Was using 25 hands per hour btw, and honestly when you think about it how terrifying, 7 years playing every day and your sample results can still mostly be determined by variance.

To expand a bit. I don't see a reason to be generally pessimist, like I was bound fo the worst outcome (and of course not the best either). I find it reasonable to expect I'm not among the 1/3 unluckiest. So one could subtract the outliers here and get what to expect. For 500k that would mean -2.0 to 1.4, that's a difference of 3.4bb/100. Still substantial, but kind of manageable.

Out of curiosity...

Consider a regular game with a group of six friends, five of whom enjoy poker but don't really take it all that seriously. No one is an absolute noob mind you, but also no one has really advanced much farther than being at best a break-even $1-2 or $1-3 NL player. Meanwhile, Player A is the only one in the group who has studied the game, takes it somewhat seriously, etc. No one would look at his play and think, "Oh wow, this guy should be thinking about the f-ing Mirage" but let's say he has put in enough time at the online and live tables to comfortably conclude that he's a 5BB/100 (after rake) in the live $1/3 NL games.

TL;DR In a regular six-handed game, Player A is a proven 5 BB/100 winner at $1/3 NL, while Players B-F are break-even or losing players at a similar stake.


Let's further define the term "highly probable" as having a 85 percent chance or better. Why that number? Because having no idea what Mason meant by it, I did a quick Google search for "highly probable percentage." The first result was some random LinkedIn article from a guy who cited two separate studies examining Korean and Australian probability terms in financial statements. (Whaddya want, it's half-assed internet research.) I chose the Australian number of 85% for "highly probable" because, well, K-pop blows.

Now, if you're the type who has NewJeans or BlackPink posters up in your man-cave, then by all means, go with the Korean standard of 90 percent. I think I just got Bart Hanson's attention.

Given the above, here are three questions:
1. After 3,500 hands, what is the approximate probability that Player A is ahead, even if by one dollar? (Note: I will accept either the Australian or the Korean standards.)
2. How many hands would need to be played for it to be "highly probable" that Player A is ahead, even by only a single dollar?
3. How many fewer posts would be in this thread had Mason nodded to Anders Ericsson by saying 10,000 hours rather than 100?

Many, many years ago when Full Tilt was a legitimate thing and Poker Tracker was accepted as normal. I had a 300,000 - 500,000 hand plateau where I was literally a breakeven player after normally crushing the games.

Looking at poker tracker it was clear. Over that range of hands, I was a net negative at playing flopped sets.

I wasn't stupid. It wasn't like I was regularly blasting off money on mono suit boards. It just seemed that everytime I flopped a set, I would lose to a bigger set or the flush/straight draw would chase and get there.

TDLR, I was extremely unlucky whenever i flopped a set. For almost 500k hands. For 6+ months I lost more than I won when flopping a set.

It happens.

Luckily i was crushing the games enough that I was breaking even otherwise even though I couldn't win with a set.

The whole episode made me realize just how much luck plays a part in this game.

I often think that there is a player who is a crusher in poker. They might be an elite level thinking who is great at reading their opponents and adjusting to them, yet he is a breakeven to middling player simply because he is literally the unluckiest player in poker.

I feel sorry for that guy.

Right.

By definition there are people who's results are a ridiculous number of standard deviations better or worse than their skill.

Obviously by definition these people are extremely rare. Almost a unicorn. But they do exist. Simply because of the immense number of players playing poker.

Do you disagree?

Almost always does not equal always. Outliers may be outliers, but they exist. Statistics demands it.

I am pretty sure the takeaway from the OP is that varian6is greater than you think.

"Ridiculous number of standard deviations" really doesn't make much sense because the probability will be so small.

Mason

Without knowing what the standard deviation is, you can't answer any of these questions.

Mason

One thing to point out is that even though you played a lot of hands, your sample size for flopping sets isn't large. Specifically, you first have to be dealt a pair that you play and then you have to flop a set. An estimate for this is to flop a set approximately once every 150 hands. So in 300,000 total hands, you're looking at a sample size of approximately 2,000. And since flopping a set produces a hand that has high variability, while 2,000 is usually a large enough sample size to draw conclusions, it may not be in this case.

Mason

Another thing to point out in this simulation is that the six bots are all playing against each other which means their results are not independent of each other. That is if one bot wins, the other bots, as a group, must all lose to account for the winning bot.

It would have been much better to just track the results of a bot in a regular game and then based on his results to get an estimate of the standard deviation.

Mason

I think the whole point was to just show just how many hands you need yo reach your expected value, expected value of bots are 0bb in this case yet one of them had 6.5 after 500k hands

That would require infinite hands.

But I think we'd already established that running 6.5bb over/under EV over 500k hands is an extreme outlier:
1 in 500,000 with a std.dev of 100bb/100
1 in 1000 with a std.dev of 150bb/100
1 in 100 with a std.dev of 200bb/100

Except that looking at results that are highly correlated with each other won't necessarily show this.

Mason

what sites/screen names does Matt Marinelli play under?

did your parrot write this?

ILuvAvrilLavigne91 on ACR

Aren't live win rates insanely higher than online? If that old saying that the best regs at a given stake are winning 10bb/hr is true that would translate to 40bb/100 assuming 25 hands/hr. Obviously this "cap" starts to be lower at 5/10 and higher, but for a "good" reg at 1/3 or 2/5 surely they would be winning at 25bb/100 at least even assuming a slightly higher hands/hr with auto shufflers. Then after 100 ish hours with 3500 hands you are very very likely (>93% chance) to be profitable. A "top" reg would almost certainly be profitable after 100 hours. I made the standard deviation a little higher than what was suggested for full ring cause live typically plays deeper than online with larger opens and straddles.

I'm not a data expert like others in this thread bur live guy can never play enough hands.

One thing being missed is that everything in life is determined by luck.
1% of people will die in a car crash
6% of people will get cancer
Yeah, a tiny % of people will way underperform their winrate over their entire career (millions of hands for online)

Your poker career is less determined by luck than so many other things that are part of games you don't choose to play

All of these vague pronouncements are silly because a knowledge of how standard deviations work covers these subjects a lot more precisely than most of these posts or individual experiments do.

https://www.amazon.com/Probability-S...s%2C649&sr=8-1

I think % of people who get cancer in their lifetime is actually close to 50

Huh? That is extremely disappointing. You should do better.

A person who really understands stats recognizes that small probabilities occasionally happen, especially given a large enough sample size.

Oh, come on. You are far better than this. Way better.

Don't let your stubbornness overcome your intelligence. Is isn't really worth losing credibility over?

Exactly.

I cannot believe this needs to explained to someone who has a background in stats.