wanna know if your good?

[stupidsucker]

Forget ITM
Forget ROI

While reading people pick apart Irie's latest post I read something very important.

[ QUOTE ]
At the poker tables, however, bad decisions are often rewarded (caught my 2-outer on the river) and good decisions are punished (the other guy caught his 2-outer on the river). At worst this trains our brains to play less than ideally, and at best it just confuses us. -Misfire(paraphrasing Slansky?)

[/ QUOTE ]

Poker Tracker, Spreadsheets.. whatever else we use to track our data leaves out what I feel is the most important factor in SnGs.

Are we making the right decision? Yes Yes this has been said before, maybe even beaten into the ground.... BUT

How many of you know what your roi is? ITM? yeah, you do.

But how many of you know how often you were dominated because of a bad call? Or know exactly what your EV compared to AV is? This is the true test of running hot or running cold.

Does anyone track their results this way? I personaly want to start. When I look at my figures at the end of the week/month/day whatever I dont want to care what my roi is. All I care about is if I am making the right choice.

Its fine to be results oriented. Just have to look at the right results. roi isnt the result, neither is itm. Sngs comes down to 1 thing. Do I have positive $EV when the chips go in?(not to be confused with chip EV)

Spreadsheet makers.... is this possible to do? This is the data that is important to look at in the long run the short run.

[Jibbs]

I've been wondering about this lately too. ITM and ROI tell you how your doing but don't rate how you got there. You could just be on a nice fat hot streak. I have wrote some software to parse hand histories last month and was thinking of writing something that would give me a little better information on how I am actually playing. One thing I wanted to see was what how often I was getting into all-in situations with the best (or worst) hand and how those situations were turning out.

[Slim Pickens]

I was doing stuff like this, especially EV vs. AV, for a set of 180 $11 SnG's. It took me (guessing) 40 man-hours to generate what I thought were useful results, and I only got up to the bubble. PM me if you want to take a look at my spreadsheets. It very well could be automated by someone with VB skills. If someone did the work to generate such a program, it would rightly cost you thousands of dollars to purchase.

Slim


Analysis of Streaks III (http://forumserver.twoplustwo.com/showthreaded.php?Cat=&Number=1907441&page=&view=&s b=5&o=&vc=1)

[The Student]

[ QUOTE ]

Does anyone track their results this way? I personaly want to start. When I look at my figures at the end of the week/month/day whatever I dont want to care what my roi is. All I care about is if I am making the right choice.



Spreadsheet makers.... is this possible to do? This is the data that is important to look at in the long run the short run.

[/ QUOTE ]

Amen. this is a great idea, one that is so good, that i'm amazed that i haven't thought of/read/heard of it before. after the endless threads of people threatening to give up their spreadsheets, Yugo's prompting that we should forget about the #s and just play, and even Gigabet's philosophizing about what the difference is between winning and losing, it really is all about making the right decisions. to really improve our game, this is ultimately all we care about - are we making the right decisions?

unfortunately, i could barely program in logo on my old apple 2c, but some ideas about what this program should look for:
1. definitely needs something like Eastbay's SNG analyzer to help with bubble play and push/fold decisions.
2. needs to be able to determine whether or not you're making the right decisions in regards to your draws and pot odds (more of an issue in the early/middle levels)
3. other ideas?

can anyone help out with this project?
ts-

[jcm4ccc]

I disagree completely. The point of SnGs is to win money. Here's an example of the error in this way of thinking.

Let's say it's the first level of an SnG. I'm dealt AQo UTG. I go all-in. I am called by someone with A8o. I lose. According to your proposed calculations, I'm just a victim of bad luck, because I was called by a worse hand and I lost.

Or let's say there are 4 players left, blinds 100/200, I'm in the SB with K7o. I have 1000 chips, everybody else has 3000 chips. Two folds, I push, I'm called by AKs, I still win the hand. According to your calculations, I'm lucky rather than good.

[gumpzilla]

[ QUOTE ]
But how many of you know how often you were dominated because of a bad call? Or know exactly what your EV compared to AV is? This is the true test of running hot or running cold.

[/ QUOTE ]

Let's say your opponent has pushed their last ten hands. You conclude from this that they are likely pushing any two, and you look at 99 in your hand. You gleefully put your money in. Your opponent flips over AA. Did you make a bad call? It seems to me that your brand of analysis would say yes, but it's equally obvious that you didn't make a bad call here. (Yes, we can construct situations where it would be a bad call, but in general it's going to be sweet.) The results you're looking at are no different from the standard results-oriented stuff, as far as I can tell.

This would be a good way to look at things, except it's very hard to concretely prove that various calls are good or bad, since a lot of what makes them good or bad is what kind of range you put your opponent on. There's not a good way to show that your range is too loose; you can find out if your range was missing relevant hands, but that's about the only way you can find a mistake. And as far as determining whether calls are good or bad based on the range you've put them on . . . well, that's what all the math we talk about here is for.

[Apathy]

[ QUOTE ]
Sngs comes down to 1 thing. Do I have positive $EV when the chips go in?(not to be confused with chip EV)


[/ QUOTE ]

This is a vast over simplification. There are plenty of times where you should pass on +$EV situation and there are also plenty of times you should take a -$EV gamble.

Also I'm not sure if I understand entirely what it is you want but if you are wondering how well you do in all in situations compared to what the EV is, this is also an over simplification of running well. If you are always dealy KK when someone else has AA for example you are running bad but this wouldnt be quantifiable by the sheet you want.

[gumpzilla]

[ QUOTE ]

This is a vast over simplification. There are plenty of times where you should pass on +$EV situation and there are also plenty of times you should take a -$EV gamble.

[/ QUOTE ]

Assuming adequate bankroll, I disagree with this.

A couple of things I think you might mean:

1) Are you sure you don't mean chip EV, not $ EV?

2) Are you talking about making long term image plays against players you see often that are +$EV in the long term but -$EV right now? I'd argue these are still + $EV.

EDIT: Thinking about 30 seconds more, you probably are making the better spots argument. But in order to get your true $EV, these kinds of things would be taken into account; that is, the EV of passing on a current play will increase if it's likely that there are better plays later, and if this increase is sufficient then it becomes the best play, $EV wise. You're still trying to maximize that $EV.

[Slim Pickens]

[ QUOTE ]
Let's say it's the first level of an SnG. I'm dealt AQo UTG. I go all-in. <font color="red">Regardless of the outcome of this particular situation, that was a stupid play because on average it loses money given how you fare against the range of hands that typically call you, how often someone calls, and how much you stand to win when no one calls.</font>

[/ QUOTE ]

[ QUOTE ]
Or let's say there are 4 players left, blinds 100/200, I'm in the SB with K7o. I have 1000 chips, everybody else has 3000 chips. Two folds, I push. <font color="red">Regardless of the outcome of this particular situation, that was a good play because on average it wins money given how you fare against the range of hands that typically call you, how often someone calls, and how much you stand to win when no one calls.</font>

[/ QUOTE ]

I think you're making results-oriented statements.

EDIT: ...but I think we just disagree on the OP's proposed calculations. When I did mine, I also looked at random hand EV vs. my actual EV, as in was it "BB gets AA week" or something similar.

[gumpzilla]

Of course he is, because he's pointing out the silliness of the results-oriented program that the original poster seems to be advocating.

[The Student]

[ QUOTE ]
I disagree completely. The point of SnGs is to win money. Here's an example of the error in this way of thinking.

Let's say it's the first level of an SnG. I'm dealt AQo UTG. I go all-in. I am called by someone with A8o. I lose. According to your proposed calculations, I'm just a victim of bad luck, because I was called by a worse hand and I lost.

Or let's say there are 4 players left, blinds 100/200, I'm in the SB with K7o. I have 1000 chips, everybody else has 3000 chips. Two folds, I push, I'm called by AKs, I still win the hand. According to your calculations, I'm lucky rather than good.

[/ QUOTE ]

I won't speak for the OP, but personally, these aren't the first things I'm going to look for to come out of this program. What I'd first analyze is a variation of your second example. Let's say I'm the SB with 1000 chips and K/Js, but let's say there's a push in front of me from the button with 3000 chips and I call. If his VPIP is 60%, is it correct for me to call based upon his probable range of hands? How does this change when his VPIP is 15%? This is what I want the program to tell me.

But even if we use your second example, it's still possible that you made the right decision based upon how tight the BB is playing. Just because you were up against a better hand and got lucky doesn't mean that you made an incorrect decision for that particular situation. the ideal program should be able to determine whether or not a push is a $EV positive move here.

ts-

[gumpzilla]

You appear to be asking for a poker playing AI, nothing less.

When you have questions about hands like these, there are two basic kinds of errors that you can make:

1) Putting your opponent on a faulty range;

2) Not calculating or miscalculating the appropriate thing to do given this range.

Programs exist (eastbay's SNG analyzer is the prime example, but PokerStove and dethgrind's ICM calculator should work just as well if you don't mind doing a little bit of work) to help with the second one, but I don't think you're going to write a computer program to help you with the first one very easily.

[Phil Van Sexton]

I wasn't going to respond because I wasn't sure if this was serious. Apparently other people think it is, so I guess I should respond.

If a computer program could actually tell you if you were playing well, then that program would have to know how to play well. If it did, someone would write a bot and make millions rather that generating analysis of your play.

I'm all for doing analysis of your play. It sounds like Slim did something to see how often he was getting called on steal. It doesn't tell him if he's playing well, but the data might help him to identify leaks. At least I hope so given the effort he put into it.

As stated already, any eastbay analysis would require hand ranges and is completely impractical.

On the other hand, ExpectedWin vs ActualWin reports are possible, but these are just pointless "Luck Reports".

If you knew that you won 47% of allins last week and you should have won 51%, what can you do with this information? Nothing.

[The Student]

[ QUOTE ]
You appear to be asking for a poker playing AI, nothing less.

but I don't think you're going to write a computer program to help you with the first one very easily.

[/ QUOTE ]

I know that I can't write this program at all. Yeah, I guess I am looking for some kind of AI program - but it doesn't hurt to dream, right?

ts-

[Slim Pickens]

[ QUOTE ]
I'm all for doing analysis of your play. It sounds like Slim did something to see how often he was getting called on steal. It doesn't tell him if he's playing well, but the data might help him to identify leaks. At least I hope so given the effort he put into it.

[/ QUOTE ]
I started the analysis thinking it would be about steals, but for all the effort I put into it, I found the only modification to my game was to back off just a little bit on the steals. Oh and I gave up sleeping last year so the time's no big deal.



[ QUOTE ]
On the other hand, ExpectedWin vs ActualWin reports are possible, but these are just pointless "Luck Reports".

If you knew that you won 47% of allins last week and you should have won 51%, what can you do with this information? Nothing.

[/ QUOTE ]
It might make me feel better about it. For long-time players, it won't matter, but sometimes a n00b's got to demonstrate why 90 tournaments is a totally insignificant sample size and exactly how a run of bad luck on individual hands translates into his bankroll.

Slim

[hummusx]

I'm currently working towards a few related tools. I hadn't thought too much about a tool that would summarize this information, to give you some sort of 'luck quotient' or anything like that. What I am planning to do is to build a tool that will scan tournament histories based on some criteria and allow me to quickly browse through 'interesting' hands. It should be easy enough to define some criteria to do dirty estimates of push EV and call EV based on a default hand range, and then scan a few hundred tournaments and show only the hands that might be worth looking into.

Also, to answer Phil, the one thing you might say would be worthwhile about knowing that you were 47% coin flips when you should have been 51% is to compare that to other figures. If you're running 65% ROI and your coin flips are 95% but should be 51%, you can be pretty sure it's not that you are a poker god.

[hedgeyerbets]

If you play enough hands your ROI will eventually come to reflect exactly what you seem to desire it to: how well you are playing. Poker is a game in which the aim is to win money. The best way to determine how well you are playing poker, then, is to determine how good you are at winning money. Of course it takes a long time for the amount of money you are making to accurately reflect how good you are at making money, but...

[hummusx]

[ QUOTE ]
If you play enough hands your ROI will eventually come to reflect exactly what you seem to desire it to: how well you are playing. Poker is a game in which the aim is to win money. The best way to determine how well you are playing poker, then, is to determine how good you are at winning money. Of course it takes a long time for the amount of money you are making to accurately reflect how good you are at making money, but...

[/ QUOTE ]

That's kind of the whole point, isn't it. People learn by reinforcement one way or the other. Since poker takes an extremely large amount of time for trends one way or the other to emerge, tools that allow us to see and analyze these trends are necessary. 2+2 is one of those tools, because people with large amounts of experience can relate their knowledge. Books are another great tool. Applications that allow us to generalize away from the specific occurences would be another.

[Slim Pickens]

[ QUOTE ]
You appear to be asking for a poker playing AI, nothing less.

[/ QUOTE ]

I thought he was asking for a program that:

1) Goes through your PT database and finds every hand where someone was all-in and greater than 40% of your stack was involved.

2) For all of those hands, run an ICM calc on each situation to generate $EV.
a) Do this with the Villain(s) on some range of hands.
b) Do the same calc with any shown hands.

2a will tell you something about how well you're playing against your selected range of opponents, but of course is limited by your ability to put your opponents on the correct range of hands.

2b is your "luck report" in the short term, and in theory converges to some value in the long term.

Comparing 2a and 2b will, over a large-enough sample, tell you how accurate your range of hands for your generic opponents is. None of this tells you anything about how well you read specific opponents. You might be able to keep track of what ranges you had for opponents in each situation and redo the 2b calcs to get an even better idea of how you read specific opponents. I don't think this is really asking for poker AI.

Slim

[stupidsucker]

I had a better response all typed out, then I went back and re-read the fast responses piling up and realized that everyone seems to be missing the point.

I will chalk this up as my fault for not explaining properly.

What I was suggesting would be simalar to Eastbay's Program. (http://sitngo-analyzer.com/)

It would take in the HHs use a "generic range" that you define. The range is needed in order to give any authenticity to your results. After enough data is collected then the program itself could calculate true calling range and true pushing range, not the one you create.(for example... A2s will be dealt x% of the time and the BB will call an all in 27% of the time they hold it)

I never said this was an easy task.
Nor did I say it was perfect.

I simply stated that using roi/ITM alone to gague if you were doing well is not ideal. So what data is important? How do you extract and anyalize it?

I guess I know whats in my head, but I cant relay it into words. I have always been bad at that.

As for some of the examples given. Most of them didnt fit with what I wanted to do. Analying the first few levels of a SNG is too hard and intricate. Analying pushes and calls is fairly easy if you can. You can go on and on with different factors, but it can be simplified.

You either understand where I am comming from or you dont.

[hedgeyerbets]

I don't think that how often you are correct in such moves statistically would converge to the actual value any more quickly than ROI would converge to it's actual value. So this might not really accomplish anything unless you looked at each hand individually and analyzed.

[stupidsucker]

[ QUOTE ]
I disagree completely. The point of SnGs is to win money.

[/ QUOTE ]

Exactly.. and roi is perfect to guage this IF you never play another sng again.

but if you want projected roi, then you need to be looking at what is going to win you money in the long run.

ROI only tells you where you have been, not where you are going.

Also... for the record Slim gets me.

Is it perfect? no
Is it better then using roi to base your mood or poker playing on? YOU betcha!

[tech]

[ QUOTE ]
1) Putting your opponent on a faulty range;

2) Not calculating or miscalculating the appropriate thing to do given this range.


[/ QUOTE ]

My goal is to elimnate the Type 2 errors above. Every night I go through all the tournaments I played that day using Eastbay's tournament browser. I am looking for any mistakes I made (either not pushing/calling when I should have, or pushing/calling when I should not have).