PDA

View Full Version : Error in tournament EV calculation, 3 way all in



brbrbr
02-18-2010, 03:56 PM
This is the line in error, the end result should be $0, but it says $4.29.

EV Result Luck Hand Player
...
$5.41 $4.29 ($1.11) JsJc chips23f

So it shows this wrong value in the Tourney-results-data view list, both in hand and tournament list.

It seems that it missed the fact that player was out of the tournament, so, the luck part is calculated wrong, it should be (5.41). This was 3 players all in preflop.

I'm including the hand history for the hand, also the ICM text from the replayer.

This is 1.10 beta 5, Full tilt poker. Hand in replayer plays ok.

---hand log---
Full Tilt Poker Game #18569106797: $3.50+$0.30 Sit&Go (Sup Turbo) (140800650), Table 1 - 15/30 - No Limit Hold'em - 14:07:29 ET - 2010/02/18
Seat 1: tinta7 (285)
Seat 2: ldh1992 (255)
Seat 4: 86JACK86 (300)
Seat 5: AAKamda (270)
Seat 6: chips23f (300)
Seat 7: XXTerminatorXX1 (323)
Seat 8: thingsandstuff (645)
Seat 9: 1hothandz1 (322), is sitting out
86JACK86 posts the big blind of 30
The button is in seat #2
*** HOLE CARDS ***
Dealt to chips23f [Js Jc]
AAKamda folds
chips23f raises to 300, and is all in
XXTerminatorXX1 has 15 seconds left to act
XXTerminatorXX1 folds
thingsandstuff raises to 570
1hothandz1 folds
tinta7 folds
ldh1992 folds
86JACK86 calls 270, and is all in
thingsandstuff shows [Ah Qs]
86JACK86 shows [Tc Ts]
chips23f shows [Js Jc]
Uncalled bet of 270 returned to thingsandstuff
*** FLOP *** [5h 6d Th]
*** TURN *** [5h 6d Th] [6s]
*** RIVER *** [5h 6d Th 6s] [9c]
thingsandstuff shows a pair of Sixes
86JACK86 shows a full house, Tens full of Sixes
chips23f shows two pair, Jacks and Sixes
86JACK86 wins the pot (900) with a full house, Tens full of Sixes
*** SUMMARY ***
Total pot 900 | Rake 0
Board: [5h 6d Th 6s 9c]
Seat 1: tinta7 didn't bet (folded)
Seat 2: ldh1992 (button) didn't bet (folded)
Seat 4: 86JACK86 (big blind) showed [Tc Ts] and won (900) with a full house, Tens full of Sixes
Seat 5: AAKamda didn't bet (folded)
Seat 6: chips23f showed [Js Jc] and lost with two pair, Jacks and Sixes
Seat 7: XXTerminatorXX1 didn't bet (folded)
Seat 8: thingsandstuff showed [Ah Qs] and lost with a pair of Sixes
Seat 9: 1hothandz1 didn't bet (folded)




----icm----

Payouts
1 $15.75
2 $9.45
3 $6.30

Start EV End EV Diff Player
$3.44 $3.63 $0.19 tinta7
$3.11 $3.28 $0.17 ldh1992
$3.61 $8.74 $5.14 86JACK86
$3.28 $3.46 $0.18 AAKamda
$3.61 $0.00 ($3.61) chips23f
$3.85 $4.05 $0.20 XXTerminatorXX1
$6.77 $4.29 ($2.47) thingsandstuff
$3.84 $4.04 $0.20 1hothandz1

SNG EV analysis

Board -

Prizes $15.75 $9.45 $6.30



EV Result Luck Hand Player

$3.98 $3.63 ($0.35) tinta7

$3.61 $3.28 ($0.32) ldh1992

$1.53 $8.74 $7.21 TcTs 86JACK86

$3.79 $3.46 ($0.34) AAKamda

$5.41 $4.29 ($1.11) JsJc chips23f

$4.44 $4.05 ($0.38) XXTerminatorXX1

$4.32 $0.00 ($4.32) AhQs thingsandstuff

$4.43 $4.04 ($0.38) 1hothandz1



hand order TcTs > JsJc > AhQs probability 9.9790%

0 stack 285, icm ev $3.63

1 stack 255, icm ev $3.28

2 stack 900, icm ev $8.74

3 stack 270, icm ev $3.46

4 stack 345, icm ev $4.29

5 stack 323, icm ev $4.05

6 stack 0, icm ev $0.00

7 stack 322, icm ev $4.04



hand order TcTs > AhQs > JsJc probability 7.2707%

0 stack 285, icm ev $3.63

1 stack 255, icm ev $3.28

2 stack 900, icm ev $8.74

3 stack 270, icm ev $3.46

4 stack 0, icm ev $0.00

5 stack 323, icm ev $4.05

6 stack 345, icm ev $4.29

7 stack 322, icm ev $4.04



hand order JsJc > TcTs > AhQs probability 38.4110%

0 stack 285, icm ev $4.06

1 stack 255, icm ev $3.68

2 stack 0, icm ev $0.00

3 stack 270, icm ev $3.87

4 stack 1245, icm ev $10.86

5 stack 323, icm ev $4.52

6 stack 0, icm ev $0.00

7 stack 322, icm ev $4.51



hand order JsJc > AhQs > TcTs probability 7.1746%

0 stack 285, icm ev $4.06

1 stack 255, icm ev $3.68

2 stack 0, icm ev $0.00

3 stack 270, icm ev $3.87

4 stack 1245, icm ev $10.86

5 stack 323, icm ev $4.52

6 stack 0, icm ev $0.00

7 stack 322, icm ev $4.51



hand order AhQs > TcTs > JsJc probability 0.4062%

0 stack 285, icm ev $4.06

1 stack 255, icm ev $3.68

2 stack 0, icm ev $0.00

3 stack 270, icm ev $3.87

4 stack 0, icm ev $0.00

5 stack 323, icm ev $4.52

6 stack 1245, icm ev $10.86

7 stack 322, icm ev $4.51



hand order AhQs > JsJc > TcTs probability 36.1202%

0 stack 285, icm ev $4.06

1 stack 255, icm ev $3.68

2 stack 0, icm ev $0.00

3 stack 270, icm ev $3.87

4 stack 0, icm ev $0.00

5 stack 323, icm ev $4.52

6 stack 1245, icm ev $10.86

7 stack 322, icm ev $4.51



hand order TcTs = JsJc > AhQs probability 0.0000%

0 stack 285, icm ev $3.57

1 stack 255, icm ev $3.23

2 stack 450, icm ev $5.29

3 stack 270, icm ev $3.40

4 stack 795, icm ev $8.03

5 stack 323, icm ev $3.99

6 stack 0, icm ev $0.00

7 stack 322, icm ev $3.98



hand order TcTs = AhQs > JsJc probability 0.0000%

0 stack 285, icm ev $3.57

1 stack 255, icm ev $3.23

2 stack 450, icm ev $5.29

3 stack 270, icm ev $3.40

4 stack 0, icm ev $0.00

5 stack 323, icm ev $3.99

6 stack 795, icm ev $8.03

7 stack 322, icm ev $3.98



hand order JsJc = AhQs > TcTs probability 0.0000%

0 stack 285, icm ev $3.54

1 stack 255, icm ev $3.20

2 stack 0, icm ev $0.00

3 stack 270, icm ev $3.37

4 stack 622.5, icm ev $6.75

5 stack 323, icm ev $3.96

6 stack 622.5, icm ev $6.75

7 stack 322, icm ev $3.95



hand order TcTs > JsJc = AhQs probability 0.0713%

0 stack 285, icm ev $3.59

1 stack 255, icm ev $3.25

2 stack 900, icm ev $8.69

3 stack 270, icm ev $3.42

4 stack 172.5, icm ev $2.26

5 stack 323, icm ev $4.02

6 stack 172.5, icm ev $2.26

7 stack 322, icm ev $4.01



hand order JsJc > TcTs = AhQs probability 0.0386%

0 stack 285, icm ev $4.06

1 stack 255, icm ev $3.68

2 stack 0, icm ev $0.00

3 stack 270, icm ev $3.87

4 stack 1245, icm ev $10.86

5 stack 323, icm ev $4.52

6 stack 0, icm ev $0.00

7 stack 322, icm ev $4.51



hand order AhQs > TcTs = JsJc probability 0.1264%

0 stack 285, icm ev $4.06

1 stack 255, icm ev $3.68

2 stack 0, icm ev $0.00

3 stack 270, icm ev $3.87

4 stack 0, icm ev $0.00

5 stack 323, icm ev $4.52

6 stack 1245, icm ev $10.86

7 stack 322, icm ev $4.51



hand order TcTs = JsJc = AhQs probability 0.4021%

0 stack 285, icm ev $3.41

1 stack 255, icm ev $3.08

2 stack 300, icm ev $3.57

3 stack 270, icm ev $3.25

4 stack 472.5, icm ev $5.28

5 stack 323, icm ev $3.82

6 stack 472.5, icm ev $5.28

7 stack 322, icm ev $3.81

Patvs
02-18-2010, 10:22 PM
Please update to the latest version - http://www.holdemmanager.com/downloads/HmBetaUpdate.exe

You lose the hand JJ vs TT vs AQ. (preflop allin)
The buyin is $3.50. // 300 startstacks.



This is the line in error, the end result should be $0, but it says $4.29

Looks good to me.... had you WON the hand...you'd have a monster stack with a large equity to finish first of second.


Let's simplify (extreme simplification) and state:

You had a 50% to WIN the hand---> Equity -->$9
And a 50% to loose the hand--> Equity--> $0.
So you're EV difference as a direct result of this hand is $4.50. (it's irrelevant if you won/lost is.. had you won it.. the EV would be -$4.50)
(there are also some splitpot scenario's! that influence the EV)

The Minder
02-19-2010, 01:10 AM
And just for the hell of it, I'm going to throw this into the mix:
Expected Value in Poker - A Lousy Concept For the WSOP (or Any Tournament) (http://ezinearticles.com/?Expected-Value-in-Poker---A-Lousy-Concept-For-the-WSOP-(or-Any-Tournament)&id=2208350)

brbrbr
02-19-2010, 10:30 AM
Thanks Patvs, I understand the math, but if you look closely at the two ICM tables, they don't agree with each other. In the first one, chips23 ended up with EV zero, and in second one, he ends up with ev of 4.29, despite being out of the tournament.

Now that I looked more at it, seems that lines 5 and 7 of the second table are switched, since the result in the 7th line is correct for the guy that dropped out of the tournament (at least it seems so). So this is probably not a math bug but more like somewhere in handling lists of players on the table.





Start EV End EV Diff Player
$3.44 $3.63 $0.19 tinta7
$3.11 $3.28 $0.17 ldh1992
$3.61 $8.74 $5.14 86JACK86
$3.28 $3.46 $0.18 AAKamda
$3.61 $0.00 ($3.61) chips23f <---XXXXXXXXXXXXXXX
$3.85 $4.05 $0.20 XXTerminatorXX1
$6.77 $4.29 ($2.47) thingsandstuff
$3.84 $4.04 $0.20 1hothandz1

SNG EV analysis
Board -
Prizes $15.75 $9.45 $6.30

EV Result Luck Hand Player
$3.98 $3.63 ($0.35) tinta7
$3.61 $3.28 ($0.32) ldh1992
$1.53 $8.74 $7.21 TcTs 86JACK86
$3.79 $3.46 ($0.34) AAKamda
$5.41 $4.29 ($1.11) JsJc chips23f <---- XXXXXXXXXXXX
$4.44 $4.05 ($0.38) XXTerminatorXX1
$4.32 $0.00 ($4.32) AhQs thingsandstuff
$4.43 $4.04 ($0.38) 1hothandz1




I upgraded to the 1.10.02 (the latest version) but it still looks the same.

brbrbr
02-19-2010, 11:14 AM
Another theory would be that, since this is the situation where we have 3 way all in, but no side pot, there is an error in handling this case. This would be a pretty rare situation of having two players with same stacks, and another one with a larger stack.

Patvs
02-19-2010, 02:40 PM
These hands happen all the time in SUPER TURBO.

I don't fully understand your ICM tables.



Start EV End EV Diff Player
$3.44 $3.63 $0.19 tinta7
$3.11 $3.28 $0.17 ldh1992
$3.61 $8.74 $5.14 86JACK86
$3.28 $3.46 $0.18 AAKamda
$3.61 $0.00 ($3.61) chips23f <---XXXXXXXXXXXXXXX
$3.85 $4.05 $0.20 XXTerminatorXX1
$6.77 $4.29 ($2.47) thingsandstuff
$3.84 $4.04 $0.20 1hothandz1



86JACK86 had TT and he won the hand. His EV jumped to $8.74.
If chips23f had won the hand his EV would have jumped to $8.74... but he lost the hand (EV = 0)

JJ vs TT vs AQo: (I'll neglect the 0.30% splitpot)
((JJ wins 46% x $8.74) + (JJ loses 54% x 0)) = $4.02
So this hand has a $4 EV effect on the outcome of the SNG.

brbrbr
02-19-2010, 04:10 PM
Again, I agree with your math, but, again, this is not what the HM shows in the second table. The tables are from the replayer, copied from the text you get when you click the "ICM" button.

I'm aware that I was not very clear, I also tried to format the tables in a naive way (with spaces in a proportional font), but html just ignores extra spaces. So, I'll try it again.

In the ICM "report" you get from the replayer, at the top you get two lists of players and their ICM tourament values, and how they changed in the hand. In the first one, we just get value at the start, value at the end of the hand, and the difference. These are the numbers for our player:

Start ev: $3.61
End ev: $0
Diff: $-3.61.

All good so far, player started with 300 chips, ICM estimate of it is 3.61 (just above starting 3.50 value, since one player was already out). He lost all his chips in this hand, value went to zero. All perfectly clear (and correct).

The second table shows some more numbers, the first one is the expected value after the end of hand, the second number is the actual outcome, the third number is thus his "luck". And this is what the table (the second one) shows:

EV: $5.41
Result: $4.29
Luck: $-1.11

This is incorrect, the result for our player is $0, not 4.29. Thus, the "luck" calculated from this end result is also wrong. This is why I said that first and second table don't agree with each other, our player is out of the tournament, his result number should say 0.

It should be more like:

EV: $4.50 (or something close to this, perhaps even 5.41?)
Result: $0
Luck: $-4.50

In fact, those numbers we can find in the second table, but two lines down, under a different players name (the all in guy with the big stack).


Also, the dump off all the combinations of who beats who and who splits, some of the end results are impossible. The third block (JJ > TT > AQ) says 4th stack (our player) would end up with 1245 chips, this is clearly wrong since the most he can end up with is 900. (There was no small blind, and big blind was one of the three all in guys).

This would support my theory that the uncalled portion of the big stack got mixed up with the empty side pot. The list of all combinations gives both players 4 and 6 a possibility of going to 1245 chips, but only the bigger stack could do that. Also, for both it lists the possibility of going to zero, again this is impossible since only the smaller stack could go to zero.

So, there seems to be a bug in those lists, too.

Patvs
02-19-2010, 04:41 PM
EV: $5.41
Result: $4.29
Luck: $-1.11

This is incorrect, the result for our player is $0, not 4.29. Thus, the "luck" calculated from this end result is also wrong.


Yeah, you're right. Though now the big question is, is the EV difference (what results in the EV graph) also wrong..... or is it just the ICM that shows some incorrect info.

Because I still think any EV difference of $4-$4.29 for this hand is accurate.

brbrbr
02-19-2010, 06:03 PM
The wrong result gets displayed in both the "$EV Diff" column in the list of hands for the tournament, and for the "$EV Won" column in the list of tournaments, both in the Tourney - Results - Data View tab. Can't tell for the graph, its hard to see on the chart with many tournaments.

morny
02-19-2010, 06:49 PM
Can you post a screenshot or send the hand and summaries to support@holdemmanager.net