Life-changing Variance

my graph looks exactly like urs. and damn whend u move up to 2/4? i always see u grinding them 1/2.

I shot took and immediately got raped, don't worry I'm still at 1/2 with my gigantic roll because I'm scared to death of 400nl =[

Food for thought for sure

Why is everyone running so bad?

In your lifetime you may never get to the long run.

This is a good blog, but I'm confused why you listed the # of hands inaccurately. You say 300k, but the graph says its domain is 1435. Your graph has the same issue... You say 90k hands and the graph says 336 for the domain. What's up? Off the top of my head I can't even think of a likely conversion factor that turns 1,435 into 300k and 336 into 90k.

Anyway, if you're trying to understand the long run as anything other than an infinite sample you'll probably need to specify what realm of probability will satisfy your sense of accuracy. In the most extreme case, I don't think it violates any law of statistics to say that you can run good or bad over any finite sample. Outside of an infinite hand sample, any sense of "long run" where your EV converges really has to be understood as something like "over a million hands, there's a X% chance that my winnings will be within |Y|$ of (actual $EV = observed $EV).

Of course with such a large volume of players, some will run surprisingly well. There's about a one in a million chance of winning 19 flips in a row. So exactly that often, someone is gonna win 19 coin flips in a row. There's no hammer of justice in probability that is gonna then locate the guy and then make him lose 19 other flips in that same million hand sample just to even things out. Outside of those 19 hands, that player may run exactly "right:" his observed EV = to actual EV for rest of the sample.

And yeah, being that guy would be awesome, and you would be overconfident and have an inflated sense of your abilities because even though you were actually (for the sake of argument) a marginal winner, you won 19 or so more BI and had some serious extra cash. On the other hand, it's entirely possible that you could run entirely below expectation for the rest of your life, or that a meteor could blow crash into your bed while you were sleeping there, or that the atoms in your wall could move in synchronization so that it actually waved at you. Dwelling on extreme probabilities is a lot of fun as long as you don't actually believe any of it

In your life time, you'll never get to the long run.

Failsafe, the sample of 90k only has 336 hands that went all in preflop. Same deal with the other chart. I felt a need to describe the real sample before someone goes "lol only 1k sample wtf is this".

Thanks for the comments though. The issue with this is that yeah there's all sorts of things we can look at in the field of probability with regards to infinite numbers and other such things that I don't understand that well, but in reality I just find that it's a testament to this game being truly sicker than we expected (at least on the micro level). In 2005 no one cared about million hand sample sizes because the games were so good that variance didn't seem to matter as much. Nowadays the games are harder, smaller edges are contested, and variance increases.

It's just hard for me to accept because of all the material I read when starting out in 2006 or whatever, thinking that there was some justice down the road. I've played about a million hands in my career which by no means is really that much but it's disgusting the swings that can seriously occur. Especially since the mantra of "longterm" had been touted even before online poker (Sklansky etc.), when you consider that live professionals will probably never even get to play I don't know 500k hands in their entire life.

On February 23 2009 22:02 failsafe wrote: This is a good blog, but I'm confused why you listed the # of hands inaccurately. You say 300k, but the graph says its domain is 1435. Your graph has the same issue... You say 90k hands and the graph says 336 for the domain. What's up? Off the top of my head I can't even think of a likely conversion factor that turns 1,435 into 300k and 336 into 90k.

filtered to show only hands that got aipf

EDIT too slow

but they get to play endless fish worse than 2005

yeah, long term is way longer than most ppl think.

sick man sick

what program did you use to calculate the allin ev?

it really only depend on ur edge on the games u're playing as well

the long run is all hands played imo

On February 23 2009 23:27 Fayth wrote: it really only depend on ur edge on the games u're playing as well

that is totally true when discussing overall winrate. but the graphs that were posted show only AIPF coinflip scenarios. While AIPF coinflips are only a small snippet of overall play, it has absolutely nothing to do with edge and is 100% variance.

there is a phenomena that applies IRL as well as poker.

losers tend to overestimate the impact of variance and underestimate the impact of skill while winners tend to underestimate the impact of variance and overestimate the impact of skill

just human nature and the reality is somewhere in between.

and with the first graph, it works out effectively as you flipped a coin 336 times, you lost 180 and you won 156. Definitely unlucky, but not that unlikely. In fact, I think the probability of that outcome is something like 18.5% so you can consider the cumulative result over the past 90K hands AIPF as a single lost overpair vs underpair.

On February 24 2009 00:27 TenBagger wrote: there is a phenomena that applies IRL as well as poker. losers tend to overestimate the impact of variance and underestimate the impact of skill while winners tend to underestimate the impact of variance and overestimate the impact of skill just human nature

That sounds like my stance on Innate talent vs acquired skill / work ethic etc --________--

sick blog. makes me hate that i only play 4 tables so if i have a bad 100k hands thats like 4 months =[

that + still having to cash out = gg owned me.