The recent durrrr -v- Isildur madness got me thinking a bit about the variance etc involved in these high stakes matches and tables which ultimately produce the 'stars' of internet poker. People see this guy come in and take durrrr for a few $M and assume instantly that he has his number and is the new king of poker.
I did some calculations based on the PTR data available. I know that this isn't a complete dataset and its only for NL (?) but its the best I have. I took durrrr's stats just for 5/1k tables where has has made $6.2M over 67k hands or so, for a winrate of $9253/100 hands over a sample of 67k hands.
I assume a standard deviation of about 50 big bets (100 big blinds) per 100 which is not unreasonable given his playing style and the variance in these big games. If you run that through a variance simulator for 50 independent trials of 67k hands you see the results in the graph below. Its quite clear if the graph is to be believed that he is running very well if that is his true winrate.



Do you think its the case that if he had run towards the bottom of his expectation that he wouldn't be nearly as famous and talked about as he is? Obviously he's a ridiculously good player and to build the roll to play these games in the first place proves that but I'm just using him as an extreme example that maybe the stars in the poker sky aren't really the best players and maybe there's a 10/20 or 25/50 grinder thats the best poker mind out there but runs terribly and will never be heard of outside 2p2.

why do morons still believe tableratings is accurate??

Paragraph 2, sentence 2.

it's generally 99% correct for pokerstars and full tilt. breakeven and losing regs think differently of course.

Why is it not possible durrr ran above expectation and now it's catching up?

Why is Guy considered a fish when he loses that amount in <30k hands but when durrr does it, it's potentially variance?

your right. All you can do is to do the best you can and hope to not run exceptionally bad. But thats not just poker, thats life.

i feel sorry for the trial that is -$. he must have run real bad

Haha....very true!

this, please make a new thread for that poor guy

Poker fame is definitely related to variance in tourneys especially (obv), see Jamie Gold, Jez Yang etc for tourneys.

If they didnt run hot who the **** r they??

True dat, but who knows what your life EV really is, and how far below or above you are. Technically speaking a lot of people might think there life is Shi*, but little do they know, they might be above EV.

The graph is a beautiful rainbow!

EV doesn't quite work that way... There's no such thing as catching up.

Also, the better and agressive the players, the higher the variance for both.
Obv since both play very agro its not insane to have 30+ buy in swings even if u got the slight edge.

can regress towards a true mean.

I believe that is what he meant.

His true winrate over infinite hands might be A.

His exhibited winrate is B.

Over time B will approach A...IF ENOUGH data is processed.

It isnt a seemless or constant process but....you get what I mean.

hey theres my line at the bottom.

fk you full tilt

because durr has proven himself over hundreds of thousands of hands and guy hasnt?

lol pokers

that graph looks like something my 2 year old nephew would draw with a couple crayons.

I understand (I think) what OP is getting at, and was thinking about that, too.

Basically, you take a bunch of equally skilled NL hold em players, one of them will eventually outshine all of them, if for no other reason than variance. Couple that with a a marginal skill factor, and you create the very rare nosebleed player.

i don't think you should just presuppose a value for standard deviation in this match-up. the standard deviation is a factor that has a huge effect on the results the simulator will give you. with both players being sick LAGs and isildur's constand river overbetting, i'd think the stanard deviation here could be insane.

someone could just get the hands into a HEM database and let it calculate the std. dev..

Randomness is random no matter how it was generated.

no one thinks its 100% accurate. however, none of the hands on TR are fabricated.

Please explain why you use sigma^2 = ~50BB/100 ? Without a good justification for variance value, all further analysis is useless.

Fun part is, there are some regs that play quite high and think they are the one. Some are sadly just the highest lines as shown in the graph.