A Statician's take on RUNNING BAD! - Page 2

I think many people here underestimate how little of a sample we actually generate playing online and how big of an impact a run-bad/good at a certain period of time could have. A run bad during a move-up as opposed to a run-good could mean a difference of several months. To have a truly good sized sample we'd need several years of play.

How big is our poker time-line? Many of us are in it for a few years then out. In that case variance is highly important. A run-good to shoot us up into the nosebleeds as opposed to a run-bad that keeps us at 100nl could mean the difference between going full time pro or working for Acme Corp.

On August 23 2009 09:32 SpeedyJack wrote: I don't think you're giving the people on this forum enough credit, I'm pretty sure most of the players and almost all of the winning ones know your points and agree. However, if I played 20K hands this weekend and got KKvAA 50 times, I'm pretty sure we can say I was running bad. We also acknowledge running good, and if you view each players' hands as one piece of a really long "string" then it's pretty easy to see how one person can run poorly over a decent sized sample. We get it man, do you actually think anyone on LP thinks that the odds of winning a flip get better just because they lost one earlier?

I totally understand where you are coming from. The point of the article is to explain that you will run good and run bad with 100% mathematical certainty. You should expect with 100% certainty to have sessions where you lose every time you got it in as a favourite. Getting emotional is an irrational behaviour to a very common occurrence.

A quick retort to the (50 KK vs AA argument). Sure you can use that one statistic to bemoan the poker gods, but during that 20k hand, how often will you have QQ vs JJ or KK vs TT, etc. If you add up your equity in every single hand preflop you will undoubtably find that you ran normal.

On August 23 2009 04:15 PokerDoc88 wrote: I think you'll find if you ran multiple string simulations of 100 flips, you would not commonly see 2 strings of 7 2's. You don't see a single 6 or 5 string in that sequence. In a 100 flip sequence, you'd have 93*0.007 (I think) chance of seeing a sequence of exactly 7 flips in a row. It is still quite improbable.

actually if you look at player 1 there are a couple of string of 5 or 6.

I agree with most points you made in this thread, but "running bad" still exists.

You would not argue that running "good" does not exist, do you? Jamie Gold surely ran good for example, and running bad is basically the same thing as running good, it just means that you are either way above or way below your "expectation" to win.

Now, determining your true expectation to win is almost impossible due to the highly random structure of this game and due to so many factors that are impossible to measure.

What I liked the most was the part about that the human brain's perception is heavily skewed and that many people think they are running bad when in reality they are not, it's just that the bad things are memorized stronger and that when they win a pot with 80% equity most people are really thinking that they "deserved" to win 100% of the pot when in reality they "deserved" only 80% of it. That is definately a good valid point, and I couldn't agree more.

But the main thing I am getting at is that I still think that its possible to run very bad, even more than most people would think, it is just that is is VERY unlikely to happen, but it still does. Some people will definately run worse than others during their lifetime career. I would assume that the distribution resembles a bell curve here with extremely bad running players at one side(very few of them) and extremely good running players at the other side(also very few), but the main population just running about average.

The more hands you play, the more the LIKELINESS of your win rate to come closer to your expected winnings, but that doesnt mean that things will "even out".

For example, you can have expected winnings of 100$ over 100.000 hands. Now you only won 98$ over these hands. That means, you ran 2% below your expectation, or to say, you ran 2$ below your expectation.

Now, you play 100.000.000 hands, your expected winnings should be 100.000$, but you win 98.500$ so basically you came a lot closer to your expectation ( 1,5%) but in total terms of money your gap got bigger.(-1.500$ below EV).

Ah well pretty unstructured but yeah.

On August 23 2009 09:57 Maynard! wrote: I think many people here underestimate how little of a sample we actually generate playing online and how big of an impact a run-bad/good at a certain period of time could have. A run bad during a move-up as opposed to a run-good could mean a difference of several months. To have a truly good sized sample we'd need several years of play. How big is our poker time-line? Many of us are in it for a few years then out. In that case variance is highly important. A run-good to shoot us up into the nosebleeds as opposed to a run-bad that keeps us at 100nl could mean the difference between going full time pro or working for Acme Corp.

I also agree with that, very good point.

thanks to this thread, i can now lol at hands like this to people who run like jesus against me everyday hahaha

Submitted by : whamm!

PokerStars Game #31957217008: Holdem No Limit ($0.50/$1.00 USD) - 2009/08/23 8:56:32 ET
Table Kitalpha II 9-max Seat #6 is the button
Seat 1: rocco13579 ($40.35 in chips)
Seat 2: dwq2300 ($16.30 in chips)
Seat 3: orena22 ($26.20 in chips)
Seat 4: DaveyDoWin ($94.60 in chips)
Seat 5: JP917 ($157.90 in chips)
Seat 6: Hero ($100 in chips)
Seat 7: Klaus20 ($100 in chips)
Seat 9: Large1337 ($20 in chips)
Klaus20: posts small blind $0.50
Large1337: posts big blind $1

Holecards(Odds)
Dealt to Hero
rocco13579: raises $2 to $3
dwq2300: folds
orena22: calls $3
DaveyDoWin: folds
JP917: calls $3
Hero: calls $3
Klaus20: folds
Large1337: folds

Flop(Odds) (Pot : $13.50)

   
rocco13579: checks
orena22: bets $23.20 and is all-in
JP917: raises $131.70 to $154.90 and is all-in
Hero: calls $97 and is all-in
rocco13579: folds
Uncalled bet ($57.90) returned to JP917

Turn(Odds) (Pot : $230.70)

   

River (Pot : $230.70)

   

Showdown
JP917: shows (four of a kind, Sixes)
Hero: shows (a full house, Nines full of Sixes)
JP917 collected $147.60 from side pot
orena22: shows (two pair, Eights and Sixes)
JP917 collected $80.10 from main pot

Summary
Total pot $230.70 Main pot $80.10. Side pot $147.60. | Rake $3
Board  
Seat 1: rocco13579 folded on the Flop
Seat 2: dwq2300 folded before Flop (didnt bet)
Seat 3: orena22 showed and lost with two pair, Eights and Sixes
Seat 4: DaveyDoWin folded before Flop (didnt bet)
Seat 5: JP917 showed and won ($227.70) with four of a kind, Sixes
Seat 6: Hero (button) showed and lost with a full house, Nines full of Sixes
Seat 7: Klaus20 (small blind) folded before Flop
Seat 9: Large1337 (big blind) folded before Flop

On August 23 2009 04:15 PokerDoc88 wrote: I think you'll find if you ran multiple string simulations of 100 flips, you would not commonly see 2 strings of 7 2's. You don't see a single 6 or 5 string in that sequence. In a 100 flip sequence, you'd have 93*0.007 (I think) chance of seeing a sequence of exactly 7 flips in a row. It is still quite improbable.

The math is non-trivial and I actually had to research it but you can find it here:

http://mathdl.maa.org/images/upload_l.../Polya/07468342.di020742.02p0021g.pdf
http://faculty.pittstate.edu/~ananda/STATMETHODI/Theory-of-runs.pdf

To summarize the first paper, the expected value of the longest run is about 5.6 for n = 100, and 6.6 for n= 200 and follows a logorithmic rate.
He also did an experiment where a student makes up what he thinks is a random sequence and compared it to a true random sequence. He then ran it through a computer program to see which sequence was more likely to be random. The major downfall for students is that their longest run of consecutive heads or tails was too short.

cliffnotes plz

On August 23 2009 10:56 royalsu wrote: Show nested quote + On August 23 2009 04:15 PokerDoc88 wrote: I think you'll find if you ran multiple string simulations of 100 flips, you would not commonly see 2 strings of 7 2's. You don't see a single 6 or 5 string in that sequence. In a 100 flip sequence, you'd have 93*0.007 (I think) chance of seeing a sequence of exactly 7 flips in a row. It is still quite improbable. The math is non-trivial and I actually had to research it but you can find it here: http://mathdl.maa.org/images/upload_l.../Polya/07468342.di020742.02p0021g.pdf http://faculty.pittstate.edu/~ananda/STATMETHODI/Theory-of-runs.pdf To summarize the first paper, the expected value of the longest run is about 5.6 for n = 100, and 6.6 for n= 200 and follows a logorithmic rate. He also did an experiment where a student makes up what he thinks is a random sequence and compared it to a true random sequence. He then ran it through a computer program to see which sequence was more likely to be random. The major downfall for students is that their longest run of consecutive heads or tails was too short.

that's actually cool

apparently nobody got my post

its spelled "statistician" not "statician"

also, how familiar with statistics is the op? (ps, don't say "very." i'm actually quite curious as I was a math major)

sorry day, I will fix the title
i basically have a degree in pure and applied math at waterloo but I'm not a statistician. So I've taken the standard university probability/statistic courses. I was reading john allen paulos' book "INNUMERACY: mathematical illiteracy and its consequences" and he mentions how people are so amazed at hitting streaks in baseball. Or the "hot hand" in basketball. The basketball example is especially fitting, since kobe bryant shoots at about 50% and yet he can have many streaks of 10-15 consecutive made shots. It's been shown that mathematically this is expected.

Kobe bryant has taken 18162 shots in his career. If you take his sequence of made vs missed shots it should look like a random sequence of {0,1}.

i read about the first half of what you wrote and saw some hypocrisies.

for example, saying you "missed a $100 value bet on the river means you just lost $100"

by saying this you are basically saying you dont understand hand ranges, expectation and variance. suppose on the river you are ahead of 90% of your opponents range. your opponent never folds a worse hand, so he calls with better 10%, folds 70% and calls with worse 20%. this means that the EV of your bet is $10, not $100.

i mean everything you are saying is just based on coinflip math. are you serious? do you really think running bad is a myth?

running bad is not bullshit lol. just look at winrate/standard deviations. a common winrate is 2ptbb and a common standard deviation for that winrate is 50ptbb. whats that mean? that losing at 100ptbb for 100 hands is only 2 standard deviations outside of the norm. which off the top of my head is about 3% of the time? so every 3000 or so hands you should have at least 1 100 hand stretch where you lose 2+ buyins.



heres a program where you can input ptbb and standard deviation and print out samples of 100k hands. after refreshing just once i received this sample:




as you can see, this sample had a person up 20 buyins after 20k hands and he finished roughly even. that means he had an 80k hand stretch where he lost 20 buyins. he is a winning player and this simulation does not take into account tilt, missed bets, anything, just math and what a real statistician would look at, not some random sequence of coin flips rofl.

it would NOT take long to press f9 and find a large losing sample over this many hands.

for the program go here:
http://www.anskypoker.com/2009/06/100k-hand-variance-simulator/ just click on variance simulator and open it in excel, play with the ptbb/standard deviation and press f9 to get new simulations

On August 23 2009 10:56 royalsu wrote: Show nested quote + On August 23 2009 04:15 PokerDoc88 wrote: I think you'll find if you ran multiple string simulations of 100 flips, you would not commonly see 2 strings of 7 2's. You don't see a single 6 or 5 string in that sequence. In a 100 flip sequence, you'd have 93*0.007 (I think) chance of seeing a sequence of exactly 7 flips in a row. It is still quite improbable. The math is non-trivial and I actually had to research it but you can find it here: http://mathdl.maa.org/images/upload_l.../Polya/07468342.di020742.02p0021g.pdf http://faculty.pittstate.edu/~ananda/STATMETHODI/Theory-of-runs.pdf To summarize the first paper, the expected value of the longest run is about 5.6 for n = 100, and 6.6 for n= 200 and follows a logorithmic rate. He also did an experiment where a student makes up what he thinks is a random sequence and compared it to a true random sequence. He then ran it through a computer program to see which sequence was more likely to be random. The major downfall for students is that their longest run of consecutive heads or tails was too short.

really great discussion guys.

I rembember this study - where he asked students to make their own random sequences.
They had dramatically wayy too short consecutive runs, was quite surprising actually.

hit f9 4 more times and got this one. 50 buyin downswing over 70k hands. as a 2ptbb winner over 70k hands you should expect to win 28 buyins. that means this player at one point was about 80 buyins under EV after 75k ish hands.

if a player started out a poker career with a stretch like this then there is no way he will ever become the player who would reach his full potential. he would be convinced poker is not beatable and would never learn properly because he ran so completely horrible from the outset that his emotional state about poker and his confidence in the game would be crushed forever. this is another form of running bad and being unlucky that people tend to overlook.

royalsu, instead of thinking of running good/bad as a sequence of 1's and 2's, you can view a session as a random walk on a number line (stepping left when you lose and right when you win).

I like this interpretation considerably more than runs, as i think the math involved with random walks is more interesting. For instance, given any point on a number line, there is a 100% probability you'll arrive at that point using a random walk. cool stuffs

Yeah day statistics is pretty counter-intuitive sometimes. I'm going to read up on that when school starts.

Daut I'm not saying that bad runs don't exist. In fact I'm showing precisely the opposite, that bad runs are very common. You can easily go broke in poker from a bad run.

What I'm trying to point out in this article is how people think about bad runs and expected value. For example, if you're playing a session and you're up or down 5 buyins really quickly, this shouldn't be surprising and there should be no emotional feelings to a totally normal and common occurrence.

You bring up good points about that 100$ missed value bet and a lot of people never even think about it from the point of view you just described.

I really dont understand the point of this post. Bad runs are a part of poker, thus they dont exist. Que?

From talking to peeps on msn I've realized that my article is a bit misleading.

Clarification: my point is that bad runs are way more common than you would expect. The viewpoint I want to change is all those grinders who complain about ONE session being so unfair, so unlucky. Then blaming that one session for ruining their profits for the month. Those unlucky sessions are inevitable.