Originally posted by KeplerMore dishonest deflection. It begs the question of why you do this.
I thought the object was to stop cheats no matter where. Does cheating not matter now provided it isn't done in the championship?
It matters less, obviously. The time of the game mods is not infinite, so obviously they cannot test all engines with all possible settings and should therefore concentrate on higher ranked players.
Originally posted by PalynkaBut you have already implied that some engines are unsuitable for the purpose. It may well be imprudent for the game mods to post details of what they do here but the game mods rely on others to pass on suspects. If we do not know which engines are unsuitable we may well fail to do that initial work satisfactorily. It would be a tragedy if cheats were escaping just because those who do the initial detective work were inadvertently using faulty equipment. If you have reason to suspect that some engines are not suitable for the purpose at least PM that information to those who we know do pass on suspects. SquelchBelch and no1marauder would be good choices to start with.
I'm pretty sure they have much more information than me to choose which engines they test, with what time controls and how many choices. I'm fairly sure they are not as one-dimensional to believe there is one unique engine with one unique set of controls that is always best.
The reason they don't post those details here is clear, if you share my opinion. Because it would make it easier for cheaters to go around it.
Originally posted by KeplerSeriously, CAN YOU READ???
But you have already implied that some engines are unsuitable for the purpose. It may well be imprudent for the game mods to post details of what they do here but the game mods rely on others to pass on suspects. If we do not know which engines are unsuitable we may well fail to do that initial work satisfactorily. It would be a tragedy if cheats were escapin ...[text shortened]... o we know do pass on suspects. SquelchBelch and no1marauder would be good choices to start with.
No engine is "unsuitable" by design. A high match-up rate is a high match-up rate is a high match up rate. It could be with any engine. Even BattleChess. It's a question of finding the most popular ones among the top cheaters here and that cannot be done by analyzing tournaments.
Originally posted by PalynkaWhy I do what?
More dishonest deflection. It begs the question of why you do this.
It matters [b]less, obviously. The time of the game mods is not infinite, so obviously they cannot test all engines with all possible settings and should therefore concentrate on higher ranked players.[/b]
Originally posted by PalynkaYES I CAN!
Seriously, CAN YOU READ???
No engine is "unsuitable" by design. A high match-up rate is a high match-up rate is a high match up rate. It could be with any engine. Even BattleChess. It's a question of finding the most popular ones among the top cheaters here and that cannot be done by analyzing tournaments.
Ah, you missed the point. I am not trying to detect cheaters here on this site. I am not trying to detect cheaters at all, I leave that to others who know more about it than I do and have the time for such endeavours. I was asked if I had any evidence that the "obvious difference" between engines and humans is actually obvious. Once they are convinced that is possible then hopefully they will proceed to the detail of how exactly to go about the detection process. Again, this is not a detection process for this site. As far as I am aware this site already has an adequate system in place.
Some probabiliy fun:
You have a bag of 30 blue balls and 10 red balls. You pick 10 balls out of the bag at random. What is the probability of getting 5 or more red balls?
Answer: 4.98%. Odds of 19-1.
From the analysis: Top 10 on 1st choice: 5 of 10 engines, 5 of 30 humans.
Falcon 72.9
The Baron 66.7
Rybka 64.7
Hiarcs 64.6
Jonny 64.5
Anand V63.1
Fischer R 62.8
Aronian L 61.4
Svidler P 60.8
Rittner H 60.2
What is the probability of getting 7 red balls or more?
Answer: 0.0598%. Odds of 1670-1.
From the analysis: Top 10 on 3rd choice: 7 of 10 engines, 3 of 30 humans.
Falcon 91.5
Jonny 90.3
Rybka 89.9
Fischer R 88.7
Hiarcs 88.2
The Baron 87.5
Toga 87
Junior 86.4
Estrin Y 85.8
Anand V85.6
What is the probability of getting 8 red balls or more?
Answer: 0.00235%. Odds of 42,646-1.
From the analysis: Top 10 on 2nd choice: 8 of 10 engines, 2 of 30 humans.
Falcon 88.0
The Baron 82.9
Fischer R 82.2
Jonny 82.1
Rybka 81.8
Hiarcs 81.2
Toga 79.3
Sjeng 79.3
Aronian L 78.1
Junior 78.0
So are match up rates using one engine effective at detecting the use of other engines?
2 edits
Originally posted by KeplerI was mistaken for implying that you were trying to set a threshold of some kind to detect engine users on the basis of match-ups.
I am actually just trying to show that it is possible to distinguish between engines and humans. I expected the engine match ups to be higher than the human match ups. It is logical that engines should match engines at a higher rate than humans which is why i was surprised when my data came up with no statistical difference at all.
I agree that it is logical that engines should match engines at a higher rate than humans. I understand the distinctions between engines and humans very well but yesterday I somehow managed to rationalize why engines would not necessarily match engines at a higher rate than the strongest humans according to perceived and imagined flaws in your methodology.
I cannot remember the rationale I used but I had diagrammed my reasoning using X, Y, and Z variables. One of my mistakes as you noted is that I considered your test engine, Glaurung 2.1, (Z) as distinct from either population. It is distinct in that it is of different strength and style than the collective engine population but it may suffice in terms of your analysis. So I am not sure if some of what I had written in my previous posts is relevant.
I also thought of following analogy: Jack (X) and Jill (Y) have a DNA match-up 99.5% but it is not sufficient evidence that there is no significant difference between Jack and Jill – we already know that Jill is hot and Jack is not; Jack and Jill are very different people despite having a very similar DNA match-up. Jack is male and Jill is female but both have similar clothing styles and body strength. Compared with Aphrodite, (Z), we find that there is no statistically significant difference in DNA considering the DNA matchup of Jack (X) and Aphrodite (Z) with that of Jill (Y) and Aphrodite (Z) despite the fact that Aphrodite (Z) is representative of the collective Y population. My question is why Jack, Jill, and Aphrodite have such similar DNA match-ups (comparing X-Z and Y-Z matchups again) when X is male and Y and Z is female. One answer is because they are all human and such a small difference in DNA match-up can lead to significant difference in genetics.
I am not sure if the analogy works in terms in comparing how engines and humans match engines but it sure was fun. Perhaps a problem with my analogy is that moves of humans and computers differ much more than the genetics from males and females from a strictly statistical perspective.
According to Gatecrasher’s statistics, the hypothesis that engines match engines at a higher rate than humans seems to be true although his statistics do not take into the account that the engine population is stronger than the human population. The results do indicate a certain degree of variability within the human population [naturally] but also variability among engines is evident in his analysis. Increased variability in engines decreases the likelihood of engine-engine matchup; I think if we understood why this variability exists, we would understand why human-engine and engine-engine match-ups are similar if your result is correct and I think that your result is mostly correct.
There may be a significant difference in the rate that humans and engines match engines based on the fact that there are human and engine tendencies that are distinct to each population. However, despite the existence of these differences, there is certain amount of variability and overlap within and between these population. Like you, I am not completely sure if or how differences between the two populations manifest themselves statistically in match-up rates in considering X-Z (human-engine2 matchups) and Y-Z relationships (engine1-engine2 matchups).
Originally posted by GatecrasherI surmise that you are not implying that the two question that you posed above are related; I do not understand how the two questions are related, the first question a fun probability question that I did in high school, the second one something else entirely.
Some probabiliy fun:
What is the probability of getting 5 or more red balls?
So are match up rates using one engine effective at detecting the use of other engines?
My answer to your second question is that it depends on whether the engine has balls. If so, it is probably human. :-)
^^^ The real answer is clearly no. You don't even have to consider the the variability among the engines to answer that question.
Originally posted by no1marauderI agree but I don't know how high that ceiling is and I am not sure if anybody else knows how high that ceiling could be theoretically since no human plays perfect chess. I don't know if anybody has done a simulation which extrapolates the matchup of a human to an engine given that the human plays perfect chess.
It seems to me from everything I have seen that there really is a ceiling over which human players are not going to reach as regards matching up to engines on a consistent basis.
The highest matchups for first-move were Anand at 63% and Fischer at 62% in Gatecrasher's analysis. Theoretically the ceiling is slightly higher.
Originally posted by GatecrasherFor the probability of getting 8 or more I get 0.00234%, not 0.00235%
Some probabiliy fun:
You have a bag of 30 blue balls and 10 red balls. You pick 10 balls out of the bag at random. What is the probability of getting 5 or more red balls?
Answer: 4.98%. Odds of 19-1.
From the analysis: Top 10 on 1st choice: 5 of 10 engines, 5 of 30 humans.
Falcon 72.9
The Baron 66.7
Rybka 64.7
Hiarcs 64.6
Jonny 64.5
...[text shortened]... 78.0
So are match up rates using one engine effective at detecting the use of other engines?
Originally posted by YugaIt would be interesting to analyze the games of strong GMs that play CC such as the veteran grandmaster Ulf Andersson from Sweden. He is an active player at ICCF. Also, GM Rafael Leitao of Brazil plays there.
I agree but I don't know how high that ceiling is and I am not sure if anybody else knows how high that ceiling could be theoretically since no human plays perfect chess. I don't know if anybody has done a simulation which extrapolates the matchup of a human to an engine given that the human plays perfect chess.
The highest matchups for first-move were Anan ...[text shortened]... and Fischer at 62% in Gatecrasher's analysis. Theoretically the ceiling is slightly higher.