CardRunnersEV computation of EV does not match my computation

LeChiffreLeChiffre Red Chipper Posts: 140 ✭✭
Hi all,
In this thread I asked how to compute hero's EV of calling turn in the following hand.

PokerStars - $0.02 NL FAST (6 max) - Holdem - 6 players
Hand converted by PokerTracker 4

BTN: 131.5 BB
SB: 114 BB
BB: 112 BB
Hero (UTG): 100 BB
MP: 32 BB
CO: 192 BB

SB posts SB 0.5 BB, BB posts BB 1 BB

Pre Flop: (pot: 1.5 BB) Hero has :AC: :KC:

Hero raises to 4 BB, MP calls 4 BB, fold, fold, SB raises to 13 BB, fold, Hero calls 9 BB, MP calls 9 BB

Flop: (40 BB, 3 players) :3H: :JS: :AS:
SB bets 26 BB, Hero calls 26 BB, MP calls 19 BB and is all-in

Turn: (111 BB, 3 players) :QH:
SB bets 61 BB, Hero ?

After some discussion with @jeffnc we arrived at an EV but this does not match the one computed by CREV.

For the sake of argument I assumed the following ranges of the villains:
SB: AA, AK, AQ and JJ
MP: A9, AT and KT

Of course these ranges don't make much sense, it was just about the pure maths behind calling, so it is not about gameplay elements, strategy or hand reading.

This is the tree that I created in CREV. I am of course only interested in the EV of the turn call. I just picked the above ranges and recreated the line of actions, taking into account correct stack sizes and what have you.

anddzrhq6e12.png

As you can see the EV of a call on the turn is -19.4.

The problem is that this does not match the EV we computed by hand.
We looked at the EV in the main pot and side pot seperately.

Quoting Jeff here in the original thread:

"Your equity in the main pot against both players is only 12%. There were 40BB in the pot preflop, and an all-in for 19 more for a total of 97BB. So you figure to get about 12BB there.

Your equity in the side pot against SB is 26%. There is 14BB in the side pot plus the 61BB for a total of 75BB. So you figure to get about 20BB there.

Your total equity is 12+20=30. You are being asked to put in 61BB on the turn with an expectation of getting 30BB back."


So that would mean an EV of around -30, which is not the -19.4 computed by CREV.
There are three possible reasons for this in my estimation:
1) CREV did something wrong
2) Our EV computation is wrong
3) I did something wrong entering the tree in CREV

Any help would be greatly appreciated here :) cheers!


Comments

  • jeffncjeffnc Red Chipper Posts: 4,039 ✭✭✭✭✭
    LeChiffre wrote: »
    Quoting Jeff here in the original thread: "Your equity in the main pot against both players is only 12%. There were 40BB in the pot preflop, and an all-in for 19 more for a total of 97BB. So you figure to get about 12BB there.

    Your equity in the side pot against SB is 26%. There is 14BB in the side pot plus the 61BB for a total of 75BB. So you figure to get about 20BB there.

    Your total equity is 12+20=30. You are being asked to put in 61BB on the turn with an expectation of getting 30BB back."

    Actually I think that's wrong, but the question is still valid because I can't come up with the number CREV comes up with.

    It's a little confusing because you're putting $61 into the side pot, but not into the main pot, and yet you still have to pay $61 to get any of either pot.

    So I think the right thing to do is add your $61 to the side pot, but not the main pot, and then subtract your $61 call from everything, as a cost. In a normal pot you wouldn't need to do this because you'd just be adding and subtracting the $61, but here it's only associated with one of the equities and not the other.

    main pot equity * main pot$, plus side pot equity * side pot$ (including your call), and subtract out the cost of your call.

    .12($97) + .26($14+$61+$61) - $61 = -$14

    I think that's right, but still doesn't match CREV. Calling @SplitSuit ?

  • LeChiffreLeChiffre Red Chipper Posts: 140 ✭✭
    edited May 13
    LeChiffre wrote: »

    Your equity in the side pot against SB is 26%. There is 14BB in the side pot plus the 61BB for a total of 75BB. So you figure to get about 20BB there.




    FWIW this not entirely true either. We would have 26% if MP was not in the hand. The thing is that with every holding in his KT,AT,A9 range he changes my equity cause of blockers.

    Decided to just go through every combo in his range and make each individual hand dead in Equilab. My equity for every combo in MP's range against SB is then:

    AdTd 22.62
    AhTh 23.81
    KdTd 20.66
    KhTh 21.43
    KsTs 20.66
    Ad9d 24.21
    Ah9h 25.40
    AdTh 22.75
    AdTs 22.62
    AdTc 22.62
    AhTd 23.81
    AhTs 23.81
    AhTc 23.81
    Ad9h 24.34
    Ad9s 24.21
    Ad9c 24.21
    Ah9d 25.40
    Ah9s 25.40
    Ah9c 25.40
    KdTh 20.75
    KdTs 20.66
    KdTc 20.66
    KhTd 21.43
    KhTs 21.43
    KhTc 21.43
    KsTd 20.66
    KsTh 20.75
    KsTc 20.66

    Which is an average of 22.7%, not 26%. So I should get 17BB from the side pot, not 20.

  • jeffncjeffnc Red Chipper Posts: 4,039 ✭✭✭✭✭
    LeChiffre wrote: »
    FWIW this not entirely true either. We would have 26% if MP was not in the hand. The thing is that with every holding in his KT,AT,A9 range he changes my equity cause of blockers.

    Decided to just go through every combo in his range and make each individual hand dead in Equilab.

    Yeah that's absolutely true, I was going to mention that at some point that we were actually "rounding off" the dead cards so to speak. And I was going to mention there's no way to enter a range of dead cards in Equilab, so it's too much trouble, so kudos for doing all that manually, lol

  • LeChiffreLeChiffre Red Chipper Posts: 140 ✭✭
  • Adam WheelerAdam Wheeler Red Chipper Posts: 2,515 ✭✭✭✭
    Why don't you just write to Scylla instead ?
  • LeChiffreLeChiffre Red Chipper Posts: 140 ✭✭
    What do you mean?
  • LeChiffreLeChiffre Red Chipper Posts: 140 ✭✭
    OK, I figured it out. I narrowed villains' ranges to the following:

    MP:
    AdTd
    AhTh
    KhTh

    SB:
    AdAh
    AdKd
    AhKh

    Then I went over every possible scenario and computed the EV according to Jeff's formula:

    main pot equity * main pot$, plus side pot equity * side pot$ (including your call)

    MP has AdTd:

    -SB has AdAh
    Not possible

    -SB has AdKd
    Not possible

    -SB has AhKh
    36.9% of 97
    +
    39.29% of 136
    Total = 89.2

    MP has AhTh:

    -SB has AdAh
    Not possible

    -SB has AdKd
    38.10% of 97
    +
    50% of 136
    Total = 105

    -SB has AhKh
    Not possible

    MP has KhTh:

    - SB has AdAh
    3.57% of 97
    +
    7.14% of 136
    total = 13.2

    - SB has AdKd
    2.38% of 97
    +
    50% of 75+61
    total = 70.3

    - SB has AhKh
    Not possible

    Possible outcomes:
    70.3
    13.2
    105
    89.2

    Average:
    69.4

    The call to make is 61, so the EV is +8.4
    This matches the one from CREV:

    co40wigdpkej.png

    Now the question is whether there is a quicker way to do this =)
  • Adam WheelerAdam Wheeler Red Chipper Posts: 2,515 ✭✭✭✭
    LeChiffre wrote: »
    What do you mean?
    Write to support.
  • jeffncjeffnc Red Chipper Posts: 4,039 ✭✭✭✭✭
    Yeah that's a good idea to narrow down the range to test out the answer. And no I don't think there's a faster way to do it.

    So like I was saying, I had thought of this, but at the point I was going to mention it, I was off on that side experiment with giving MP a hand like 22, and of course the blocking there is so minimal it couldn't have explained the difference between our answer and CREV's, so I put it on the back burner. But of course the blocking of MP's real range is very significant when it blocks most of the cards we need, so I'm sure that explains it.

    So, looks like you have it figured out, CREV was right and we just had no good way to write an equation. Programs like that don't figure things out with equations and probabilities, they use brute force techniques (basically what you did). They either run all possible combinations out when the numbers are reasonable, and when the combinations of runouts gets outrageously large, they use things like Monte Carlo algorithms, which are way too complicated for us to deal with unless we get paid to do it :)
    https://en.wikipedia.org/wiki/Monte_Carlo_algorithm
  • LeChiffreLeChiffre Red Chipper Posts: 140 ✭✭
    Yup, so it seems. I guess that's it then. CREV provides the answer, so I will just play around there with ranges and pot sizes, etc.

    Thanks for the help!
  • SplitSuitSplitSuit RCP Coach Posts: 3,838 -
    jeffnc wrote: »
    I think that's right, but still doesn't match CREV. Calling @SplitSuit ?

    I'm not the CREV guy =(
    My latest poker course brings the popular book 'Poker's 1%' to life- The One Percent
  • jeffncjeffnc Red Chipper Posts: 4,039 ✭✭✭✭✭
    No I know, but you do combo math pretty well, lol
  • SplitSuitSplitSuit RCP Coach Posts: 3,838 -
    jeffnc wrote: »
    No I know, but you do combo math pretty well, lol

    :blushes:
    My latest poker course brings the popular book 'Poker's 1%' to life- The One Percent

Leave a Comment

BoldItalicStrikethroughOrdered listUnordered list
Emoji
Image
Align leftAlign centerAlign rightToggle HTML viewToggle full pageToggle lights
Drop image/file