Carlsen resigns on move 2

[Geoff Chandler]

In some ways that's not my major gripe though: it's the idea that you could somehow be "cued thrice" unproblematically, like it was as easy as that. How's it going to be done? The suggestion in the blog piece itself is laughable.

Hi Justin,

Without being to trace the prime source we can only guess at where Anand was coming from.
The 'how it was done' may have been covered in the original article.
At the top level these days it would very difficult, in the past before the glass screens and
players wandering about muttering to each other there will be many way to warn a player 'something is on'
.
If you want to advise a player 'something is there' a one off signal is all you need.
You do not have to get the co-ordinates to the player, him knowing there something on will suffice.
(it's strange how we always use 'he' when discussing a cheat. Has a female player ever admitted to cheating.?)

I suspect 'thrice' was pulled out of the air. (but as yet no proof - 'thrice' may have been mentioned earlier)
A player like Anand would only need it once. Here I am thinking of game 6 in the Carlsen - Anand game.
If a plant had warned Anand that 'something was on' he may have spotted and played 26...Nxe5 and won that game
taking a 2-1 lead. Instead he lost that game.

https://www.chessgames.com/perl/chessgame?gid=1778864

Anand states; 'The thing is when you’re not expecting a gift sometimes you just don’t take it.'
https://chess24.com/en/read/news/carlse ... ideo-recap

Which appears to translate as he did not expect Carlsen to leave such a move on the board, he trusted him.

[JustinHorton]

'
If you want to advise a player 'something is there' a one off signal is all you need.
You do not have to get the co-ordinates to the player, him knowing there something on will suffice.

What method of signalling do you propose

[Roger de Coverly]

But statistically, they are just data aren't they?

It's not the same data though, particularly in the early middle game where theory will have been extended. Also in the endgame, modern players have the advantage of being able to consult tablebases at some stage in their career.

Those of us who have perhaps varied between 1900 and 2100 or BCF/ECF equivalents for most if not all their adult playing career would likely find some of their earlier games unsound, particularly in the opening and early middle game. That's because knowledge has moved on.

[NickFaulks]

I said percentage rather than sigma because I was trying to avoid jargon. But does make sense to consider this way.

No, it makes no sense at all. The "table of probabilities" is based on Ken's output being normally distributed, but not only is there no reason to believe that it is, there is every reason to expect that it would not be.

[NickFaulks]

What method of signalling do you propose

I'm not going to say it is easy, or that I have any brilliant ideas, but it does feel doable with today's technology. In any case, would you accept that it is a far simpler challenge to have something go beep a couple of times in a game than to get the best move signalled every time?

[JustinHorton]

Oh, definitely. I am also sure that having something go beep is a lot more reliable than any of the hokey manual signalling methods in which I've seen people express their unwarranted confidence. (Hey, Grifter's assistant keeps walking into the room just before he finds a good move! And never any other time! Nobody's going to notice anything suspicious about that!)

But I also think that it's one thing being doable, and another thing working effectively over a substantial period.

[MartinCarpenter]

But statistically, they are just data aren't they?

It's not the same data though, particularly in the early middle game where theory will have been extended. Also in the endgame, modern players have the advantage of being able to consult tablebases at some stage in their career.

Those of us who have perhaps varied between 1900 and 2100 or BCF/ECF equivalents for most if not all their adult playing career would likely find some of their earlier games unsound, particularly in the opening and early middle game. That's because knowledge has moved on.

The overall level/type of play very likely has changed, yes but that isn't quite what he's trying to measure.

That's the likely hood of people massively outperforming their normal level for given event(s) etc. That should surely have stayed much more stable over time? People are still (hopefully!) people.

Actually, its obviously easy enough to check that the modern data has a similar distribution, I'd imagine that he's done that.

[MartinCarpenter]

Oh, definitely. I am also sure that having something go beep is a lot more reliable than any of the hokey manual signalling methods in which I've seen people express their unwarranted confidence. (Hey, Grifter's assistant keeps walking into the room just before he finds a good move! And never any other time! Nobody's going to notice anything suspicious about that!)

But I also think that it's one thing being doable, and another thing working effectively over a substantial period.

Yes, any form of human assistant is a really serious risk in the medium/long term. I can't really imagine anything using one keeping going for too long.

Full automation is difficult but not entirely implausible which will be getting people very worried if nothing else.

[NickFaulks]

When it is stated that someone is playing well, I no longer know whether this is intended to convey that they are matching top computer moves or getting good results. The two are of course correlated, but they are a long way from being the same thing - even at 2800 level.

[Roger de Coverly]

That's the likely hood of people massively outperforming their normal level for given event(s) etc. T

The weak point of the method is that it relies on knowing what a normal level actually is. The underlying theory of Elo ratings suggests that a "normal" level exists and is a Random Variable in the technical statistic sense and you attempt to measure it by using results. You compute rating change by using the statistics of how likely were the results of games against other rated players. If it was unlikely, you don't conclude that the player was receiving assistance, rather you conclude that the previous estimate of strength is out of date and correct it.

[Roger de Coverly]

. (Hey, Grifter's assistant keeps walking into the room just before he finds a good move! And never any other time! Nobody's going to notice anything suspicious about that!)

I do not think they did not notice though. There were whispers at least when it was alleged that something similar may have happened. That doesn't even need the sophistication of standing anywhere specific. Players may have seconds and historically at least they had access to phones, the press room and the playing area.

Contemporary report
https://en.chessbase.com/post/danailov- ... ng-topalov

[Mick Norris]

When it is stated that someone is playing well, I no longer know whether this is intended to convey that they are matching top computer moves or getting good results. The two are of course correlated, but they are a long way from being the same thing - even at 2800 level.

Exactly the issue, I'd like to think that playing well means making good moves unaided which may lead to good results

I've learned that "playing very accurately" seems to mean "likely to be consulting an engine somehow"

[NickFaulks]

Exactly the issue, I'd like to think that playing well means making good moves unaided which may lead to good results

Then I disagree. I think playing well means playing moves which are in practice too good for your opponent, so that you win. I don't care whether a computer can find flaws in them ( making them "bad" moves ), because you're not playing against the computer.

[Matt Bridgeman]

The latest Chessbase article should be added to the thread; https://en.chessbase.com/post/statistic ... ns-niemann

More anomalies being thrown up but who knows what it really means.

[NickFaulks]

More anomalies being thrown up but who knows what it really means.

Looks to me like another example of Nepo's comment - if you look for something weird you will find it.

Ignore his new variable STDCPL, which adds nothing to his thesis. He is saying that Niemann's ACPL is actually suspiciously high, meaning that his moves aren't good enough for an above average GM. How does that work and, if true, why hasn't anyone noticed it before?

By the way, according to these methods Carlsen's progress from 2400 to 2700 looks equally peculiar, but it all fell into line after that.