Stuey
03-05-2005, 02:02 AM
I need help understanding this. I think there is value in his method but I am not the one to extract it.
GFI by Aaron W. (http://forumserver.twoplustwo.com/showflat.php?Cat=&Number=1833461&page=4&view=colla psed&sb=5&o=14&fpart=1)
[ QUOTE ]
After my month away from posting (and playing -- the only poker playing I had was a single home game with some friends), I'm ready to get back into the swing of things.
When I was last here in January, there were plenty of people who played far too tight (I count myself among them, but not nearly as bad as some others). I was looking for a way to find hands which can be played profitably in situations where the lines between calling and folding are not clearly drawn. I think I've found it. But now comes the true test of letting others take this idea and pick it apart, either with numbers or experience which contradicts it.
One of the things I realized is that I don't really have a good sense of how often hands flop things that I like. This thought followed from reflecting on the TT essay where I tried to define "Good" and "Great" flops for that hand. I wondered if it was possible to do the same for other hands.
But rather than computing the probabilities by hand as I did for the essay, I used my minimal programming skills and wrote a little analysis program (true programmers can laugh - I used Microsoft's QBASIC for this -- at some point, I'll learn a real programming language). I did a spot check by hand and the numbers seemed to be coming out right, but if you want to double check them by repeating the simulations, I would greatly appreciate it.
The raw output is freely accessible and stored as as CSV file (comma separated value) and should be readable by any spreadsheet program. You'll probably need to save the file to your computer, then open it. (It's just a plain text document.) The links are below.
Non Pair Data
Pocket Pair Data
I hope that the classifications are clear enough. Rather than list them all, I'll list a few to give a flavor of the classes of flopped hand types for the non pocket pairs.
- 2 Pair, 2 Cards
- 2 Pair, Higher Card (meaning the higher pair is one of my kickers, like 89 with a 559 flop)
- 2 Pair, Lower Card (A6 on a TT6 flop)
- Top Pair + FD + OESD
- Bot Pair + BDFD + GSD (gutshot straight draw)
- 1 Over + None + None (1 overcard, no flush or straight draws at all)
I then took this data and to come up with a "Good Flop Index" (GFI). A "good" flop is one where I would see the turn. It says nothing about how I would play it, nor what sort of EV I would expect from it. I should point out that implicit in flopping "good" is the fact that I wouldn't go on if I didn't think it could be played profitably.
The next step was to mimic the idea of trying to come up with a measure for the less clear situations. For example, AA flopping an overpair is always going to play on, but TT with an overcard may not. So I weighted each hand type by a factor (0%, 25%, 50%, 75%, 100%) based on how often I guessed I might see the turn.
Why only those levels? It seems a bit silly to try to distinguish things too sharply here. I'm not trying to get an exact value for how often I flop "good", but just trying to build up some intuition. When you consider that there are 19600 flops and most of the categories have on the order of 100 entries, there isn't really that much of an effect when you start adding finer increments (a 10% shift there would represent a shift of about .05% in the final number).
[In order to save space, I'll include the GFI weighting scheme I used in a response to this post.]
Once each category is given a weight, I take a weighted sum to find a number which represents the total count of "good flops". This number is normalized by dividing by 19600 (giving me a percent), and this value is the GFI of a given hand.
I have also saved that as a CSV (though you can read it straight off as text because there are only three columns). The first column is the hand (obviously), the second column is the GFI, and the third column gives the odds of a good flop.
GFI values
Here are a few values:
AA 1 0
KK 0.91 0.10
QQ 0.81 0.23
JJ 0.71 0.40
TT 0.58 0.72
AKs 0.58 0.73
AJs 0.53 0.89
AKo 0.51 0.96
98s 0.36 1.75
K8s 0.35 1.86
A2s 0.30 2.31
72o 0.09 10.52
Why is this at all useful? I'm hoping that these values will give me a good sense of how hands rate in a pot that's contested between short-handed and multi-way. For short-handed play, your play depends much more on what you think of your opponent. In multi-way pots, most of the hands you play are based on either the large immediate equity (strong hands) or the large implied odds (because of the many players you expect will pay you off when the flop hits you). In the between cases, say 3-4 opponents, you need to balance the read with the chances of flopping good to make a correct decision.
What might be an application of the chart? I pick up a marginal hand such as K8s in late position against a slightly loose player whose postflop play isn't good, but not bad, either. The blinds are bad. Am I getting enough value out of my hand to make it worth mixing it up with them? According to the chart, I'm going to get a good flop about 35% of the time. There are 2.5 SB in the pot now, giving me odds to chase a good flop. I might even consider raising here because, assuming both blinds fold, I'm getting 3.5:2 = 1.75:1 on my 1.86:1 shot (the advantage of being the agressor, position, and so on make up for the measly amount I'm short). But even if the blinds don't fold, they add to the odds I'm getting preflop to catch a profitable flop.
I'm not entirely thrilled with the analysis as it stands. I definitely think there's room for improvement on balancing the weights. I think pocket pairs are a bit overvalued, flopping middle pairs might be undervalued, draws might be overvalued... I consider my experience to be insufficient to place these numbers confidently. But the hard work of getting the data is done and now it's about playing with the numbers.
By the way, if anyone cross-posts this, I'd like to know... I want to follow the discussion (if any) wherever it goes.
[/ QUOTE ]
The links in the quote don't work please see the original to review his numbers.
GFI by Aaron W. (http://forumserver.twoplustwo.com/showflat.php?Cat=&Number=1833461&page=4&view=colla psed&sb=5&o=14&fpart=1)
[ QUOTE ]
After my month away from posting (and playing -- the only poker playing I had was a single home game with some friends), I'm ready to get back into the swing of things.
When I was last here in January, there were plenty of people who played far too tight (I count myself among them, but not nearly as bad as some others). I was looking for a way to find hands which can be played profitably in situations where the lines between calling and folding are not clearly drawn. I think I've found it. But now comes the true test of letting others take this idea and pick it apart, either with numbers or experience which contradicts it.
One of the things I realized is that I don't really have a good sense of how often hands flop things that I like. This thought followed from reflecting on the TT essay where I tried to define "Good" and "Great" flops for that hand. I wondered if it was possible to do the same for other hands.
But rather than computing the probabilities by hand as I did for the essay, I used my minimal programming skills and wrote a little analysis program (true programmers can laugh - I used Microsoft's QBASIC for this -- at some point, I'll learn a real programming language). I did a spot check by hand and the numbers seemed to be coming out right, but if you want to double check them by repeating the simulations, I would greatly appreciate it.
The raw output is freely accessible and stored as as CSV file (comma separated value) and should be readable by any spreadsheet program. You'll probably need to save the file to your computer, then open it. (It's just a plain text document.) The links are below.
Non Pair Data
Pocket Pair Data
I hope that the classifications are clear enough. Rather than list them all, I'll list a few to give a flavor of the classes of flopped hand types for the non pocket pairs.
- 2 Pair, 2 Cards
- 2 Pair, Higher Card (meaning the higher pair is one of my kickers, like 89 with a 559 flop)
- 2 Pair, Lower Card (A6 on a TT6 flop)
- Top Pair + FD + OESD
- Bot Pair + BDFD + GSD (gutshot straight draw)
- 1 Over + None + None (1 overcard, no flush or straight draws at all)
I then took this data and to come up with a "Good Flop Index" (GFI). A "good" flop is one where I would see the turn. It says nothing about how I would play it, nor what sort of EV I would expect from it. I should point out that implicit in flopping "good" is the fact that I wouldn't go on if I didn't think it could be played profitably.
The next step was to mimic the idea of trying to come up with a measure for the less clear situations. For example, AA flopping an overpair is always going to play on, but TT with an overcard may not. So I weighted each hand type by a factor (0%, 25%, 50%, 75%, 100%) based on how often I guessed I might see the turn.
Why only those levels? It seems a bit silly to try to distinguish things too sharply here. I'm not trying to get an exact value for how often I flop "good", but just trying to build up some intuition. When you consider that there are 19600 flops and most of the categories have on the order of 100 entries, there isn't really that much of an effect when you start adding finer increments (a 10% shift there would represent a shift of about .05% in the final number).
[In order to save space, I'll include the GFI weighting scheme I used in a response to this post.]
Once each category is given a weight, I take a weighted sum to find a number which represents the total count of "good flops". This number is normalized by dividing by 19600 (giving me a percent), and this value is the GFI of a given hand.
I have also saved that as a CSV (though you can read it straight off as text because there are only three columns). The first column is the hand (obviously), the second column is the GFI, and the third column gives the odds of a good flop.
GFI values
Here are a few values:
AA 1 0
KK 0.91 0.10
QQ 0.81 0.23
JJ 0.71 0.40
TT 0.58 0.72
AKs 0.58 0.73
AJs 0.53 0.89
AKo 0.51 0.96
98s 0.36 1.75
K8s 0.35 1.86
A2s 0.30 2.31
72o 0.09 10.52
Why is this at all useful? I'm hoping that these values will give me a good sense of how hands rate in a pot that's contested between short-handed and multi-way. For short-handed play, your play depends much more on what you think of your opponent. In multi-way pots, most of the hands you play are based on either the large immediate equity (strong hands) or the large implied odds (because of the many players you expect will pay you off when the flop hits you). In the between cases, say 3-4 opponents, you need to balance the read with the chances of flopping good to make a correct decision.
What might be an application of the chart? I pick up a marginal hand such as K8s in late position against a slightly loose player whose postflop play isn't good, but not bad, either. The blinds are bad. Am I getting enough value out of my hand to make it worth mixing it up with them? According to the chart, I'm going to get a good flop about 35% of the time. There are 2.5 SB in the pot now, giving me odds to chase a good flop. I might even consider raising here because, assuming both blinds fold, I'm getting 3.5:2 = 1.75:1 on my 1.86:1 shot (the advantage of being the agressor, position, and so on make up for the measly amount I'm short). But even if the blinds don't fold, they add to the odds I'm getting preflop to catch a profitable flop.
I'm not entirely thrilled with the analysis as it stands. I definitely think there's room for improvement on balancing the weights. I think pocket pairs are a bit overvalued, flopping middle pairs might be undervalued, draws might be overvalued... I consider my experience to be insufficient to place these numbers confidently. But the hard work of getting the data is done and now it's about playing with the numbers.
By the way, if anyone cross-posts this, I'd like to know... I want to follow the discussion (if any) wherever it goes.
[/ QUOTE ]
The links in the quote don't work please see the original to review his numbers.