-rates like we talk about here.
What you describe about Carlson is a nice example, too. In the end it is the nature of his mind and his willingness to follow certain things he learned and how he wants to use them. That's certainly also his 'style'.
I think that Kasparov has prooven in some of his games against Deep Blue that he can play in a way to defeat machines that he studied. This doesn't mean people can easliy now beat machines - but anyone can look more into engine play and find certain 'style' there, too.
One more question: what do you think would be Kasparovs or Carlsons match-up rate ONLY FROM CC GAMES?
Noone has ever been able to measure those - who knows, maybe they would be much higher?
07 Jan 12
Originally posted by tharkeshby your way of thinking, the earth cannot be proven to exist. because there's an infinitely small but existing chance of all its matter randomly quantum leaping, well, anywhere.
You have no evidence. I know enough of statistics to understand your numbers well. They are NOT a clear proove!
They are an indication, yes. But circumstancial evidence is missing...
there's such a thing as statistical proof. basically it means 'beyond REASONABLE doubt'. is it reasonable to think that skeeter et al. score consistently MUCH higher matchup rates than ANY past CC master or CURRENT OTB master has been found capable of?
well, as far as mathematical certainty goes, sure it's possible. but the odds of that happening with these unnecessary high margins of error? I think we'll run out of people on earth and time in the universe first.
"No amount of evidence I can produce will convince you,
so I'm not playing your game any more. "
No sulking....You are the one crowing she is a cheat and spent hours
gathering the info. You must allow us time to digest it.
Need those numbers I asked for mate.
Just those and you will see where I'm going......you may have something.
If you are not willing to allow your worthy work to be discussed then all you
have is series of numbers.
I could run those games through a ZX81 and get a new set of numbers.
You have to look at the games.
07 Jan 12
Originally posted by tharkeshcan't know until they'll start playing CC.
One more question: what do you think would be Kasparovs or Carlsons match-up rate ONLY FROM CC GAMES?
Shouldn't it be much higher as they have even more time to find a better move? Wouldn't that as well change their match-up rates (and as computers are beating humans now, more time to think = finding better move = closer to a machine move)...
but, it has been shown regarding other players that the increase in matchup rate is surprisingly small.
07 Jan 12
Originally posted by wormwoodLet me summarize a few points, which we can not exclude from being true for Skeeter:
...
well, as far as mathematical certainty goes, sure it's possible. but the odds of that happening with these unnecessary high margins of error? I think we'll run out of people on earth and time in the universe first.
1) GMs and IMs are studying and analyzing games with the help of engines, too. The style of engines influences the style of the student and may contribute to higher match-up rates (as believed by some GMs).
2) She is playing corresponding chess. As Gatecrasher says, some of the best CC players of the past had above average match-up rates.
3) We do not know the match-up rates of current GMs in CC. It can not be excluded, that the limits for match-up would have to be raised to higher numbers.
4) As you said, every statistic has its outlier, in particular when the starting conditions are violated they are more likely to happen. IF they happen, that does not mean, that they are not possible and have to be excluded because of low chances of appearing.
These are just four points, taken alone would not be enough, I guess. Certainly a combination of them is possible - all taken together give serious doubt. Not just little.
Originally posted by tharkeshYes that is true.
You have no evidence. I know enough of statistics to understand your numbers well. They are NOT a clear proove!
They are an indication, yes. But circumstancial evidence is missing...
What has happened is that I tested an objective batch of 20 games vs 2000+ rated players which all had 20 or more non-database moves & a program analysed them & spat-out the results about 14 hours later.
Then I compared the overall match rate %'s for top 1, 2, 3 & 4 choice moves & in skeeter's batch they happen to all be higher than those for any tested OTB Super GM (including modern players like Kramnik, Carlsen, Anand) and also higher than the old pre-computer era CC World Championship specialists who had many days to ponder each move.
There could be many reasons why skeeter's engine match rate %'s are so high.
Many more obvious moves than in any of the benchmark threshold games, a sheer fluke (get enough chimps at enough typewriters & one will type the complete works of Shakespeare) & probably a few others.
You do have to ask yourself what the most likely reason is for such an engine-like set of games is though.
There are other factors which help decide if a player has used an engine.
I'm not going to go into details, but these are circumstantial elements & also involve human judgement.
Match rates will always be the single most critical part of any case.
If the match rates are low you simply cannot ban for consistent, blatant engine use.
Edit:
You can't test for what rates the best modern GM's have in CC, simply because you cannot prove 100% that they haven't consulted an engine, so this is a moot point!
07 Jan 12
Originally posted by wormwoodThat is interesting to know. It shows that there was an increase in match-up rates. I wonder how big the increase would be for a Kasprov dedicated as much to CC on RHP as Skeeter was, allowing him games with 3/7 timeout... I am not mocking you, I am really curious. The top RHP players are mad to some extend, spending so much time on CC per day...
...
but, it has been shown regarding other players that the increase in matchup rate is surprisingly small.
07 Jan 12
Originally posted by ZygalskiHey Zygalski, thank you for that post! I am really happy to read there clearly, that you are strongly aware of the possible pitfalls of the system.
Yes that is true.
What has happened is that I tested an objective batch of 20 games vs 2000+ rated players which all had 20 or more non-database moves & a program analysed them & spat-out the results about 14 hours later.
Then I compared the overall match rate %'s for top 1, 2, 3 & 4 choice moves & in skeeter's batch they happen to all be higher than ...[text shortened]... se you cannot prove 100% that they haven't consulted an engine, so this is a moot point!
As you say, there are things based on human judgment and the like. I have done so, too and might have been totally wrong - and I might have burned my fingers, giving her the benefit of the doubt, even though I don't know for sure.
07 Jan 12
Correct WW.
We are not talking about Kasperov and Carlsen (actually one of their
chief OTB assets, especially Kasparov, is/was their presense at the board. One
of the reasons K lost to Deep Blue when he was slaughtering everyone else.)
We are talking about a middle of the road good player playing loads of
games against (hopefully) players of the same class playing loads of games.
Running one game through a box where a player has used a box should
catch them (though I would prefer if the hunter knew and used the same kit
as those being hunted.)
Zyglo believes in the numbers - I go for the actual games and see if I can
get a whiff of silicon. If Zyglo can give me the figures I ask for you will
see where I'm at.
I don't need 20 games. If I am convinced on just one game that a box has
been used then that is good enough for me.
I went through loads of Skeeters game looking for combinations.
(it was like pulling teeth.) I quit when I hit six. These games are new to me.
07 Jan 12
Skeeter played over 38202 moves in 1756 games here since May 2003.
User 34969
May 2003-Jan 2012 = about 91 months or about 2730 days.
38202/2730 = 14 moves per day.
That's not bad, considering it doesn't take into account work, holidays, sickness, internet connection problems & so on.
Ok, so some games will have been against very much lower rateds, plenty of obvious moves & so on, but she wasn't hanging around & always had at least a dozen games in progress, if I recall.
07 Jan 12
Originally posted by greenpawn34absolutely not!! it's not unheard of that even low rated amateurs score 100% matchup on a single game!
Running one game through a box where a player has used a box should
catch them
winning one time in lottery is good luck. winning 20 times in a row is cheating!
Originally posted by greenpawn34PM me your email address & I can send the full Batch Analyzer output file in Notepad format.
...If Zyglo can give me the figures I ask for you will
see where I'm at...
The file has each position with the output in the format:
20. c4 { c4 49 20 : Bb3 37 20 : 0-0 18 20 : cxb4 17 20 }
Where on White's 20th move they played c4.
This is [b ]bold[/b ] bracketed, which indicates where the played move is in relation to the top 4.
The score for the move was +0.49 and the score was reached at a depth of 20 ply.
The next best move was 20. Bb3, which scored +0.37, also at a depth of 20 ply and so on.
In other words, you can see how often the top choice moves were made when there were several other candidates which may have been within a few centipawns of the optimal move.
You can also get a very good idea of how many (or few!) errors there are, by looking at any moves which are sub-optimal & comparing them to the best engine score.