2NL: good winrates by position

[gabl136]

Hello!

I'm a beginner player. I've played 55k hands in the last two months playing 6+ tables with a winrate of 10BB/100+.

What I'm wondering is what are good winrate stats by position. I think 20bb/100 is possible overall, so i'm not yet satisfied, and stats by position would tell me what to aim for. Specifically, I'm losing so much on the blinds i'm embarrassed to say. Also, I lose a good chunk of what a win from the fish to good regs in all positions.

If there is someone with a big enough sample size and a good / crushing winrate please post a screenshot, that would help me alot.

[MadMojoMonkey]

Don't worry too much about this. It will vary from player to player.

There is nothing you can directly do to change your winrate, anyway. All you can do is to try and change your play. When you think about what you need to improve as a player, always choose things which you can directly control.

You can control when and where you play. You control whether you play when you're tilted. You control your own ranges and bet sizes.

You do not control your ability to win any given hand. You cannot control your overall winrate by any direct, single action.

It will stress you out more than necessary if you let yourself feel accountable for things which are beyond your control.

***
In the case that you were to fold every hand from every position, then you'd see this for your winrate
BB: -100 bb/100
SB: -50 bb/100
everything else: 0 bb/100

From the blinds, your "zero point" is already negative. When you look at the number that your tracking software tells you for these positions, be aware that they are skewed. They do not show your personal winrate directly. If you add back the 50/100 to the stated winrate, then you see your contribution to the winrate, ignoring the fact that you are forced to place a blind bet.

In general, these will be pretty low. In general, your highest winrate is from the BTN, and it decreases as you move around the table.

***
If you're winning 10 bb/100 overall, then you'll be moving up in stakes soon enough. Don't worry about trying to be the best pro at the micros. If you're crushing the micros, then you'll be (at least) winning the smalls. So there's no point in building 100s of thousands of hands at the micros just to prove you can have a maxed out winrate...
unless that's your thing.

I guess it could be your thing. I assume you're playing for money and moving up in the stakes, though.

[gabl136]

I know what i can and what i can't control. That's why i'm asking for positional stats. After i know what kind of winrates are possible i can rethink my ranges and everything else

[MadMojoMonkey]

I know what i can and what i can't control. That's why i'm asking for positional stats. After i know what kind of winrates are possible i can rethink my ranges and everything else

Whatever your winrates are, they will improve as you focus on your individual leaks, and not vice versa.

For a more direct answer to your question, if you're winning 10 bb/100 overall, then whatever your positional stats are, they're good.
A rock-solid pro, stooping to the micros to make a point, can have an overall winrate of ~20 bb/100.

Different players will have different winrates from different positions, but in general, the closer you are to the button, the higher your winrate.

So.... take your current winrates for each position and multiply by 2. If you were a world-class pro, then that's what you might expect to see after 100's of thousands of hands played at the micros.

Bear in mind that the variance on a winrate in NLHE is usually ~80 - 100 (bb/100)^2. So the standard deviation is 9 - 10 bb/100.
This means that you'd expect any winrate stat to fluctuate by 9 - 10 bb/100 if you ~tripled your data set.

In general, this is a huge swing in a winrate, as you can see that your winrate of 10 bb/100 could be statistically 0 - 20 bb/100 with a confidence of 68%.

EDIT: Not expect (see underline above), so much as "not be surprised by." Ultimately, we do expect the current value to be the "correct" value.

[OngBonga]

Pretty sure everyone is losing money on the blinds. Even then, if you're losing too much, then you're probably not folding enough.

Post a list of hands on the blinds, hands you consider tight folds, marginal calls, questionable decisions, etc, and see what others think. That's the best way to improve your winrate from the blinds imo.

[Eric]

Post a list of hands on the blinds, hands you consider tight folds, marginal calls, questionable decisions, etc, and see what others think. That's the best way to improve your winrate from the blinds imo.

Agreed, we need to see some hands.

[Schuppen]

Hello,

Love this answer MadMojoMonkey . I like this approach of how to deal with your winrate a lot.
I am an experienced poker player myself, know the things about floating , probe betting, using hand range with equity and whatever..But i have no knowledge of pure statistics(by that i mean standard deviation and variance and thinks like that)
My "win" rate is around -1BB /100... Definitely not that bad. Of course there are better players and a very lot of players who think they are better..
Anyway I would like to know more about it.. Can you maybe give me another example of those variable winrates ,
let's say 100 000 h and a winrate of 0 BB/100 with a confidence of 50 , 66 and 75 percent?

Also is it possible to make such assumptions with the EV line (i mean the yellow line in PT4 so not the effective green winrate line).

Thankx

[WeldPhaser]

yeah i said this to dude in other post


post hands man, and convert them..

[MadMojoMonkey]

Hello,

Love this answer MadMojoMonkey . I like this approach of how to deal with your winrate a lot.

Hello. Thanks.

Ultimately, my approach to winrate is to look at it after the fact and say, "yep, that's my winrate." It's going to vary so much over a sample of less than thousands of hands that it's simply not a strong indicator of my play.

Anyway I would like to know more about it.. Can you maybe give me another example of those variable winrates ,
let's say 100 000 h and a winrate of 0 BB/100 with a confidence of 50 , 66 and 75 percent?

When we're talking statistics about statistics (not a typo), we need to be very clear about what question we're asking. I'm not sure what you're after, here.

If you say the winrate is 0 bb/100, then I expect it to be 0 bb/100 after any number of hands. This means that no matter how wiggly the line of your winrate in a graph of winnings vs. hands, it is mostly flat.

If your "true" winrate is 0 bb/100, and you want to know what the expected value of your bankroll will be after 100,000 hands, at various confidence intervals, then:
@50% CI -> 0 +/- 1,908 bb
@66% CI -> 0 +/- 2,699 bb
@75% CI -> 0 +/- 3,254 bb

Notice that the expected value is always the "true" value - in this case 0 bb. Notice that as your confidence goes up, the spread in the value goes up. If you want an precise answer (with small +/-), then you get a very low confidence interval. If you want high confidence, then you get a very wide answer (with large +/-).

Also is it possible to make such assumptions with the EV line (i mean the yellow line in PT4 so not the effective green winrate line).

Thankx

Yep, the math is not specific to the cause of the input. It's just taking the data points and quantifying the spread in them. Assuming randomness means what we assume it means and we don't dig too deep into that (lalalalala, random is random, cause my brain accepts it w/o question, lalalalala) - ahem - assuming that, then it doesn't matter whether it's dice or cards or people making choices or whatever.

We're just quantifying our lack of understanding and making predictions which stand up to a formal mathematical description. This formal description, thankfully, describes many observable phenomena and has proven to be quite robust at stating its limitations explicitly.

[Schuppen]

We're just quantifying our lack of understanding and making predictions which stand up to a formal mathematical description. This formal description, thankfully, describes many observable phenomena and has proven to be quite robust at stating its limitations explicitly.

Thanks for the comment Madmojomonkey.
So if i understand it correctly ( based on the the mathematical description @66% CI -> 0 +/- 2,699 bb (100 000h)) a poker player who has after 100 000 hands a "true" winrate of 0 bb/100 , can say that there is a good chance (66%) that his winrate will be between -2,7 bb/100 and +2,7bb/100 .
If you translate this to the mind of a poker player you can say it can go both sides .he can be a winner or a loser He just has to play solid aggressive poker , look out for leaks in his game and thats it. wait and see.
In reality a beginning poker player psychologically can think he is the best player in the world where in fact the cards just do the work.
I mean a player who don't play lots of hands or plays in casinos, can literally think for years that he is a fantastic poker player while his winrate may be -1BB/100
That's why i think there always will be fish
On the other side i think there are also break even players on the long run who have all the bad luck in the world in their first months of poker and eventually give up and hate the game forever

I think statistics help you realise that.
Plus more importantly , it gives you the tranquility to not to be too much influenced by a player his stack while i m sure people psycologically make other decisions just based on someone 's stack. As if they are intimitated by a big stack , you know

[MadMojoMonkey]

Thanks for the comment Madmojomonkey.
So if i understand it correctly ( based on the the mathematical description @66% CI -> 0 +/- 2,699 bb (100 000h)) a poker player who has after 100 000 hands a "true" winrate of 0 bb/100 , can say that there is a good chance (66%) that his "apparent" winrate will be between -2,7 bb/100 and +2,7bb/100 .

@first underline: This "after" should be "throughout." This number is the same for every hand that was played throughout the 100k hands under scrutiny. Which makes sense, because most players talk about their winrate as the total of all hands played throughout the period and as such assume that they haven't made any changes or adjustments to their own play which were "permanent."

This is the core of why focusing on winrate is not too much help. If you're playing at the micros, you should be constantly improving your skills and making permanent changes to your play which are designed to eliminate leaks.

Note that we still had a pretty wide spread at +/- 2.7 bb/100 after 100k hands. Most people don't ever play that many hands at a single micro-stake. There are so many simple improvements to make as a player that will have your bankroll and skills ready to move up out of the micros in such short order that playing 100k hands at what could never be called a "living wage" is a waste of time.

***
It is naive to look at 2 data points to deduce the the slope of a wavy line. If all you have as data is your starting BR, your ending BR and the number of hands you played, then this is the best you can do. However, you have the graph of your BR on a hand-by-hand basis. You can take ALL the data points into account and do a "linear regression" to create a "linear best-fit line" and look at the slope of that line. This is a much more robust estimate of your winrate.

***
I added the 2nd underline to distinguish that the apparent winrate is distinct from the true winrate. I'm pretty sure you get this (which is a surprise, because most people have a million questions the first time I bring up this stuff).

[Schuppen]

It is naive to look at 2 data points to deduce the the slope of a wavy line. If all you have as data is your starting BR, your ending BR and the number of hands you played, then this is the best you can do. However, you have the graph of your BR on a hand-by-hand basis. You can take ALL the data points into account and do a "linear regression" to create a "linear best-fit line" and look at the slope of that line. This is a much more robust estimate of your winrate.

[/QUOTE]



Yeah you made it very clear for me. So super; . I would like to know more about statistics . Just out of interest and to look at poker from another point of view. Maybe i'm going to follow a course in the future.. To be honest i focus too much on my winrate but i 'm really trying NOT to do that.

So a linear best-fit line is a graph to define your "apparent" winrate and it is a more precise estimation because you have more data if i inderstand it correctly. And what do you mean with the data points ? Can you say that one data point is one hand in correspondance with a value expressed in dollar? So in fact, is the linear best-fit line based for instance on the green line in PT4 so the profit on the Y as and the number of hands on the x as ?

[MadMojoMonkey]

I would like to know more about statistics . Just out of interest and to look at poker from another point of view. Maybe i'm going to follow a course in the future.

Do it! It was tough, but not bad compared to other math courses. The stuff you learn is applicable in every field, though.

So a linear best-fit line is a graph to define your "apparent" winrate and it is a more precise estimation because you have more data if i inderstand it correctly.

You do.

And what do you mean with the data points ? Can you say that one data point is one hand in correspondance with a value expressed in dollar? So in fact, is the linear best-fit line based for instance on the green line in PT4 so the profit on the Y as and the number of hands on the x as ?

Yes.

Here's how to do it w/o any math:

Open Excel.
In A1 put a 0
In A2 put a 1
In B1 put a 0
In B2 put "=B1 + 0 + 0.2*RAND() - 0.1" (but w/o the quotes)

Spoiler: Show

The + 0 part is the true winrate we're simulating. If you want to simulate a winrate of 2.7 bb/100, then you'd type in "=B1 + 2.7 + 0.2*RAND() - 0.1"
The +0.2*RAND()-0.1 part is to simulate a standard deviation of 10 bb/100. That is, +/- 0.1 bb/hand.

Now highlight A2 and B2 by click and drag. The next step is a separate click and drag.
Click and drag on the lower right corner of the highlighted bold boundary box (it should have a little square on it) and drag it down as far as your heart's content.

Go to the column headings and click and drag across A and B. This should highlight the entire columns.

Insert-> Charts -> Scatter -> (the one with jagged, straight lines)

It gives you a graph with a jagged line. This is meant to simulate a BR of a player with a "true" winrate of 0 bb/100. See spoiler above to change the value of the simulated true winrate.

Right click the jagged line and select "Add Trendline"
It opens a window, the only thing you probably want to do is to click the check box next to "Display Equation on Chart"

Voila! You have a linear regression of a plot of your BR vs. hands, and the slope of that best-fit line is an estimate of your apparent winrate. If we were to put error bars on that slope, then we'd have a robust estimate, but we're already leagues closer than just looking at the ending value of the BR and the number of hands.

Now mash F9 a bunch of times (this tells Excel to recalculate the sheet, and it will re-roll the random numbers)

Look at all that variation in outcomes! That's simply expected variance in your BR.


All you have to do now is replace the simulated outcomes in column B with your actual BR after each hand (which your tracker program must have somewhere in order to make the green line). See if you can copy paste that data into column B and make sure you have a number in column A for each entry in column B. Using click and drag to fill in like you did above is quick and easy.