Why I hate "Look at my 500-SNG penis" posts [Archive]

Irieguy

10-25-2005, 01:38 AM

One of the most common generic types of posts on this forum is the "check-up" post where somebody posts their results (always good) from their last 100-1000 SNGs.

These posts suck.

We all want to get better at SNGs and we all want to know what is obtainable and sustainable in terms of earning potential. These posts don't help in either regard, and can be quite anti-productive in the latter regard.

You don't have to be a statistician to understand how selection bias deems the data contained in these post worthless.

So, the topic got me thinking about a little thought experiment:

Imagine if every registered user of 2+2 played a set of 500 SNGs.

Furthermore, for the sake of demonstration, imagine if every 2+2 poster was actually a LOSING poker player. Assume for a moment that we all had a "true" expected ROI of -9%

I asked our resident Monte Carlo simulator expert to plug-in a -9% ROI for every registered 2+2 user and have them play 500 SNGs.

How many "winners" would there be? 3703
Would somebody have a 20% or better ROI? Yep

So, even if we were all losing poker players playing 500 SNGs a month there could be thousands of posts by "winning" poker players each month talking about their last set of 500 SNGs.

If you want to know if you are any good, ask a long term winner to watch some of your HHs.

If you want to know what's sustainable, ask a long term winner what their ROI/ITM at a certain level has been. (Or search one of the 9384268268726127967234 posts on the topic.)

Irieguy

Seth Money

10-25-2005, 01:41 AM

Amen. My Brother. Amen.

Seth

SammyKid11

10-25-2005, 01:44 AM

YEA!!!!! I think I may have posted one of these stupid posts a few months ago on my first 1000 SnG's. I really only did it because I'd seen a lot of those posts, and it's only been in the past few weeks or so that I've realized how incredibly stupid they are.

I say, if you want to make a results post, it should be minimum 5000 SnG's at a given level. If you're winning after 5000, you're probably a winner.

Even then, though, who really cares about my stats? No one except me and my landlord. It's just self-indulgent. And if you're posting 500 bad SnG's, it's a plea for pity. Yuck. Good post, Irie.

bones

10-25-2005, 01:44 AM

I imagine the typical "My last 500 sngs at $XX buyin" poster expects this sort of validation.

http://hometown.aol.com/pimpen82/images/2girls%20kissing%20me.jpg

splashpot

10-25-2005, 01:45 AM

Dude, that is one low res picture.

SammyKid11

10-25-2005, 01:45 AM

You can bet that if a Penis post actually did RECEIVE this kind of validation, I would be posting my stats often.

SCfuji

10-25-2005, 01:45 AM

you mean my 5000% roi over my last three sitngos isnt because of my 1337 skills?

bones

10-25-2005, 01:47 AM

It's hard to find a quality pic of 2 girls kissing some douchebag for no good reason.

johnnybeef

10-25-2005, 01:47 AM

he introduced me to Vegas strip clubs.....

Irieguy

10-25-2005, 01:47 AM

Bones,

If that picture is of you, you are my type of dude.

Irieguy

bjb23

10-25-2005, 01:49 AM

excellent post. i very much agree.

anyway, you said to ask a "long term" winner about roi and such; im curious as to how you would define "long term" in terms of number of sngs. i have heard some speculative responses to this question but your opinion would be greatly appreciated on the subject (if you have previously stated a number then i obviously missed it).

thanks,
bj

bones

10-25-2005, 01:49 AM

[ QUOTE ]
Bones,

If that picture is of you, you are my type of dude.

Irieguy

[/ QUOTE ]

Unfortunately I don't run that well.

zipppy

10-25-2005, 01:55 AM

sounds like someone's bitter cause their last 500 SNGs weren't good enough to post...

/images/graemlins/grin.gif

Slim Pickens

10-25-2005, 01:55 AM

On the other hand, those 3703 losers with winning records would be anywhere from 51% to 99.9% confident, statistically speaking, they were winning players. Go figure.

I could care less if people want to post stats, but I do find it annoying when it's done under the guise of "How am I doing?" It's a lot like the guy with a 12" dong stepping out of the gym shower and asking everyone if he's shrinking a little because the hot water heater is broken. No one wants to answer, nor should they. It's just a way to trick people into looking at your (e)penis.

Irieguy

10-25-2005, 01:58 AM

[ QUOTE ]
excellent post. i very much agree.

anyway, you said to ask a "long term" winner about roi and such; im curious as to how you would define "long term" in terms of number of sngs. i have heard some speculative responses to this question but your opinion would be greatly appreciated on the subject (if you have previously stated a number then i obviously missed it).

thanks,
bj

[/ QUOTE ]

Well, it's not like there's some magic number where you are suddenly a "long term" winner. As your data set increases in size, your confidence intervals narrow. You can calculate how likely you are to be a winner after any number of SNGs, and you can define your confidence intervals whenever you would like. Aleo's spreadsheet will even do it all for you.

But, to answer your question more specifically: if for some goofy reason I wanted to know what ROI could be maintained at a given level, I would want to speak with someone who had beaten that level for more than 3,000 SNGs.

Irieguy

SlackerMcFly

10-25-2005, 01:59 AM

Wow, if that is a 12oz beer, you and your friends are really, really tiny people!

If you absolutely must carry around a gigantic beverage, at least make it a Heineken.

I mean really! Bud Light? In a can? That large? Yikes.

Here is my promise never to post a 50/100/500/3000 SNG results thread. Also, never to be seen with a Bud Light that big....

Perhaps you have found your new avatar?? /images/graemlins/grin.gif

MGDMcSlack

Irieguy

10-25-2005, 02:00 AM

[ QUOTE ]
No one wants to answer, nor should they. It's just a way to trick people into looking at your (e)penis.

[/ QUOTE ]

Brilliant.

Irieguy

Jman28

10-25-2005, 02:19 AM

[ QUOTE ]

We all want to get better at SNGs and we all want to know what is obtainable and sustainable in terms of earning potential. These posts don't help in either regard, and can be quite anti-productive in the latter regard.

[/ QUOTE ]

This is an excellent point. I don't really mind reading the posts. I usually just skip over them. But they probably hurt our perception of what really is sustainable at given levels, and how most people are really doing. That is a problem.

Also, My ROI at the STEP 5s is 41%. What can I do to improve it? Is that around what I should expect?

chisness

10-25-2005, 02:23 AM

i did one of these when i finished my first 500 (and the results happened to be good). the problem with this is that until you've done a bunch more than 500 you don't realize how dumb it is to make this type of post.

Jman28

10-25-2005, 02:24 AM

[ QUOTE ]
i did one of these when i finished my first 500 (and the results happened to be good). the problem with this is that until you've done a bunch more than 500 you don't realize how dumb it is to make this type of post.

[/ QUOTE ]

Yeah, I just realized that I made at least one post like this early on. My bad Irie.

EDIT: When I did, if I remember, I didn't intend to brag. However, if my results had been terrible, I wouldn't have made the post out of embarassment. That may be what is skewing our data.

bones

10-25-2005, 02:24 AM

[ QUOTE ]
Also, My ROI at the STEP 5s is 41%. What can I do to improve it? Is that around what I should expect?

[/ QUOTE ]

http://www.nightclubchicks.com/kissingagainladies.jpg

Chicks dig Step 5s.

Irieguy

10-25-2005, 02:33 AM

[ QUOTE ]
[ QUOTE ]
Also, My ROI at the STEP 5s is 41%. What can I do to improve it? Is that around what I should expect?

[/ QUOTE ]

http://www.nightclubchicks.com/kissingagainladies.jpg

Chicks dig Step 5s.

[/ QUOTE ]

Bones,

You are really going to enjoy the STTF-HUC II.

Irieguy

freemoney

10-25-2005, 02:37 AM

I don't like going around and telling people that I am freinds with or see on a daily basis really how I am doing in poker. As a 19 year old kid the forum provides an outlet where people can say "oh crap, i was making 9/hr delivering food in the rain, now i make 1124393/hr playing a game I enjoy". That initial realization is pretty incredible and if the forum helps to exercise your e-penis a little bit, I really see nothing wrong with that. Its a negative sum game, to be a significant winner is exciting, especially for young kids who do not possess the ability to make 1/10th that kind of money ever doing anything else. How is that any different than you, Irie, wanting to make social posts about seeing some Prince impersonator or brag that you had girls strip for you? You want this forum to provide certain things for you just like everyone else. If you wanna post about Yugo coming to LV go ahead, but yet you dont understand why a 20 year old kid wants to share that he just made 100000 over the last 20 hours in a game hes been working hard to beat with other people trying to accomplish the same goals?

ilya

10-25-2005, 02:41 AM

bones

10-25-2005, 02:42 AM

[ QUOTE ]
Bones,

You are really going to enjoy the STTF-HUC II.

Irieguy

[/ QUOTE ]

I plan on it. I'm gonna be in vegas in early December for a gin tournament. I hear the Spearmint Rhino is the place to be, so I might have to get some pre-STTF scouting in while I'm there.

Big Limpin'

10-25-2005, 03:16 AM

if someone achieves something they are proud of, and post hopeing for a pat on the back and words of encouragement...i have no problem with it.

Bigwig

10-25-2005, 03:18 AM

[ QUOTE ]
if someone achieves something they are proud of, and post hopeing for a pat on the back and words of encouragement...i have no problem with it.

[/ QUOTE ]
Yeah, I don't either. Some people insist on being forum sheriffs around here. Becoming better at poker is as much psychological as it is technical, and if someone wants reassurance that they're on the right track, I'm happy to give it to them.

microbet

10-25-2005, 03:40 AM

Is having a wife that will make out with strippers sustainable?

10-25-2005, 03:50 AM

No, because eventually she will sober up enough to realize the strippers are hotter than you are.

TheNoodleMan

10-25-2005, 03:52 AM

There comes a point where if you are a winning player, you know it. Once this happens, it really doesn't even matter what your recent results look like. I still import all my hands in to PT, but I have quit looking at the stats. Ignoring my ROI has been one the most +ev developments my game has ever seen, I just can't prove it with numbers. /images/graemlins/tongue.gif

microbet

10-25-2005, 03:55 AM

Since you are replying to my post you must be talking about strippers.

How many strippers did your wife have to make out with before you knew it is sustainable?

curtains

10-25-2005, 04:23 AM

[ QUOTE ]
Since you are replying to my post you must be talking about strippers.

How many strippers did your wife have to make out with before you knew it is sustainable?

[/ QUOTE ]

READ THE [censored] FAQ!!!

mackthefork

10-25-2005, 04:28 AM

[ QUOTE ]
On the other hand, those 3703 losers with winning records would be anywhere from 51% to 99.9% confident, statistically speaking, they were winning players. Go figure.

I could care less if people want to post stats, but I do find it annoying when it's done under the guise of "How am I doing?" It's a lot like the guy with a 12" dong stepping out of the gym shower and asking everyone if he's shrinking a little because the hot water heater is broken. No one wants to answer, nor should they. It's just a way to trick people into looking at your (e)penis.

[/ QUOTE ]

It was meant to be a rhetorical question. Sorry. /images/graemlins/grin.gif

Mack

quinn

10-25-2005, 04:38 AM

[ QUOTE ]

Furthermore, for the sake of demonstration, imagine if every 2+2 poster was actually a LOSING poker player. Assume for a moment that we all had a "true" expected ROI of -9%

I asked our resident Monte Carlo simulator expert to plug-in a -9% ROI for every registered 2+2 user and have them play 500 SNGs.

How many "winners" would there be? 3703
Would somebody have a 20% or better ROI? Yep

[/ QUOTE ]

Well, this is just wrong. The chance that you'd come out with 3703 winners is really small, and it isn't certain that any of them would have a 20% or better ROI. I assume you did one simulation and came up with this, but this conclusion is just horribly presumptuous.

curtains

10-25-2005, 04:43 AM

[ QUOTE ]
[ QUOTE ]

Furthermore, for the sake of demonstration, imagine if every 2+2 poster was actually a LOSING poker player. Assume for a moment that we all had a "true" expected ROI of -9%

I asked our resident Monte Carlo simulator expert to plug-in a -9% ROI for every registered 2+2 user and have them play 500 SNGs.

How many "winners" would there be? 3703
Would somebody have a 20% or better ROI? Yep

[/ QUOTE ]

Well, this is just wrong. The chance that you'd come out with 3703 winners is really small, and it isn't certain that any of them would have a 20% or better ROI. I assume you did one simulation and came up with this, but this conclusion is just horribly presumptuous.

[/ QUOTE ]

Actually I believe Irie's figures are pretty close to correct on this matter.

Arnfinn Madsen

10-25-2005, 04:49 AM

[ QUOTE ]
Also, My ROI at the STEP 5s is 41%.

[/ QUOTE ]

http://www.sneakycracker.net/images/macros/epenis.jpg

raptor517

10-25-2005, 04:56 AM

[ QUOTE ]
[ QUOTE ]
[ QUOTE ]

Furthermore, for the sake of demonstration, imagine if every 2+2 poster was actually a LOSING poker player. Assume for a moment that we all had a "true" expected ROI of -9%

I asked our resident Monte Carlo simulator expert to plug-in a -9% ROI for every registered 2+2 user and have them play 500 SNGs.

How many "winners" would there be? 3703
Would somebody have a 20% or better ROI? Yep

[/ QUOTE ]

Well, this is just wrong. The chance that you'd come out with 3703 winners is really small, and it isn't certain that any of them would have a 20% or better ROI. I assume you did one simulation and came up with this, but this conclusion is just horribly presumptuous.

[/ QUOTE ]

Actually I believe Irie's figures are pretty close to correct on this matter.

[/ QUOTE ]

i second that notion. these numbers are pretty spot on. holla

bennies

10-25-2005, 05:35 AM

[ QUOTE ]
[ QUOTE ]
if someone achieves something they are proud of, and post hopeing for a pat on the back and words of encouragement...i have no problem with it.

[/ QUOTE ]
Yeah, I don't either. Some people insist on being forum sheriffs around here. Becoming better at poker is as much psychological as it is technical, and if someone wants reassurance that they're on the right track, I'm happy to give it to them.

[/ QUOTE ]

I agree. And OP shouldn't worry that reading stat posts are -EV. The readers can figure this out for themselves without a shrink.

bawcerelli

10-25-2005, 06:28 AM

It's not really the size of your ROI, it's how you use it.

PrayingMantis

10-25-2005, 06:59 AM

Irieguy,

I generally agree with the points you make about 500 SNGs "bragging posts", which are very biased and do not give any real picture about anything (although they have some psychological role, both to the posters and readers).

But I think there's some logical loop in the way you present it in your post, which, in a way, is unavoidable, and thus I don't mean to point out a "flaw" in _your_ particular reasoning or something.

The point is, that even if all of 2+2 registered posters are -9% ROI players, there would still be a pretty healthy chance that some of them are in fact "long term" winners too, and very few might even be actually "crushing" the games (that is - there IS the possibility). Naturally confidence intervals will be different, etc, but still, even considering the existence of "long-term winners", you are clearly not able to say with 100% confidence whether this game is +EV at all - in the sense of being beatable for players with a particular set of abilities.

Therefore, there will always be some amount of "belief" involved here, a "belief" in the fairness and beatability of the game which you cannot "prove" or "refute", even if you take your advice only from "long term" winners (that is, unless you have some direct access to the RNG itself and full results of ALL players, and not even then, in some cases). And in that sense, and in a deep way, a 10K games sample is not essentially different than a 500 games sample, and if you take the "biased" element, and put it into posts about "winning over 10K games" (of course, they are very very few in comparison), you get basically the same picture as with the common 500 games posts, i.e, you could write the same post about those people posting those (rare, to say the least) "bragging 10K-games winning post", and all your points would still be valid.

...

Reading the above, I'm not sure I made myself clear, so I'll rephrase it:

X (number of) "500-games bragging posts" aren't much helpful in determining whether SNGs are beatable (and for what ROI, if at all), as much as Y number (Y<X by a big margin) of "10K-games bragging posts" aren't helpful.

BTW, those "10K-games bragging posts" posters don't neccessarily need to actually POST their "bragging posts", since they are simply the most prominent posters, that everybody "knows" they are winning over a long term. But surely if you consider the possibility that this is actually a forum that discusses some -EV game, there's a good chance that there would still be those few long-term winning players, and obvisouly, they'll be the most prominent, and a "proof" that this -EV game is actually beatable, and for what "win-rate".

Double Down

10-25-2005, 07:30 AM

I say we get to the bottom of this and really try to figure out how many games get into the long run. There is a way to figure this out. Let's say for the sake of argument that the average vig for sngs is 10%. So all players on average have a -10% ROI. Now, someone who knows how to calculate standard deviation, figure this out:

After how many sngs would being within let's say -8 to -10% happen 99.999% of the time?

After how many sngs would being within -20 to 0% happen 99.9999% of the time?

Let's say it was 2,000. And after 2,000 sngs you were showing a ROI of 15%. This is probably a good indicator that you are a winning player because statistically, you should overwhelmingly be between -20% and break even, so unless you're an incredible statistical anomaly, you are probably a winning player.

You could calculate the SD a different way. By entering your number of sngs and your ROI, you can calculate within 1, 2, and 3 standard deviations (which I think are 66, 95, and 99.9 percent respectively, yes?) the range in which your ROI probably is.

Would someone who knows more about standard deviation care to delve into this a little deeper?

flyingmoose

10-25-2005, 07:54 AM

[ QUOTE ]
I say we get to the bottom of this and really try to figure out how many games get into the long run. There is a way to figure this out. Let's say for the sake of argument that the average vig for sngs is 10%. So all players on average have a -10% ROI. Now, someone who knows how to calculate standard deviation, figure this out:

After how many sngs would being within let's say -8 to -10% happen 99.999% of the time?

After how many sngs would being within -20 to 0% happen 99.9999% of the time?

Let's say it was 2,000. And after 2,000 sngs you were showing a ROI of 15%. This is probably a good indicator that you are a winning player because statistically, you should overwhelmingly be between -20% and break even, so unless you're an incredible statistical anomaly, you are probably a winning player.

You could calculate the SD a different way. By entering your number of sngs and your ROI, you can calculate within 1, 2, and 3 standard deviations (which I think are 66, 95, and 99.9 percent respectively, yes?) the range in which your ROI probably is.

Would someone who knows more about standard deviation care to delve into this a little deeper?

[/ QUOTE ]

This has been done, and after a certain number of SnGs you can certainly know the odds that you are a winning player. Of course, all math needs assumptions, and to find your win rate (or evidence that you're really a winning player) you would have to assume that the skill level of you and of the average opponent has not changed since you started playing. Considering the skill level of the average opponent changes based on what time of the day you play, this would be a very big assumption.

Basically, you never know your win rate, you have to assume it. Then again, it doesn't really matter what your exact win rate is, unless you feel like strutting it all over the forum in a pair of chapps.

Double Down

10-25-2005, 07:58 AM

Could you explain to me how to figure it out? Like if my ROI is 15% after 2,000 then what are the odds that my true ROI is between 14-16%? What are the odds that I might still be a losing player?

PrayingMantis

10-25-2005, 08:04 AM

[ QUOTE ]
I say we get to the bottom of this and really try to figure out how many games get into the long run. There is a way to figure this out. Let's say for the sake of argument that the average vig for sngs is 10%. So all players on average have a -10% ROI. Now, someone who knows how to calculate standard deviation, figure this out:

After how many sngs would being within let's say -8 to -10% happen 99.999% of the time?

After how many sngs would being within -20 to 0% happen 99.9999% of the time?

Let's say it was 2,000. And after 2,000 sngs you were showing a ROI of 15%. This is probably a good indicator that you are a winning player because statistically, you should overwhelmingly be between -20% and break even, so unless you're an incredible statistical anomaly, you are probably a winning player.

You could calculate the SD a different way. By entering your number of sngs and your ROI, you can calculate within 1, 2, and 3 standard deviations (which I think are 66, 95, and 99.9 percent respectively, yes?) the range in which your ROI probably is.

Would someone who knows more about standard deviation care to delve into this a little deeper?

[/ QUOTE ]

This was discussed and "solved" many times in the past, but it is still not relevant to the points I've made. I'm talking about a completely different kind of problem, which has to do with the information you gain from OTHER players/posters and their results, and how biased it is - i.e, what defines your particular sample.

Obviously, information coming from ONE person and that's it (in your particular case - this person is "you", but it doesn't matter for this discussion) is a very small indication of anything, since it's impossible to determine one's level of "luckiness" (that is - how "normal" are one's results) without a reference or comparison to a sample of any kind. There is no meaning to standard deviation in a vacuum, unless you _assume_ something about what should be normal results (or "believe" in something to begin with), and that's the theoretic problem you have here, or the logical difficulty.

Double Down

10-25-2005, 08:14 AM

I thought that the point of standard deviation WAS to determine your "luckiness" versus your "skills", essentially, based on results, what are the odds that it is a good or bad swing and what are the odds that it is mathematically where it should be.

We can do it for other subjective gambling endeavors such as a card counter in blackjack, why not something with a set buyin such as a SNG? It seems like we should be able to. Educate me on the subject, I'm very interested. /images/graemlins/smile.gif

10-25-2005, 08:34 AM

The look at my dick posts are usually good for a laugh but they are getting old.

AA suited

10-25-2005, 09:49 AM

[ QUOTE ]
If you want to know if you are any good, ask a long term winner to watch some of your HHs.

If you want to know what's sustainable, ask a long term winner what their ROI/ITM at a certain level has been. (Or search one of the 9384268268726127967234 posts on the topic.)

Irieguy

[/ QUOTE ]

thx for this post irie. it made me look at how many sngs i've played.

i never knew the total b4 because databases being full, thus needing to create a new one. and also new accts and user names, but got errors when i tried to alias them to another.

so basically i had a few separate datasets.

wow.. i've played 4000 50+5's. and the roi makes my e-penis go bong! /images/graemlins/smirk.gif /images/graemlins/cool.gif

but you dont want me looking at your HH. here's a sample of my interpretations: Karak567 HH Thread (http://forumserver.twoplustwo.com/showflat.php?Cat=&Number=3761885&page=0&view=colla psed&sb=5&o=14&fpart=1#)

thx again for posting this Irie. now to get a cart to wheel around my e-penis /images/graemlins/grin.gif

kevstreet

10-25-2005, 09:55 AM

[ QUOTE ]
[ QUOTE ]
Since you are replying to my post you must be talking about strippers.

How many strippers did your wife have to make out with before you knew it is sustainable?

[/ QUOTE ]

READ THE [censored] FAQ!!!

[/ QUOTE ]

Very nice Curtains...

zipppy

10-25-2005, 10:11 AM

[ QUOTE ]
if someone achieves something they are proud of, and post hopeing for a pat on the back and words of encouragement...i have no problem with it.

[/ QUOTE ]

pokerlaw

10-25-2005, 10:30 AM

[ QUOTE ]

Imagine if every registered user of 2+2 played a set of 500 SNGs.

[/ QUOTE ]

what was this number (i.e. how many registered users are there on 2+2)?

splashpot

10-25-2005, 10:31 AM

Go to the "main index" and scroll all the way down.

Indiana

10-25-2005, 10:56 AM

Irie,

2 questions. 1: How did you estimate the variability in your simulations? Your answer will depend on the noise.
2: What initial sample size did you start with to come up with the result of 3700? I can easily write an SPLUS script to validate your results.

Indy

splashpot

10-25-2005, 11:03 AM

I've found that you can do similar experiments with rvg72's ROI simulator (http://forumserver.twoplustwo.com/showflat.php?Cat=&Number=3218857&page=&view=&sb=5& o=&vc=1).

PrayingMantis

10-25-2005, 11:09 AM

[ QUOTE ]
I thought that the point of standard deviation WAS to determine your "luckiness" versus your "skills", essentially, based on results, what are the odds that it is a good or bad swing and what are the odds that it is mathematically where it should be.

We can do it for other subjective gambling endeavors such as a card counter in blackjack, why not something with a set buyin such as a SNG? It seems like we should be able to. Educate me on the subject, I'm very interested.

[/ QUOTE ]

You can look at your own post above and see why there's a deeper problem here (It touches on one of the most debateable issues in stats/probabilities at large, btw):

You say:

[ QUOTE ]
There is a way to figure this out. Let's say for the sake of argument that the average vig for sngs is 10%. So all players on average have a -10% ROI. Now, someone who knows how to calculate standard deviation, figure this out:

After how many sngs would being within let's say -8 to -10% happen 99.999% of the time?

After how many sngs would being within -20 to 0% happen 99.9999% of the time?

Let's say it was 2,000. And after 2,000 sngs you were showing a ROI of 15%. This is probably a good indicator that you are a winning player because statistically, you should overwhelmingly be between -20% and break even, so unless you're an incredible statistical anomaly, you are probably a winning player.

[/ QUOTE ]

Suppose you play a game that is in fact equivalent to flipping a non-biased coin, but you don't know that. It costs you $10 to play a game, paying 1:1. You play, say, 1000 games. Now suppose you are a "significant" winning player: you have 550 wins and 450 loses, for a $100 net win, which is basically 1% ROI. Surely according to a normal calculation with regard to how "confident" you can be about being a "winning player", you'll get some result that will be more in the direction of being a winner than a loser - but it won't make any sense, since the game is 0EV by defintion, only you don't know that, and was simply lucky.

More extreme example: suppose you play 1000 of these games, and win 650. This is, by definition, possible. You were very lucky. This has nothing to do with being a "winning player", of course. However, if you use some assumption (=belief) about this game being "beatable" in some way, then you are clearly a "winning player", a "good player", which is absurd.

Obviously, you can run hot enough to have 1000 straight wins. This is extremely improbable, but still, the fact that it's very improbable has nothing to do with what you might call the "actual" result of having 1000 straight wins, which by itself has _the same probability as any other *specific* result with a *specific sequence*_ in this game, and is just as unprobable, or probable. Now if you had such a run, you would believe with an extremely high level of confidence that you are a "winning player" in this game, a real master, but you'll be completely wrong, of course.

In fact, your mistake about your "true" ROI and confidence about being a "winning/losing" player, will grow as you'll be running hotter/colder.

In other words: according to a different (but legitimate) assumption with regard to your results, it is possible to interpret them as being simply nothing more than an indication for how lucky you were running, and nothing else. That is - getting to ANY conclusion about your confidence with regard to any "true ROI" of any kind, is impossible and absurd, according to this.

You'll have the same problem, theoretically, no matter how big is your sample, as long as it isn't "infinite".

Whenever you make any kind of calculation with regard to your confidence of being a winner, you are using some "hidden assumption/s" about the game you're playing. The results of the calculation you make _can not_ prove this assumption, however, many many people seem to be missing this point, by trying to do exactly that: using the results to prove the assumption.

....

This discussion is getting very theoretical, and might seem even absurd to some (although it isn't absurd at all!), but since this whole thread started with a theoretical notion, I think there's place for a somewhat different point of view as well.

pokerlaw

10-25-2005, 11:13 AM

[ QUOTE ]
Go to the "main index" and scroll all the way down.

[/ QUOTE ]

thx

fnord_too

10-25-2005, 11:14 AM

[ QUOTE ]
Irie,

2 questions. 1: How did you estimate the variability in your simulations? Your answer will depend on the noise.
2: What initial sample size did you start with to come up with the result of 3700? I can easily write an SPLUS script to validate your results.

Indy

[/ QUOTE ]

(I ran the simulation for him) I assigned a 10% chance each to coming in first, second, or third and used the random number generator from java.util. For each of 41,000 virtual players, it simulated 500 STT's (deduct buy in from running total, roll the dice, add any money won to running total).

fnord_too

10-25-2005, 11:21 AM

My only comment is that when you do these analyses, you know how likely certain things are, and that those liklihoods are always strictly less than 1. A lot of people consider something like 95% to be 100%, but of course they are wrong (which is I think the essence of what you were trying to say). (Also, a lot of people don't even bother to do the statistics).

Indiana

10-25-2005, 11:30 AM

Regardless of the #s, Irie makes a valid point. Its just story-telling at 500 SNGs. We don't need interim results unless the individual has prospectively defined success and futility stopping boundries for quitting poker.

Indy

PrayingMantis

10-25-2005, 11:31 AM

[ QUOTE ]
My only comment is that when you do these analyses, you know how likely certain things are, and that liklihoos is always strictly less than 1. A lot of people consider thing like 95% to be 100%, but of course they are wrong (which is I think the essence of what you were trying to say). (Also, a lot of people don't even bother to do the statistics).

[/ QUOTE ]

True, and I'd like to add that in some cases and according to some different assumptions, even this "95%" figure is "in fact" lower or much lower, and alternatively, low figures "should be" higher.

I'm using all those "" "" because this is a tricky subject, as any issue that has to do with "confidence" and "levels of belief".

fnord_too

10-25-2005, 11:31 AM

One more reply to this:

This gets at the heart of another big misconception in the world: that science proves things. A lot of people are ignorant that science consists of theories which are supported (or not) by experiments. There is no "LAW of Gravity," for instance, it is the "Theory of Gravity," which can (and has been to some extent) invalidated. (Clarification: Newtonian Gravity does not explain some observations. This resulted in the relativity theories. Newtonian gravity is a very good approximation in most of the cases humans have to deal with, but not entirely accurate. Relavativity could likewise be wrong in ways we have yet to discover.)

fnord_too

10-25-2005, 11:34 AM

[ QUOTE ]

I'm using all those "" "" because this is a tricky subject, as any issue that has to do with "confidence" and "levels of belief".

[/ QUOTE ]

Yeah, statistics can get really hairy and assumptions have huge impacts. Also, when hypothesis testing one can get different results depending on which null hypothesis is chosen. Lies, damn lies, and statistics indeed!

PrayingMantis

10-25-2005, 11:42 AM

[ QUOTE ]
Regardless of the #s, Irie makes a valid point. Its just story-telling at 500 SNGs. We don't need interim results unless the individual has prospectively defined success and futility stopping boundries for quitting poker.

[/ QUOTE ]

When you say this, you just need to realize that ANY specific number of results (i.e, sample size), particularly one that is achievable by a human-being, is no more than "interim results" in the pure sense. Therefore, according to your own logic, you shouldn't be interested in any kind of results posted, no matter how relatively big is the sample size, and who is the poster.

Indiana

10-25-2005, 11:44 AM

When I say interim I mean part of the way to the necessary sample size for showing such a result with reasonable power. True, all data is "interim" data if you consider all data that could be collected but this isn't what I meant.

Indy

PrayingMantis

10-25-2005, 11:56 AM

[ QUOTE ]
This gets at the heart of another big misconception in the world: that science proves things. A lot of people are ignorant that science consists of theories which are supported (or not) by experiments. There is no "LAW of Gravity," for instance, it is the "Theory of Gravity," which can (and has been to some extent) invalidated. (Clarification: Newtonian Gravity does not explain some observations. This resulted in the relativity theories. Newtonian gravity is a very good approximation in most of the cases humans have to deal with, but not entirely accurate. Relavativity could likewise be wrong in ways we have yet to discover.)

[/ QUOTE ]

Good points and analogy.

[ QUOTE ]
Relavativity could likewise be wrong in ways we have yet to discover

[/ QUOTE ]

You probably know that, but there are aspects of quantum mechanics that completely contradict basic assumptions of "relativity", and that's why in some senses relativity was already "discovered" to be wrong, or at least not consistent with a different and very valid and successful theory. Einstein was very unhappy about it too, although he had a big part in the emergence of the qunatum theory as well.

Yes, I think it's about time we discuss these things here!

/images/graemlins/grin.gif

stupidsucker

10-25-2005, 12:19 PM

[ QUOTE ]
You probably know that, but there are aspects of quantum mechanics that completely contradict basic assumptions of "relativity", and that's why in some senses relativity was already "discovered" to be wrong, or at least not consistent with a different and very valid and successful theory. Einstein was very unhappy about it too, although he had a big part in the emergence of the qunatum theory as well.

Yes, I think it's about time we discuss these things here!

[/ QUOTE ]

I find these things extremely interesting. If only I were smart /images/graemlins/frown.gif

microbet

10-25-2005, 12:23 PM

We know God doesn't play dice, but does he play poker?

Oh yeah, Stanzee!

PrayingMantis

10-25-2005, 12:24 PM

Hey stupidsucker, if you really find these things interesting you can find quite a lot of popular good science/physics books directed at people who don't have any advanced educations at these matters. It is indeed interesting and you don't need a lot of math, or to be exceptionally smart, IMO, in order to understand at least some of the ideas/theories.

PrayingMantis

10-25-2005, 12:28 PM

As someone once wrote somewhere:

God DOES play dice, but they're loaded, so it's OK.

In other words:

God does play poker, but only on-line which is obviously completely rigged so Einstein shouldn't be bothered. /images/graemlins/grin.gif

cleinen

10-25-2005, 12:28 PM

I think this is all over my head but i will give some thoughts. I don't believe you can ever have a definitive sample size to know your true ROI. It will get closer and closer to a true number the more you play but that number will have a margin of error. So to be a definitive sample size wouldn't you have to encountered every situation possible many times to get a true read of your play. I think this is impossible due to the fact of the variable the villian brings into the situation. I believe this creates so many variables and other factors that the it would be impossible to ever know your "true ROI".

Corey

Slim Pickens

10-25-2005, 12:32 PM

[ QUOTE ]
[ QUOTE ]
You probably know that, but there are aspects of quantum mechanics that completely contradict basic assumptions of "relativity", and that's why in some senses relativity was already "discovered" to be wrong, or at least not consistent with a different and very valid and successful theory. Einstein was very unhappy about it too, although he had a big part in the emergence of the qunatum theory as well.

Yes, I think it's about time we discuss these things here!

[/ QUOTE ]

I find these things extremely interesting. If only I were smart /images/graemlins/frown.gif

[/ QUOTE ]

It all starts when a nulecule comes out of its shell.

Double Down

10-25-2005, 12:37 PM

Yes, it's impossible to know your true ROI, but as you said, you will tend to drift towards a number the more SNGs you play, and by using standard deviation, you can find out the chances of your actual ROI being within certain ranges.

For example, after 10,000 SNGs, if your ROI is 15%, there is probably a 99% chance that your ROI is between 13% and 17%. And after 20,000, there's probably a 99% chance that it's truly between 14.5% and 15.5%. I'm pulling these figures out of my ass, but someone who knows how to utilize standard deviation could figure out the actual numbers.

microbet

10-25-2005, 12:37 PM

Does Forumbot count as a registered user in this simulation?

cleinen

10-25-2005, 12:41 PM

[ QUOTE ]
For example, after 10,000 SNGs, if your ROI is 15%, there is probably a 99% chance that your ROI is between 13% and 17%. And after 20,000, there's probably a 99% chance that it's truly between 14.5% and 15.5%. I'm pulling these figures out of my ass, but someone who knows how to utilize standard deviation could figure out the actual numbers.

[/ QUOTE ]

I guess a point/question is...can you even ever be sure of what the standard deviation is?

Slim Pickens

10-25-2005, 12:50 PM

Basically, you've got it. This often gets ignored in the statistical analysis, but one's "true" ROI is a dynamic quantity. Changing game conditions, mostly your skill relative to the opponent pool, prevent your ROI from converging. In that way, the 500-SNG posts are as valuable as that sort of information gets. "My ROI is 14.0% over 500 and that gives me a statistical winning confidence of 97.04%" is as good as you'll ever be able to do. Skill is much more likely to be a constant over 500 than over 5000.

The checkup posts can be valuable when, rather than "look at my epenis," they're a roadmap to beginning players. How did someone start with $250 and worked it up to 4-tabling the 55's? What happened along the way and what should I expect? The easiest way to show someone why 100 is an inadequete sample size is to post a graph of 500. Ditto for 500 and 3000, and 100 and 3000. It's like explaining to a middle-school boy why he's always pitching random tents (sample size 100 SNG's/a few games of Spin the Bottle) but can't get the girls to notice him. His high-school senior older brother (sample size 500 SNG's/lots of dry-humping and the one drunk prom date) has one opinion that's certainly more knowledgeable, but still distorted, and he really wants to share it because then his younger brother thinks he's cool. Dad (sample size 5000 SNG's and ...) knows the whole deal, but won't tell his sons because they have to figure it out for themselves.

SuitedSixes

10-25-2005, 12:50 PM

[ QUOTE ]
[ QUOTE ]
For example, after 10,000 SNGs, if your ROI is 15%, there is probably a 99% chance that your ROI is between 13% and 17%. And after 20,000, there's probably a 99% chance that it's truly between 14.5% and 15.5%. I'm pulling these figures out of my ass, but someone who knows how to utilize standard deviation could figure out the actual numbers.

[/ QUOTE ]

I guess a point/question is...can you even ever be sure of what the standard deviation is?

[/ QUOTE ]

This is an age-old fight here, but yes. The standard deviation is going to fall within a pretty tight tange for everyone. Irie made a post about 9 months ago about this.

Think of ROI as batting average. Players will have greater swings in their BA in April and May than they will in September, just because the effect of one at bat will have less impact the more at bats they have.

As an aside, in the ring game world, 10k hands is considered to be somewhat statistically signifcant. I now have over 500k hands in my PT database (50 10k samples) and I have had 5 10K losing periods, yet according to the "winning confidence" formula I am at 99.83%.

I also hope everyone appreciates the irony that (the one who I consider to be) The King of the E-Penis posts made an appearance in this thread.

cleinen

10-25-2005, 01:22 PM

[ QUOTE ]
I also hope everyone appreciates the irony that (the one who I consider to be) The King of the E-Penis posts made an appearance in this thread.

[/ QUOTE ]

Slacker? /images/graemlins/grin.gif

pergesu

10-25-2005, 01:37 PM

[ QUOTE ]
The checkup posts can be valuable when, rather than "look at my epenis," they're a roadmap to beginning players. How did someone start with $250 and worked it up to 4-tabling the 55's? What happened along the way and what should I expect? The easiest way to show someone why 100 is an inadequete sample size is to post a graph of 500. Ditto for 500 and 3000, and 100 and 3000. It's like explaining to a middle-school boy why he's always pitching random tents (sample size 100 SNG's/a few games of Spin the Bottle) but can't get the girls to notice him. His high-school senior older brother (sample size 500 SNG's/lots of dry-humping and the one drunk prom date) has one opinion that's certainly more knowledgeable, but still distorted, and he really wants to share it because then his younger brother thinks he's cool. Dad (sample size 5000 SNG's and ...) knows the whole deal, but won't tell his sons because they have to figure it out for themselves.

[/ QUOTE ]
Nobody may ever question your genius.