PDA

View Full Version : The luck factor


me454555
09-01-2004, 11:49 PM
We all know that winning is poker is a combination of luck and skill. I've played fantasy baseball for a few years and there are some stats that quantify luck. I.E % of balls put into play by a hitter. I was wondering if anybody knows any way to quantify luck? A way to tell whether you're playing bad or just unlucky. Are there any stats out there that can do this?

Kopefire
09-02-2004, 12:41 AM
There is no "luck" in poker, period.

There are short term deviations the norm, but the act of playing poker is making decissions that maximize your earn rate given the information you have and the probabilities you know.

If what you are asking is something like "how do I know if I'm losing money because I'm making good decissions and just not hitting my outs?" Then there are plenty of formulas for that. The easiest one to use is to look at your hourly performance (track your win or loss total every hour) and you can easily calculate a confidence intervale around your earn rate.

Basically just take the list of hourly measurements (120 minimum) of your winnings or loses, and calculate the non-parametric reference interval (non-parametric because if you are a winning or a losing player then they measurements should not fit a normal distribution!)

That will give you the confidence interval over which you can find your expected earn rate.

If you are collecting data on-line, you can do polynomial regression analysis between your earn rate and your oponents earn rate.

It's easy enough to do this in excel . . just write down your win rate for the previous hour every hour . . .

me454555
09-02-2004, 02:23 AM
[ QUOTE ]
There is no "luck" in poker, period.


[/ QUOTE ]

Then why did that guy hit his 1 outer against me last hand /images/graemlins/wink.gif Why did I flop a fullhouse when my oppoent flopped quads?!?!?! I think thats unlucky.

[ QUOTE ]
Basically just take the list of hourly measurements (120 minimum) of your winnings or loses, and calculate the non-parametric reference interval (non-parametric because if you are a winning or a losing player then they measurements should not fit a normal distribution!)


[/ QUOTE ]

What does that mean? I'm not very statically inclined so I'm not really sure what you're talking about

[ QUOTE ]
If you are collecting data on-line, you can do polynomial regression analysis between your earn rate and your oponents earn rate.

It's easy enough to do this in excel . . just write down your win rate for the previous hour every hour . . .

[/ QUOTE ]

How would I do this? Again I don't quite understand

Kopefire
09-02-2004, 02:45 AM
[ QUOTE ]
[ QUOTE ]
There is no "luck" in poker, period.


[/ QUOTE ]

Then why did that guy hit his 1 outer against me last hand /images/graemlins/wink.gif Why did I flop a fullhouse when my oppoent flopped quads?!?!?! I think thats unlucky.


[/ QUOTE ]

Ok. think it's lucky. As long as you think that way you do not understand the game.

That will happen x %-age of the time in that situation. That it happened that way at that time is just the way the cards fell. It's expected. You will get bad beats, and you will make your quad on the river and everything in between. And it will happen, after you play enough hands, pretty much as expected. You already know the percentages up front. Your job at the poker table is to make your choices based upon the knoweldge of those odds.

If you do that consistantly, you win. If you don't you lose. Luck doesn't factor into it.

[ QUOTE ]

[ QUOTE ]
Basically just take the list of hourly measurements (120 minimum) of your winnings or loses, and calculate the non-parametric reference interval (non-parametric because if you are a winning or a losing player then they measurements should not fit a normal distribution!)


[/ QUOTE ]

What does that mean? I'm not very statically inclined so I'm not really sure what you're talking about

Every hour write down how much you won or lost in that hour.
Do that for 120 hours.
Take those measurements and compute various indicators to indicate the central tendancy of the data.

That means, figure out what the "average" is. You want to do that non-parametrically, however, because you can't assume a normal distribution. A normal distribution is a collection of data that you expect to have fit more or less to a normal curve. Because you expect that you are either a winner or a loser, you would expect that your data is skewed either to the winning side or to the losing side.

You can then compute the confidence intervale for that central tendancy. So you can say something like "I have a 95.5% confidence interval that my earn rate is between -$5.50/hr and +10.20/hr."

Depending on your actual numbers you can get a good sense of the probability that you are a winner or a loser in general.

Oh, the 120 measurements is a good guideline number to make sure that you have enough measurements to get a good CI. More is better, and you can work with fewer, but at 120, you've got enough amples that you can make some meaningful statements.

[ QUOTE ]

[ QUOTE ]
If you are collecting data on-line, you can do polynomial regression analysis between your earn rate and your oponents earn rate.

It's easy enough to do this in excel . . just write down your win rate for the previous hour every hour . . .

[/ QUOTE ]

How would I do this? Again I don't quite understand

[/ QUOTE ]

You would collect the data. Enter it into Excel and use the statistcal functions of the program.

Read the help files in Excel for formulas, they're actually fairly good at explaining this stuff if you have a basic background in statistics. If you don't have that background, then pick up something like StatKing from ConJelCo.

pzhon
09-02-2004, 03:50 AM
[ QUOTE ]
[ QUOTE ]
[ QUOTE ]
There is no "luck" in poker, period.


[/ QUOTE ]

that guy hit his 1 outer against me last hand ... I think thats unlucky.


[/ QUOTE ]

Ok. think it's lucky. As long as you think that way you do not understand the game.

[/ QUOTE ]

You must use a nonstandard definition of luck. Why insult the understanding of everyone who uses the usual definition?

Do you have a better term for the short-term fluctuations?

me454555
09-02-2004, 11:21 AM
The point of the post was that I DON'T have a baseic background in stats. I'm lost on that stuff. I needed help so I made the post. Can you explain it to me or not?

Kopefire
09-02-2004, 01:46 PM
Pick up a copy of Statistics for the Utterly Confused) by Lloyd Jaisingh (http://www.amazon.com/exec/obidos/tg/detail/-/0071350055/qid=1094147013/sr=8-1/ref=pd_ka_1/102-5396522-5690548?v=glance&s=books&n=507846) it's a very good book for the kind of stats you're interested in figuring out. It explains things in enough depth that you'll be able to understand what you're doing with the numbers even if you have no statistical background at all.

I don't have a good reference text for Excel, I just use the help files. But I'm sure you could find something just browsing the shelves at your local book store.

pzhon
09-02-2004, 05:32 PM
[ QUOTE ]
I was wondering if anybody knows any way to quantify luck? A way to tell whether you're playing bad or just unlucky.

[/ QUOTE ]
It is useful to estimate the standard deviation of your results, and normalize your net results in terms of this. This is neither complicated nor esoteric.

If your records are broken into sessions of roughly equal length (and the same type of game was played in each session), add up the squares of the results. You can estimate your lifetime standard deviation as the square root of the sum. Your actual results are usually within 1 standard deviation of your expected results. It is quite unlikely to be more than 2 standard deviations away from your expected results.

The square root of the average of the squares of the session results is an estimate for the standard deviation for a session.

Example: You have played 10 sessions of $1-$2 for about 3 hours each. The results are as follows:

1: +20
2: +60
3: -120
4: +10
5: +70
6: 0
7: +40
8: -50
9: -50
10: +80

So far, you are ahead $60, 30 big bets in 30 hours. That sounds good.

The sum of the squares of the results is 36400. The square root of that is about 190. So, you are ahead less than 1/3 of a standard deviation. That means the evidence is very weak that you are a winning player. You need a lot more data to conclude that you are a winning player from the results alone. If you are a break-even player, typical results would be to end up ahead or behind $190.

The standard deviation for each session is about sqrt(36400/10) ~ 60. Winning or losing $60 in a session is typical.

There are more accurate ways to estimate the standard deviation if your sessions have different lengths. See this 2+2 essay (http://www.twoplustwo.com/mmessay8.html) on estimating your hourly standard deviation.

LinusKS
09-02-2004, 10:54 PM
Aleo Magus did us all the favor of creating a spreadsheet with a confidence calculator, and a lot of other cool stuff as well.

It's better than PT, IMO.

You can find the spreadsheet here.

http://forumserver.twoplustwo.com/showflat.php?Cat=&Number=907554&page=&view=&sb=5&o =&fpart=2&vc=1

The confidence calculator, to my understanding, tells you how likely it is that your actual results are similar to your real - long-term - results. That's based on the number of data points in your sample. The bigger your sample, the more likely it is to be similar to your actual rate of return.

That's not exactly the same as calculating the luck factor, in a given series of games. I'd be very curious to see if there was a method to do this, (and I'm sure there is) but I haven't figured a way to do it.