Cludi Deciding to Leave

Originally posted by Katonah
Well what Tricky chess said is absoultely valid. I happen to know a PH.D in mayhematical algorithms and what TC says is almost exactly what this professor said. The probablity of what the site is claiming is stasitcally far fetched. That it would take a very long time to prove up to 80-85 % any probability of cheating. Anyone else here have one of those degrees?

As it happens, yes. There would be a problem with the statistical tool if it were the sole evidence used to judge a player. As I understand it that is not the case, it is just an indicator of a need for further investigation. The human analysis of games is the most important element in the final decision.

Originally posted by TrickyChess
I hate to interrupt and create dissention among the ranks but I consulted a few PH.D's regarding the statistical rosetta stone the a certain player created and is used as damning evidence of cheating here at this site. Of these several scholarly individuals at several Uni's their agreement is this: The writing of such a program is trivial. Their contenti ...[text shortened]... less to even 80% to prove one player was a computer engine user. Sorry to burst a few bubbles.

Your post has to be understood before it can create dissention or burst any bubbles. Would you like another attempt.

The parts of your post that I'm having most difficulty comprehending are: "statistical rosetta stone the a certain player", "input the CC grouping asserted to here: Berliner to present as a control model", "a mass mean of probability of less to even 80%"

Originally posted by Dragon Fire
With Cludi being suspected I fear no one is above suspicion.

Unfortunately I have heard accusations on at least 4 other top 10 players at one time or another.

Surely no-one should be above suspicion. After all we are all just aplaying name on here.

How many people have actually met anyone in the top say 300 players and been able to verify their true identity?

To establish the integrity of the new game mods, surely they first have to be investigated, by each other and the findings presented to site admin.

They could then have regular checkups, every say 3-6 months, when say 2-3 games for each mod is investigated by every other mod to confirm their continued integrity.

With say 4 game mods this would mean between 6 and 9 games, chosen at random by site admin, would be investigated.

That way we can have more confidence in the Game Mod system.

Originally posted by adramforall
Surely no-one should be above suspicion. After all we are all just aplaying name on here.

How many people have actually met anyone in the top say 300 players and been able to verify their true identity?

To establish the integrity of the new game mods, surely they first have to be [b]investigated, by each other and the findings presented ...[text shortened]... admin, would be investigated.

That way we can have more confidence in the Game Mod system.[/b]

I agree that game mods should be investigated before approving to mod post and also after that each few (probably 6) months. And mod should be removed from its post if he is only becoming suspicious, even if evidence is not overhelming yet.

Originally posted by Gatecrasher
Your post has to be understood before it can create dissention or burst any bubbles. Would you like another attempt.

The parts of your post that I'm having most difficulty comprehending are: "statistical rosetta stone the a certain player", "input the CC grouping asserted to here: Berliner to present as a control model", "a mass mean of probability of less to even 80%"

Yes I was wondering about that too but just assumed it was over my head. I sort of understood that it was about requiring a large number of games to make valid comparisons...so many that it would take years to input them. But if that was the argument then it's correct in that it would take years to input so many games but overlooks the fact that this work has already been done and is widely available.

Originally posted by Mahout
Yes I was wondering about that too but just assumed it was over my head. I sort of understood that it was about requiring a large number of games to make valid comparisons...so many that it would take years to input them. But if that was the argument then it's correct in that it would take years to input so many games but overlooks the fact that this work has already been done and is widely available.

Thank you. I thought it was just me. 🙂

Originally posted by Mahout
Yes I was wondering about that too but just assumed it was over my head. I sort of understood that it was about requiring a large number of games to make valid comparisons...so many that it would take years to input them. But if that was the argument then it's correct in that it would take years to input so many games but overlooks the fact that this work has already been done and is widely available.

A large collection of master games would only be needed if the tool were comparing suspect player moves with other human moves. This is not the case here. As i understand it the statistical tool compares the player's games with likely engine moves by running the moves through the engine and seeing if Fritz (or whatever engine(s) is used) agrees with the player's moves. The only reason large numbers of master games would be needed is to check that the tool is doing its job and not suggesting that Capablanca had access to Fritz in the 1920s or that Deep Blue was not an engine. Those games are readily available online so it would not present a huge problem.

Originally posted by TrickyChess
I hate to interrupt and create dissention among the ranks but I consulted a few PH.D's regarding the statistical rosetta stone the a certain player created and is used as damning evidence of cheating here at this site. Of these several scholarly individuals at several Uni's their agreement is this: The writing of such a program is trivial. Their contenti ...[text shortened]... less to even 80% to prove one player was a computer engine user. Sorry to burst a few bubbles.

This is ridiculous. The amount of cheaters already caught flies in the face of arguments like this. Unless you want to argue that more than 20% of the "convicted" so far were innocent, which is clearly delusional.

Originally posted by Sydrian
You can not prove anything specifically, in a statistical sense. Statistics, by nature are a generality.

I don't agree with you.

There's a lot of garbage being written on this thread by people who clearly understand nothing about statistics.

Originally posted by TrickyChess
I hate to interrupt and create dissention among the ranks but I consulted a few PH.D's regarding the statistical rosetta stone the a certain player created and is used as damning evidence of cheating here at this site. Of these several scholarly individuals at several Uni's their agreement is this: The writing of such a program is trivial. Their contenti ...[text shortened]... less to even 80% to prove one player was a computer engine user. Sorry to burst a few bubbles.

I'm willing to bet you burst not a single bubble.

P-

Originally posted by murrow
There's a lot of garbage being written on this thread by people who clearly understand nothing about statistics.

Just over 62% of all statistics are made up by the person quoting them.

Originally posted by murrow
I don't agree with you.

There's a lot of garbage being written on this thread by people who clearly understand nothing about statistics.

Well actually your both right 😉

The "statistical" tool can ONLY be used for an indication on whether or not a player is using engine assistance. Why is that ?

We can use Fritz to evaluate a players moves and we can find out how many are 1. choice, 2. choice, .....

From Cludi's post I can see that the "statictical" tool developed gives a p-value (that is the probability that we will observe something "worse" ). The p-value(s) are closely associated with the confidence interval(s). So how do we determine the confidence interval(s) - and especially we need to find the 95%-quantile (or the 99% quantile some would argue).

To do that we must know the distribution of 1. choices, 2. choices, .... so how are theese numbers distibuted ? Could they be approximated to a normal distribution (maybee a Gamma-distribution)? Maybee they could. No matter what we need to make an assumption on the distribution of theese numbers. Otherwise we are not able to do any testing of hypothesis. (Maybee a large database could help here - I don't know .....)

But let's say that for example that the 1. choices are normal disributed with some mean value and varians. Then for every test on a hypothesis we must determine what parameters we will use for the mean value and varians in the model - and what should thoose be ? For strong players we might need to "push" the probalility mass towards the right because they are more likely to make better moves. And still we need to determine the maen value and varians.

Therefore in my oopinion it is impossible to develop a general tool, which can state a p-value for whether or not someone is using engine assistence. There are simply to many variable subjective factors:

- The approximation to a normal distribution (I think a Gamma-distribution would be better).

- Choosing the parameters of the distribution (theese have to differ depending on the players strengh).

- The choosing of the game sample to be investigated (strong players are more likely to make better moves against stronger opponents - and also to make better moves in specific tournys).

But it could surely be a good INDICATION on whether or not further investigation is neccecary.

With all this said we could have people using assistance for just a few moves - or for specific games - if you are smart enough I think it is always possible not to get caught.

So in the end it all comes down to the game mods. We have to trust their judgements (and I think all of us do), but the "statistical" tools can only be used for indications. To make a decission on whether or not someone is using engine assistance they have to look for "computer" moves and make a subjective deccision on whether or not the number of such moves could have been made without using computer assistance.

I have faith ...... Let's get a new super team up and running 😀

Originally posted by Richardt Hansen
The "statistical" tool can ONLY be used for an indication on whether or not a player is using engine assistance.

That's all statistics can ever do anyway. Statistical analysys cannot "prove" anything without uncertainty when there is randomness or disturbances in the data and samples are finite.

It can, however, provide us with evidence. Which is exactly what you admit that it does.

Originally posted by Richardt Hansen
Well actually your both right 😉

The "statistical" tool can ONLY be used for an indication on whether or not a player is using engine assistance. Why is that ?

We can use Fritz to evaluate a players moves and we can find out how many are 1. choice, 2. choice, .....

From Cludi's post I can see that the "statictical" tool developed gives a p-v ...[text shortened]... assistance.

I have faith ...... Let's get a new super team up and running 😀

I know one would only end up with a p-value (that is, a probability).

That is not what we were arguing about.

Sydrian believes that statistics CANNOT be used to evaluate specific allegations of cheating in certain games, only general patterns (see page 24).

Obviously (to someone with knowledge of statistics), the smaller the set of games, the higher the threshold of statistical significance. But there's no reason in principle why allegations relating to a specific selection of games could not be tested statistically.