Pick any player from the top 20....

Originally posted by Zygalski
Obvious moves are also included in all the benchmark human games.
Why go to the bother of attempting to filter them out unless you think for some reason there might by far more in one large batch of games than another?
What subjective criteria do you use to say "well that's a fairly obvious move" when looking at a chess engine multi pv eval?
It all sound ...[text shortened]... inded. Especially when all the human benchmarks are so consistent at the highest level of play.

I agree that the benefit may be small, but so is the effort. It only requires a pre-analysis with a short horizon (which is much faster). The benefit is in removing the moves that have little value in distinguishing computer moves from human moves. This makes the differences in match-ups more pronounced, which would make the analysis a stronger argument in a cheater/not-cheater discussion.

I disagree that the criteria are subjective. I have explained before how to filter the "obvious" moves out in an objective way, so I won't repeat that again.

Establishing new benchmarks is probably hard work. I guess that's the biggest argument to keep the current method. I was merely brainstorming how to put more focus on the few clearly distinguishing moves between humans and computers. What would be your arguments if you'd have to start from scratch?

P.S.: which top-20 players did you pick eventually?

Originally posted by caissad4
David Tebb is a known OTB player and I have played him a number of times. I know you may not comprehend that a strong player can be seen by their games but it can. Have you ever played OTB or are you just another "internet chessplayer" ? 🙄🙄🙄

Oh ok.
So David Tebb is pitting his mind against multi-core engine users & staying afloat. Thanks for that.

ok can i pick mihai52, he is the only one i have played I think, I managed one draw and one loss, although it was ages ago when i think he was 1800 rated player and I 1600-1700.

robbie v mihai52 2009


mihai52 v robbie

Originally posted by robbie carrobie
ok can i pick mihai52, he is the only one i have played I think, I managed one draw and one loss, although it was ages ago when i think he was 1800 rated player and I 1600-1700.

robbie v mihai52 2009
[pgn][Event "Three Days"] [Site "http://www.timeforchess.com"] [Date "2009.01.26"] [EndDate "2009.05.02"] [Round "1"] [White "robbie carrobie"] [Bl ...[text shortened]... 7xb7 Qc7xb7 33. Qe2a2 Qb7d7 34. Ra4a7 Bc8b7 35. Ra7a5 Bf8e7 36. Ra5a7 Qd7e6 37. Ra7xb7 1-0[/pgn]

Yeh, i think bothering to analyse this guy is pointless. You only need to move his graph back to the beginning, and you can clearly see the date he started using an engine. You don't go from an average 1800 player to a 2300 in a straight line. I wish you could! But it just doesn't work like that...

It's not actually a straight line, but there is indeed a remarkable change in the rating profile.

Originally posted by caissad4
David Tebb is a known OTB player and I have played him a number of times. I know you may not comprehend that a strong player can be seen by their games but it can. Have you ever played OTB or are you just another "internet chessplayer" ? 🙄🙄🙄

Zygalski has played OTB but his rating has dropped off the ECF list due to inactivity since 2010. Time must have played a few tricks on his memory however as he has inflated his grade in his profile by nearly 10%.

I think it is a shame that someone who has decided not to play here for a number of years to choose to return in a thinly disguised attempt to discredit this site. Russ awards no titles, offers no prizes, he just so happens to have written code that provides one of the nicest interface to play casual correspondence style chess and with lively forums. If he has decided that he could not trust the game moderator system, if decisions about banning accounts for engine use were causing him sleepless nights when he would rather be in the pub spending subscription money I, for one do not blame him for giving up on it. We all know they come back with another name anyway.

Russ has not sold out to IM's etc trying hawk their wares, hosting their videos, like chess.com, where incidentally I believe the engine use is more prolific, and where the OP's opinion is has greater influence with the admin than here.

Everyone for whom it matters knows what goes on, can analyze the games if they wish and see for themselves. It happens to a greater or lesser extent on every site. That is the landscape you can choose how you wish to navigate it.

When this thread was opened there were about 12,500 RHP users in the rating list (who had therefore moved in the last 100 days). Since I have manged to win and draw games against players up to the top 100 here with less effort than I put into OTB play then I think it is fair to say that there are less than 100 who are blatantly cheating which is less than 1%. That is, in my experience, amongst the best on the web.

So far Zygalski has had to very little work since the vast majority of what he has posted here is the same as posted on the ECF forum this time last year, and constitutes positive arguments for his statistical system. I think that argument has been won and is certainly generally accepted. But, I have yet to see the results of a Benchmark Test for a computer on computer tournament in any of these threads which means one of two things. Either no one has done one or the results are not clear.

We are in a changeover phase where computers are now teaching the chess elite things about the game. You can hear people like John Nunn talk about database and tablebase mining which is where humans are trying to derive principles from best computer practice. This may mean that human play may become more engine like over time. Also it is possible to argue that as engines become stronger the best unassisted humans may achieve a lesser match up as computer play drifts away from human limits. It all depends on what you believe about chess truth, whether it is a fight, or if there is a sole correct move or by how much it adheres to mathematical principles.

Personally I don't care if David Tebb or anyone has decided to use an engine to assist him with his games. He makes his choice to play here for whatever reasons as do I. He has an OTB grade which tells us exactly how good he is, as do I. Others may not. My results here - getting a grade inside the top 500 when I am only inside the top 3000 in the National OTB list - proves that the vast majority of players here are casual and hobby players who are engaged in playing each other and don't care either.

Originally posted by tvochess
It's not actually a straight line, but there is indeed a remarkable change in the rating profile.

If you move forward through the graph you will reach a point where it literally is a straight line from 1850ish straight up to 2400.

Originally posted by Marinkatomb
Yeh, i think bothering to analyse this guy is pointless. You only need to move his graph back to the beginning, and you can clearly see the date he started using an engine. You don't go from an average 1800 player to a 2300 in a straight line. I wish you could! But it just doesn't work like that...

Ok perhaps I am being naive, I wanted his games analyzed to exonerate him for I don't think he was using an engine when he played me, no way.

Originally posted by Marinkatomb
If you move forward through the graph you will reach a point where it literally is a straight line from 1850ish straight up to 2400.

It's not literally (!) a straight graph, even in that period, as it still goes down occasionally. 😉

I cannot believe this. This is potty.

David Tebb a engine user...Why would he?

David is not facing strong engines and keeping afloat by using an engine.

David has a playing style of chess which is very much anit-computer.
No risk, no ½ baked attacks, simple ordinary chess which in the hands
of a very good player is very difficult to beat.
(sometims boring to me but I am so jealous that I cannot play like that.)

And every move he faces will not be an engine move.
Thanks to these threads on how to catch cheats these guys are now much wiser.

He is playing against blunder - checkers and stealth users.

You will need to explain (again) why we need the top 4 choices.
Any good player (and now we are talking about a very good chess player)
in any position will play a move a computer has permed any one from four.

Why not the one move a computer would play (the top choice) v the move a human played.
Surely that is the acid test.

Originally posted by Zygalski
Obvious moves are also included in all the benchmark human games.
Why go to the bother of attempting to filter them out unless you think for some reason there might by far more in one large batch of games than another?
What subjective criteria do you use to say "well that's a fairly obvious move" when looking at a chess engine multi pv eval?
It all sound ...[text shortened]... inded. Especially when all the human benchmarks are so consistent at the highest level of play.

Obvious moves are also included in all the benchmark human games.
Why go to the bother of attempting to filter them out unless you think for some reason there might by far more in one large batch of games than another?

Because if you filter out noise you get a clearer signal. You simply need an objective criterion for counting candidate moves which humans won't instantly reject. It wouldn't be difficult to analyse some old correspondence games and some games of trusted players here as well as of some known cheats to see what the distribution of score deficits they have is. You could then use that to build a model to distinguish humans and engines on the basis of the distribution of moves they make, which would be very hard to fake for an engine cheat.

Originally posted by greenpawn34
I cannot believe this. This is potty.

David Tebb a engine user...Why would he?

David is not facing strong engines and keeping afloat by using an engine.

David has a playing style of chess which is very much anit-computer.
No risk, no ½ baked attacks, simple ordinary chess which in the hands
of a very good player is very difficult to beat.
(s ...[text shortened]... a computer would play (the top choice) v the move a human played.
Surely that is the acid test.

Well if he is using an engine it could be because he is fed up with pazters firing-up 3200 Elo rated free download engines & zooming to the first few pages of player tables.
You say David isn't playing strong engines. I'd like to know what that is based on.
As I say, bans for engine use on this site ended about 3 years ago. It would be foolish to think that there are now less engines than back then, when the top 20 used to change on a weekly basis because a few idiots got banned.
If he isn't using an engine, then David is doing an incredible job. No doubt about that. A freely available engine like Houdini counters outdated anti-engine openings. It is designed to take calculated risks breaking out of closed positions which used to scupper many a program a few years ago.
Anti-engine tactics don't work nearly as well as they used to, is the long & short of it.
Personally, I'd rather see a strong OTB centaur up the top of the charts than a no-name like myself with a Megabase & several hours to kill with the latest version of Stockfish, so whether DT is getting assistance is a bit of a moot point. I'm certainly not about to re-analyse his more recent games from the last year or so, so no worries there.

The benchmarks & the suspect analysis both look for top 4 match rates because the games analysed are generally fairly evenly matched & contain many positions where there are several decent engine moves, all within a few centipawns of one another. Only an idiot would continually pick the top choice, and anyway, different engines on different systems naturally give slightly different results - thus the top 4 match rates are important, not just the top 1 move choice. Anyway, the top choice fluctuates as the depth increases so what could be top choice after for instance 20 seconds & 18 ply might not even be in the top 4 at 30 seconds & 20 ply.
I didn't really want to use this thread to teach the basics of match rate engine detection, but ho-hum...

Originally posted by DeepThought

Obvious moves are also included in all the benchmark human games.
Why go to the bother of attempting to filter them out unless you think for some reason there might by far more in one large batch of games than another?

Because if you filter out noise you get a clearer signal. You simply need an objective criterion for counting candida ...[text shortened]... is of the distribution of moves they make, which would be very hard to fake for an engine cheat.

It all becomes far too subjective.
When you sit behind a multi-core pc running Deep Rybka at a position and see a move that is +0.35 better than the others at 32 ply depth you might think "aha - an obvious better move" and discount it. An unassisted strong player may not see it as an obvious choice at all.
You only have to analyse a few Kasparov games to see that he fairly regularly ignores "obviously good" 1st choice engine moves for strong human plans.
Who decides what is an obvious move? What is the limit at which a move becomes obvious? Even a forced series of exchanges often leads to a slight advantage to one player. Should they be excluded too?
The only way to limit subjectivity is to remove it completely.

Originally posted by caissad4
David Tebb is a known OTB player and I have played him a number of times. I know you may not comprehend that a strong player can be seen by their games but it can. Have you ever played OTB or are you just another "internet chessplayer" ? 🙄🙄🙄

FYI
FM Stefan Docx has been banned from here & chess.com too.
IM Yelena Dembo got banned for engine use from chess.com & several other confirmed titled players have also been banned from that site.
Goodness knows why they do it. Why did Winona Ryder shoplift when she was a millionaire?
Answers on a postcard...

Originally posted by Zygalski
It all becomes far too subjective.
When you sit behind a multi-core pc running Deep Rybka at a position and see a move that is +0.35 better than the others at 32 ply depth you might think "aha - an obvious better move" and discount it. An unassisted strong player may not see it as an obvious choice at all.
You only have to analyse a few Kasparov games t ...[text shortened]... r. Should they be excluded too?
The only way to limit subjectivity is to remove it completely.

In general, if you do not filter data then it is much harder to analyse.

Assuming a human will not choose a move that is more than some threshold worse than whatever the machine thinks is best means that you have a criterion for excluding moves that are likely to generate false positives. It also gives you a basis for stating a confidence level for your match-up rates. You can generate that threshold by analysing previous games to see what the distribution of score deficits is and then work out a method of assigning a probability that the human would have found it.