all in luck calculator

[stinkypete]

a few months ago a bunch of people were working on "all in luck" calculators. (i can't find the thread now)

did anyone ever end up making a program that did this correctly?

i know phil is still working on poker EV but he doesn't seem particularly interested in the all in luck part of the program. but it would be nice to have such a program that a) calculates multiway pots correctly (or even just ignores them) and b) uses the entire pot size rather than just the amount that went in on the last street.

[_dave_]

if it helps your searching, advis0r posted the java program.

I think the source was available, not sure tho.

EDIT: original (PHP)( version was posted by wildzer0 - that proba had source.

[Phil153]

[ QUOTE ]
uses the entire pot size rather than just the amount that went in on the last street.

[/ QUOTE ]
I'm not interested in giving people a flawed method that overestimates their good or bad luck. I'd rather take a somewhat smaller component that isn't overstated for people like setminers. That's the only reason I haven't added the "pot size" functionality. Total pot size on the all-in street was the first method I used when putting together the all-in graphs, before seeing it was flawed and changing it to my current method. You couldn't be more wrong in thinking the full pot size method is superior.

To answer your question, I don't think there were a bunch of people doing this, apart from the above two. The one other was the pokerluckometer guy (maker of set-o-meter) who said this was in the works a few months back. He may have made some progress since then.

[stinkypete]

[ QUOTE ]
I'm not interested in giving people a flawed method that overestimates their good or bad luck. I'd rather take a somewhat smaller component that isn't overstated for people like setminers. That's the only reason I haven't added the "pot size" functionality. Total pot size on the all-in street was the first method I used when putting together the all-in graphs, before seeing it was flawed and changing it to my current method. You couldn't be more wrong in thinking the full pot size method is superior.

[/ QUOTE ]

i'm not sure why you think this. i'd love to see some sort of analysis that suggests that you're right. i read what you wrote in the other thread and none of it was the last bit convincing.

your method *may* be a better method for determining short-term luck (though i haven't analyzed it enough to know if that's the case), but my method is a better method for determining luck in the long term and hence a better method for nailing down your winrate. and that's all i'm interested in. and it's something a lot of others would be interested in too.

[Phil153]

[ QUOTE ]
your method *may* be a better method for determining short-term luck (though i haven't analyzed it enough to know if that's the case), but my method is a better method for determining luck in the long term

[/ QUOTE ]
What do you base this on? My method is identical to your method, except that your method assumes that the previous streets went in with the same equity as the all-in street. I ignore the component from previous streets, since it can overestimate luck as the relationship between assumed and actual equity will based on how you play (i.e. setminers vs lags).

What use is a metric that uses that whole pot, but doesn't take into account the amount of money that went in on that street? According to your method, it would be perfectly valid to run all preflop all-ins as occurring on the flop - since this situation is identical to one where you saw a flop with $1 left behind and then put it in. That such strange situations can occur - causing very different and unpredictable deviations - suggests to me that this method is probably better avoided.

To try and explain, let's take two scenarios where you lose a hand.

Scenario 1
Preflop: $1000 stacks. You have KK. He has AA. 100% of your stacks go in preflop.

Say 20% equity preflop...your expected total win is $400. Your actual total win is $0. You ran bad to the tune of $400.

Scenario 2

Preflop: $1000 stacks. You have KK. He has AA. 80% of your stacks go in preflop. You flop a set, and the final 20% goes in on the flop.

You have (say) 90% equity on the flop...your expected total win is $1800. Your actual total win is $0. You ran bad to the tune of $1800.

The luck and action in both situations is nearly identical...and yet one gives you a luck factor of $1800, and the other gives you a luck factor of $400. That's a 70PTBB differential between two almost identical situations - I don't like that idea. I prefer something that intuitively matches up with the $ value of EV for the entire hand - and ignores the previous streets where this is unreliable.

The problem with your method arises when your early street and late street equity aren't congruent, and this differential can vary based on WtSD% and typical opponent play. In other words, your deviation can be exaggerated entirely by your style, even though luck doesn't change. I choose to ignore this portion of luck, and calculate a smaller portion where your deviation isn't exaggerated by your style.

But honestly, the differences aren't huge, since in almost all scenarios the majority of the money goes in on the final street. I guess if you shortstack it might be more noticable.

If you have the patience, I'm open to hearing why you think your method is a better estimate of long term deviation (as opposed to short term), and why you think this potential overestimation is worth the somewhat larger portion of luck that's calculated under your method.

[stinkypete]

80% of stacks go in preflop. with AA vs KK. it doesn't matter whether a set is flopped, the money is going in anyway. 8 times out of 9 AA is ahead on the flop and is something like a 90% favourite. 90%*$2000*8/9 = $1600.


1 out of 9 times AA will be behind with like 10% equity:
10%*$2000/9 = $22.2.

so AA average equity is $1622.22.


let's say you had AA in all of these situations and you run bad and KK wins 3 of these 9 times. your total win is 6x$2000=$12000 when you should have won 9*$1622.22 = $14600. you ran bad by $2600 by my method.

if you only consider the money that went in on the flop, you ran bad by $260.

clearly $2600 is a better estimate than $260 in this case. if you only consider preflop equity, which is reasonable here, you should have won $14750.28, or $2750.28 more than you actually won.

[stinkypete]

ie. in the long run, your method will underestimate the luck factor. in the very short run, it may be a better estimator in extreme cases.

[Adrian20XX]

[ QUOTE ]
I prefer something that intuitively matches up with the $ value of EV for the entire hand - and ignores the previous streets where this is unreliable.


[/ QUOTE ]

How about a method that weight's previous streets? I think it might be better than both methods.

I'll explain with an example, and I'll take AA against TT instead of AA against KK (cause TT can fold on other flops, KK against AA it usually ends all in always).

Stacks Pre Flop are 1000.
Pre Flop bet is 4BB, $40.

Flop now comes Td7c3h.

Equity on the flop is: 91.41% for TT and 8.59% for AA
Now we calculate the equity pre flop.

[Adrian20XX]

[ QUOTE ]
I prefer something that intuitively matches up with the $ value of EV for the entire hand - and ignores the previous streets where this is unreliable.

[/ QUOTE ]

Sorry, pressed the button by mistake in the middle.

How about a method that weight's previous streets? I think it might be better than both methods.

I'll explain with an example, and I'll take AA against TT instead of AA against KK (cause TT can fold on other flops, KK against AA it usually ends all in always).

Stacks Pre Flop are 1000.
Pre Flop bet is 4BB, $40.

Flop now comes Td7c3h.

Equity on the flop is: 91.41% for TT and 8.59% for AA

Now we calculate the equity pre flop, weighted by the percentage of time TT is going to have EV 91.41% on the flop or better. So, we take it from the following Equity Distribution Graph for TT againt AA on the Flop

So we aproximate the number that TT has 91.41% EV or more against AA to be 5% (I've seen somewhere that the graph can be obtained as a table, but I don't remember where).

Now we can compute the weighted EV pre flop for TT and AA, and it is -$36 for TT and +$36 for AA.

And, we can compute the weighted EV on the flop for TT and AA, and it is +$39,8496 for TT and -$39,8496 for TT. So, total weighted EV for TT for the hand is $3,8496 and for AA is -$3,8496.

Hand Board Stack Pre Flop Bet Flop % EV Pre Flop Bet % EV Flop Flop Bet EV Flop Bet EV
TT Td7c3h 1000 40 5,00% -36 91,51% 960 39,8496 3,8496
AA Td7c3h 1000 40 95,00% 36 8,49% 960 -39,8496 -3,8496

We can compute in the same way with bets of 20% of the stacks and 80% of the stacks pre flop, and we have EV of +$ 146 and +$711 for AA, as we expect EV grows for AA as more money is put pre flop.

Hand Board Stack Pre Flop Bet Flop % EV Pre Flop Bet % EV Flop Flop Bet EV Flop Bet EV
TT Td7c3h 1000 200 5,00% -180 91,51% 800 33,208 -146,792
AA Td7c3h 1000 200 95,00% 180 8,49% 800 -33,208 146,792

Hand Board Stack Pre Flop Bet Flop % EV Pre Flop Bet % EV Flop Flop Bet EV Flop Bet EV
TT Td7c3h 1000 800 5,00% -720 91,51% 200 8,302 -711,698
AA Td7c3h 1000 800 95,00% 720 8,49% 200 -8,302 711,698

[Phil153]

Adrian - this is the way it's done on the game analysis tab. The method is flawed as the results can skew depending on how you play,because you self select hands that go to showdown.

If you prefer to use this method, check "all-in before river" on the game analysis graphs tab.