Organizing a project to determine which sites are legit or rigged

[nineinchal]

[ QUOTE ]
[ QUOTE ]
What? This test couldn't be any more flawed.

[/ QUOTE ]

Well, tell us what could be improved in it then! [img]/images/graemlins/wink.gif[/img] That is the whole purpose of this thread, to get some discussion going about what tests could be done.

[/ QUOTE ]

That is not the WHOLE purpose, merely part of the purpose. My aim is to develop a device or method of determining which sites are offering a legimate gaming venue, and which ones are not giving a sucker an even break.

[helter skelter]

I think the first thing you would have to do is ensure the integrity of the data. Since you won't get co-operation of the sites for auditing purposes, the data sources should be limited to a perhaps a few long-time trusted members of 2+2, either from their own hands or their mining. They should be assured that no-one would see the raw data except the project manager and trusted statisticians.

Once that is agreed to, then you need one or more competent statisticians to analyze the data. You might find those in the probability forum. I think you need someone like that to determine whether sample sizes are large enough for whatever analyses you are doing and also to do the analyses correctly. Some of us (like me) can make mistakes in our analyses or interpretation of data because we don't really know enough about statistical methods.

Then you need to prepare and report the ongoing analysis in a coherent and meaningful way.

All in all, it would require a great deal of commitment to organize and manage such a project.

[nineinchal]

[ QUOTE ]
I think the first thing you would have to do is ensure the integrity of the data. Since you won't get co-operation of the sites for auditing purposes, the data sources should be limited to a perhaps a few long-time trusted members of 2+2, either from their own hands or their mining. They should be assured that no-one would see the raw data except the project manager and trusted statisticians.

Once that is agreed to, then you need one or more competent statisticians to analyze the data. You might find those in the probability forum. I think you need someone like that to determine whether sample sizes are large enough for whatever analyses you are doing and also to do the analyses correctly. Some of us (like me) can make mistakes in our analyses or interpretation of data because we don't really know enough about statistical methods.

Then you need to prepare and report the ongoing analysis in a coherent and meaningful way.

All in all, it would require a great deal of commitment to organize and manage such a project.

[/ QUOTE ]

I realize this. I am sure there is a lot more involved as well.

Keep contributing ideas such as this to move the project along.

[BradleyT]

This will be hard. In SnGs my AK never wins vs. 33 and when I get 33 I never win vs. AK - however the winrates for both AK and 33 would appear normal (because the other guy does win with AK and 33 putting both at 50:50).

[nineinchal]

[ QUOTE ]
This will be hard. In SnGs my AK never wins vs. 33 and when I get 33 I never win vs. AK - however the winrates for both AK and 33 would appear normal (because the other guy does win with AK and 33 putting both at 50:50).

[/ QUOTE ]

Nobody is saying it will be easy. All of those threads have already been written. Unfortunately many of those posts were just losing rants, blowing off steam.

The difficult part will be determining who's got a legitimate beef, and why. This task is equivalent to having DNA testing accepted as credible eveidence in a court of law. In the end, it will ultimately be the right thing to do, as well as save a bunch of time and effort for everyone.

[razarcrius]

Pardon me if im being stupid but doesnt observational evidence mean anything? I mean all you have to do is play on stars for a few hours to realize something is strange.

[MyTurn2Raise]

ummmm....I make too much money from online poker

please don't prove they are rigged

[nineinchal]

[ QUOTE ]
Pardon me if im being stupid but doesnt observational evidence mean anything? I mean all you have to do is play on stars for a few hours to realize something is strange.

[/ QUOTE ]

When you play a lot of winning poker, a lot ot things seem strange, even on stars.

More shall be revealed...

[nineinchal]

[ QUOTE ]
ummmm....I make too much money from online poker

please don't prove they are rigged

[/ QUOTE ]

Me too.

However, I lost a lot of money with online poker to make a lot of money on online poker; and I have my suspicions that not all sites are fair to the players...

[stompin]

[ QUOTE ]
[ QUOTE ]
Asking how many hands we need for such a study is the wrong question. First we should ask what kind of checks we would like to do on the data.

[/ QUOTE ]

That's a good point. What are you wanting to check?

If you want to know if the cards come out of the deck in the right "odds" as they are meant to that would be easy enough to do sample size wise.
However if you wanted to make sure certain players weren't "Rigged" then that would almost be impossible because you would never get a large enough sample size (unless they were silly enough to upload their HHs)

If someone was really worried about a person/group having all this "important" data then you'd just need to obscure the information before sending it The players names aren't important. So someone can write a really simple program that just replaces player names with UTG UTG+1 SB BB etc. People then run their hand histories through that parser. Then upload it to the central database.. Really does player names matter?
All you want is one HH file of every hand ever played that's gone to showdown right? people seem to usually complain about the showdown riggedness when there hand doesn't hold up. All of this information is available when observed.

[/ QUOTE ]

For certain types of analysis, like determining if there is such a thing as boomswitch / doomswitch or rigging to help fish survive etc, you would need to each player uniquely identified...because you'd be trying to determine how one player runs versus the population....or how a group of players with similar attributes (ie high vpip/high wtsd) perform versus the population. But the unique identifier doesnt have to be the screenname.