I'm a stats donk so I'm not sure how to interpret:

$/hour StD: $171.48
$/hr earned: $14.04

Over a decent-sized sample, how do I interpret these stats? Admittedly, this is over 100 hours so I know lolsamplesize, but I really don't know what these figures even mean.

Get all STDs looked at ASAP

The winrate is pretty straightforward:
Total $ / total hours

It means near nothing at 100 hours, but the fact that it is +$17 rather than -$25 is a good indicator that you will make it to hour 200 and beyond

The standard deviation (std # you posted) basically is a scale of how wide your results range in a given set sample period. The period most often used is an hour.

In theory, the majority of your hourly results would fit in between -$157 and $185. The distribution of results is by no means bound by -$157 and $185, it is just that the further a result falls from +$17 in one hour, the rarer it is. >99% of your results would fall between -3 standard deviations and +3 standard deviations. The distribution of these results would look like a bell curve.
** this is all for actual independent random data over big samples - 100 hours of poker does not actually meet either requirement... but in the absence of a better method - we will abuse sound statistical theory.

Results compounded many times over still will have a range of outcomes... these include heaters, downswings, etc... each specific result (like money made in the next 500 hours) - would have a probability associated with it. This is why these numbers are important when determining potential downswings / probability of that occurring / bankroll requirements. However, by the time any of us have enough personal data to make any use of it.. well, by then the value of knowing those numbers has passed (after you have played 10,000 hours - you don't need to know the BR requirements). So all of us are best off just going by the anecdotal advice of grinders past... play with 20 buy ins. If you lose it at NL - get better or run better.

(* hourly results are too fine grained to actually be normally distributed, but any distribution compounded many times approaches the bell curve)

By the way - don't follow BR advice at 1/2... just focus on trying to win AND play with an amount that you feel comfortable with.

3200 hours. Standard deviation is $1722 per hour. Win rate is 100/hour.

Currently in worst downswing ever.


Sent from my SCH-I545 using 2+2 Forums

Wow, you're awesome!

Home games are.

Sent from my SCH-I545 using 2+2 Forums

Stakes / BI?

The last year the game has been 10-25.

It's all plo and the cap on the buy in is 1k when we start. I know I know. Silly rule. As the chips fly around you can reload up to the big stack.

I rarely get thru one bullet. I'm usually in for 5kish. (And recently a lot more) The beginning part of my sample was much smaller 2 5 and some 5 10, mostly hold em.

The home game scene shifted a little over a year ago and now it's almost all plo.

The avg night there is 60k or so on table. The biggest session ever was last summer when there was 300k+ on table.

My biggest downswing was 17k in late 2013. Then 33k in summer 2014. Records are made to be broken and I just hit a 66k downer.

I'm in a 300 hour break even streak and my longest breakeven stretch came in 2013. 500 hours.

Sent from my SCH-I545 using 2+2 Forums

You know, I've just started teaching myself statistics and have been wondering about this. We really have no reason to believe that hours of poker playing are normally distributed--ESPECIALLY because if you are playing a single session each hour probably isn't even independent of the other hours (i.e. you or your opponents might play differently depending on what has happened previously), meaning the Central Limit Theorem may not even apply.

Is the reason we use normal distributions just because we know it's bad but don't have a better way?

Yeah - pretty much. Think of it like the difference between an engineer and a professor... at the end of the day, somebody has to build the ****ing bridge.

Yes. Mostly.

Sometimes we know the data isn't normal, but don't know any better way of dealing with it and just say "****it" because in practice (engineer here) it doesn't really matter.

Other times (far more often) people don't realize that the super basic statistics they think they know, about means and standard deviation, don't really apply but they use them anyway.

But this is a case where if you pretend the data are normally distributed, it predicts things that are patently absurd.

For example, let's assume a normal distribution with mean 14 and standard deviation 171.48. I think, if I am reading the z-score table correctly, that winning $300 or more in one hour, or approximately 1.66 standard deviations from the mean, will happen less than 5% of the time.

In a 1/2 game with $300 stacks, this would seem to predict that on average we will double up once every 20 hours. This to me is clearly wrong (I'd think it would happen much more often but be averaged out by losses in other hours), and so much so that we should not be using a normal distribution. To extend your analogy, we shouldn't be saying we built a bridge if it won't hold up a car.

^ You have to get more into central limit theorem and also how standard deviation applies for data sets without regards to distribution.

Even bound distributions compounded repeatedly turn into normal distributions. The only argument against poker (is not what one hour results look like) - it is the dependence of results on previous results.

I bow out here... in the end - this kind of info gets used incorrectly far more than correctly. (*probably include me in that too.)

"Yep, you run bad... buy in again" is the message that needs to get out there

I made $24.76/hr with a std. Dev/hr of $59,604 in a 900 hr sample at $1/2. Am I a maniac or is my poker income app degenerate?

You probably have corrupt data in there (like zero hour results and such)

DK ^ reorganize this as a histogram and then people will understand better.

Ah down to $133.95. Hourly dropped to $24.40. Embarrassing ****.

. Very solid results ^. Good stuff spike

Little uptick, getting close to 1000 hours at the RR $1/2. $2/5 going pretty well, recent results have been good, starting to nail down the regs' ranges. Current stats and giraffes:

I play in SoCal where the game structure is $200 min/max buy in with $3/5 blinds (Commerce), or $100-300 buy-in with $2/5 blinds (Hustler) so I have been forced to learn short stacking strategy since the lower level games are a joke ($6-7 DROP - not rake) per flopped hand.

Over 300+ hours I have a winrate of $40/hr (smaller sample size I know) but that may be due to the fact that the whole table is shortstacked all the time so it's not really apples to apples.

19 bb/hr over ~900 hours at 1/2 NL? Impressive wj, way too crush.

GjellyG

Not as hard as you think.

Have seen $40/hr in 20 - 60bb game during "daytime" game ONLY, not even in the juicy weekend and night games.

Commerce also has 5/5 $500 buyin fyi

There's just no way you're winning $40/hr at 1/2 over 1k hours, just impossible.