HM1/HM2 Icm '$EV Diff' wrongly imported?

[kampiuceris]

Hello,

Recently I noticed that HM calculates ICM ev correctly, but imports it differently almost every hand. I looked at HM1 ICM calculations (HM2 also imports the same numbers) and took 3 examples to show (All hands are Fifty50 (icm = 60 10 10 10 10). I did Icm calculations myself and can verify they are correct in the table (when you click 'ICM' button in HM1 hand replayer). Why imported numbers are slightly different?

-----------------------------------------------------------------------------------------------

PokerStars Hand #92153312207: Tournament #671608783, $57.25+$2.75 USD Hold'em No Limit - Level IX (100/200) - 2013/01/11 6:01:09 EET [2013/01/10 23:01:09 ET]
Table '671608783 1' 10-max Seat #5 is the button
Seat 2: kampiuceris (5525 in chips) 
Seat 5: lafouine460 (2115 in chips) 
Seat 6: AndreasEM (900 in chips) 
Seat 7: cuongpoker89 (1955 in chips) 
Seat 8: crispoker10 (1925 in chips) 
Seat 10: craigbor18 (2580 in chips) 
kampiuceris: posts the ante 25
lafouine460: posts the ante 25
AndreasEM: posts the ante 25
cuongpoker89: posts the ante 25
crispoker10: posts the ante 25
craigbor18: posts the ante 25
AndreasEM: posts small blind 100
cuongpoker89: posts big blind 200
*** HOLE CARDS ***
Dealt to kampiuceris [Qc As]
crispoker10: folds 
craigbor18: folds 
kampiuceris: raises 200 to 400
lafouine460: folds 
AndreasEM: folds 
cuongpoker89: raises 1530 to 1930 and is all-in
kampiuceris: calls 1530
*** FLOP *** [Qh 3s 6d]
*** TURN *** [Qh 3s 6d] [9d]
*** RIVER *** [Qh 3s 6d 9d] [9s]
*** SHOW DOWN ***
cuongpoker89: shows [7s 7c] (two pair, Nines and Sevens)
kampiuceris: shows [Qc As] (two pair, Queens and Nines)
kampiuceris collected 4110 from pot
cuongpoker89 finished the tournament in 6th place
kampiuceris wins the tournament and receives $203.81 - congratulations!
craigbor18 finished the tournament in 2nd place and received $106.01.
lafouine460 finished the tournament in 3rd place and received $97.13.
crispoker10 finished the tournament in 4th place and received $93.51.
AndreasEM finished the tournament in 5th place and received $72.04.
*** SUMMARY ***
Total pot 4110 | Rake 0 
Board [Qh 3s 6d 9d 9s]
Seat 2: kampiuceris showed [Qc As] and won (4110) with two pair, Queens and Nines
Seat 5: lafouine460 (button) folded before Flop (didn't bet)
Seat 6: AndreasEM (small blind) folded before Flop
Seat 7: cuongpoker89 (big blind) showed [7s 7c] and lost with two pair, Nines and Sevens
Seat 8: crispoker10 folded before Flop (didn't bet)
Seat 10: craigbor18 folded before Flop (didn't bet)

calculates -$43.91, imports -$44.24

-----------------------------------------------------------------------------------------------

PokerStars Hand #92154482188: Tournament #671618725, $57.25+$2.75 USD Hold'em No Limit - Level VIII (75/150) - 2013/01/11 7:02:47 EET [2013/01/11 0:02:47 ET]
Table '671618725 1' 10-max Seat #10 is the button
Seat 2: ripzone22 (870 in chips) 
Seat 4: crispoker10 (2665 in chips) 
Seat 5: Boxrod (2820 in chips) 
Seat 6: poormenVN (1610 in chips) 
Seat 7: GrossUfa (1915 in chips) 
Seat 9: TrapoLeproso (1270 in chips) 
Seat 10: kampiuceris (3850 in chips) 
ripzone22: posts the ante 20
crispoker10: posts the ante 20
Boxrod: posts the ante 20
poormenVN: posts the ante 20
GrossUfa: posts the ante 20
TrapoLeproso: posts the ante 20
kampiuceris: posts the ante 20
ripzone22: posts small blind 75
crispoker10: posts big blind 150
*** HOLE CARDS ***
Dealt to kampiuceris [Qc Kh]
Boxrod: folds 
poormenVN: folds 
GrossUfa: folds 
TrapoLeproso: folds 
kampiuceris: raises 150 to 300
ripzone22: raises 450 to 750
crispoker10: folds 
kampiuceris: raises 450 to 1200
ripzone22: calls 100 and is all-in
Uncalled bet (350) returned to kampiuceris
*** FLOP *** [4d As 9d]
*** TURN *** [4d As 9d] [7s]
*** RIVER *** [4d As 9d 7s] [4c]
*** SHOW DOWN ***
ripzone22: shows [Kc Js] (a pair of Fours)
kampiuceris: shows [Qc Kh] (a pair of Fours - Ace+King+Queen kicker)
kampiuceris collected 1990 from pot
ripzone22 finished the tournament in 7th place
*** SUMMARY ***
Total pot 1990 | Rake 0 
Board [4d As 9d 7s 4c]
Seat 2: ripzone22 (small blind) showed [Kc Js] and lost with a pair of Fours
Seat 4: crispoker10 (big blind) folded before Flop
Seat 5: Boxrod folded before Flop (didn't bet)
Seat 6: poormenVN folded before Flop (didn't bet)
Seat 7: GrossUfa folded before Flop (didn't bet)
Seat 9: TrapoLeproso folded before Flop (didn't bet)
Seat 10: kampiuceris (button) showed [Qc Kh] and won (1990) with a pair of Fours

calculates -$12.08, imports -$11.92

-----------------------------------------------------------------------------------------------

PokerStars Hand #92154439333: Tournament #671618725, $57.25+$2.75 USD Hold'em No Limit - Level VII (60/120) - 2013/01/11 7:00:15 EET [2013/01/11 0:00:15 ET]
Table '671618725 1' 10-max Seat #5 is the button
Seat 2: ripzone22 (1050 in chips) 
Seat 3: Stuggots222 (405 in chips) 
Seat 4: crispoker10 (3025 in chips) 
Seat 5: Boxrod (2130 in chips) 
Seat 6: poormenVN (1730 in chips) 
Seat 7: GrossUfa (1510 in chips) 
Seat 9: TrapoLeproso (1510 in chips) 
Seat 10: kampiuceris (3640 in chips) 
ripzone22: posts the ante 15
Stuggots222: posts the ante 15
crispoker10: posts the ante 15
Boxrod: posts the ante 15
poormenVN: posts the ante 15
GrossUfa: posts the ante 15
TrapoLeproso: posts the ante 15
kampiuceris: posts the ante 15
poormenVN: posts small blind 60
GrossUfa: posts big blind 120
*** HOLE CARDS ***
Dealt to kampiuceris [9s Ah]
TrapoLeproso: folds 
kampiuceris: raises 120 to 240
ripzone22: folds 
Stuggots222: raises 120 to 360
crispoker10: folds 
Boxrod: folds 
poormenVN: folds 
GrossUfa: folds 
kampiuceris: raises 120 to 480
Stuggots222: calls 30 and is all-in
Uncalled bet (90) returned to kampiuceris
*** FLOP *** [2s 3s Ac]
*** TURN *** [2s 3s Ac] [2d]
*** RIVER *** [2s 3s Ac 2d] [4d]
*** SHOW DOWN ***
kampiuceris: shows [9s Ah] (two pair, Aces and Deuces)
Stuggots222: shows [8h 9h] (a pair of Deuces)
kampiuceris collected 1080 from pot
Stuggots222 finished the tournament in 8th place
*** SUMMARY ***
Total pot 1080 | Rake 0 
Board [2s 3s Ac 2d 4d]
Seat 2: ripzone22 folded before Flop (didn't bet)
Seat 3: Stuggots222 showed [8h 9h] and lost with a pair of Deuces
Seat 4: crispoker10 folded before Flop (didn't bet)
Seat 5: Boxrod (button) folded before Flop (didn't bet)
Seat 6: poormenVN (small blind) folded before Flop
Seat 7: GrossUfa (big blind) folded before Flop
Seat 9: TrapoLeproso folded before Flop (didn't bet)
Seat 10: kampiuceris showed [9s Ah] and won (1080) with two pair, Aces and Deuces

calculates -$8.53, imports -$8.80

[Patvs]

This is a known bug, bugticket
HMONE-127
HM-5295

It should be fixed (for HM2) in the next update.

[kampiuceris]

This is a known bug, bugticket
HMONE-127
HM-5295

It should be fixed (for HM2) in the next update.

Still not fixed. Any progress?

[Patvs]

I've added your email address to the bug ticket, so you'll be automatically notified when it is fixed

[Patvs]

Hey kampiuceris


I imported the handhistory file you sent me via private message in both HM1 and HM2:

24a.JPG

24.JPG


See screenshots. All EV Diff numbers are the same. (they're actually more accurate in HM2 due to the new lookup table) A good example of this is the A9o hand which shows -$5.59 in HM1 vs -$4.08 in HM2. A relatively big difference, but the hand is A9o vs A9s.


The key hand is the 24s where you fold which shows +$39.33 in HM1 vs only +$1.94 in HM2.
This is the hand responsible for the completely different final EV Diff value of the game.

PokerStars Hand #98939930077: Tournament #736337223, $95.86+$4.14 USD Hold'em No Limit - Level XII (200/400) - 2013/05/22 15:46:53 EET [2013/05/22 8:46:53 ET]
Table '736337223 1' 10-max Seat #6 is the button
Seat 2: metzi69 (4278 in chips) 
Seat 3: samy212 (1638 in chips) 
Seat 4: BigMadDrongo (2165 in chips) 
Seat 6: craigbor18 (3513 in chips) 
Seat 8: hero (456 in chips) 
Seat 10: Quadchrazs (2950 in chips) 
metzi69: posts the ante 50
samy212: posts the ante 50
BigMadDrongo: posts the ante 50
craigbor18: posts the ante 50
hero: posts the ante 50
Quadchrazs: posts the ante 50
hero: posts small blind 200
Quadchrazs: posts big blind 400
*** HOLE CARDS ***
Dealt to hero [2c 4c]
metzi69: folds 
samy212: folds 
BigMadDrongo: raises 1715 to 2115 and is all-in
craigbor18: raises 1348 to 3463 and is all-in
hero: folds 
Quadchrazs: folds 
Uncalled bet (1348) returned to craigbor18
*** FLOP *** [Kc Jh 7c]
*** TURN *** [Kc Jh 7c] [2d]
*** RIVER *** [Kc Jh 7c 2d] [3c]
*** SHOW DOWN ***
BigMadDrongo: shows [Ad 9d] (high card Ace)
craigbor18: shows [9s Ah] (high card Ace)
BigMadDrongo collected 2565 from pot
craigbor18 collected 2565 from pot
*** SUMMARY ***
Total pot 5130 | Rake 0 
Board [Kc Jh 7c 2d 3c]
Seat 2: metzi69 folded before Flop (didn't bet)
Seat 3: samy212 folded before Flop (didn't bet)
Seat 4: BigMadDrongo showed [Ad 9d] and won (2565) with high card Ace
Seat 6: craigbor18 (button) showed [9s Ah] and won (2565) with high card Ace
Seat 8: hero (small blind) folded before Flop
Seat 10: Quadchrazs (big blind) folded before Flop

It's an interesting hand, because
This actually is a known bug, HM-5469. This is a splitpot bug.
But in this hand you fold, and aren't even involved in the splitpot.
And this bug -likely- affects some of your other games as well.

I'll update bugticket HM-5469 and add you to the ticket.

[kampiuceris]

Hey Patvs,

I've developed my own algorithm for calculating EV (I graduated software engineering). I don't say my algorithm is ultimate and undeniable, but it's unlikely since it passed all tests which I manually calculated on paper and it's very very unlikely I made a mistake on paper. I don't know if you are a programmer and if you have anything to do with current HM EV algorithm. But I have few things to notice (will give some calculations so you could pass them to developers if you aren't one).

First of all I don't know which version of hm1 you are using in the picture you posted above, but mine shows different numbers than yours.

I looked deeper in this case and made a picture to compare:




You mentioned that 4c2c hand (marked blue in my picture) and I didn't understand what u wanted to say: either hm1 or hm2 was incorrect? Regardless of that mine hm1 and hm2 both calculated correctly. These are the calculations for 4c2c hand:

------------ Possible hand endings: ------------
Ad9d = 9sAh      probability 90.66912%
Player Name         Stack     Icm     
metzi69             4228      0.2398884314     
samy212             1588      0.1449764733   
BigMadDrongo        2565      0.1822260738  
craigbor18          3913      0.2291656811   
kampiuceris         206       0.0238597140   
Quadchrazs          2500      0.1798836265

Ad9d > 9sAh      probability 7.15761%   
Player Name         Stack     Icm      
metzi69             4228      0.2401207402  
samy212             1588      0.1460890153   
BigMadDrongo        5130      0.2705215063     
craigbor18          1348      0.1358803913
kampiuceris         206       0.0268889046      
Quadchrazs          2500      0.1804994422      

Ad9d < 9sAh      probability 2.17327%   
Player Name         Stack     Icm     
metzi69             4228      0.2409333333     
samy212             1588      0.1529333333       
BigMadDrongo        0         0.0000000000      
craigbor18          6478      0.3159333333    
kampiuceris         206       0.1068666667   
Quadchrazs          2500      0.1833333333     

------------ EV of the hand: ------------
Player Name         Icm          ($)
metzi69             0.2399277677 ($229.99)
samy212             0.1452290288 ($139.22)
BigMadDrongo        0.1845856486 ($176.94)
craigbor18          0.2243743821 ($215.09)
kampiuceris         0.0258804973 ($24.81)
Quadchrazs          0.1800026757 ($172.55)

------------ Result of the hand: ------------
Player Name         Stack     Icm     
metzi69             4228      0.2398884314     
samy212             1588      0.1449764733   
BigMadDrongo        2565      0.1822260738  
craigbor18          3913      0.2291656811   
kampiuceris         206       0.0238597140   
Quadchrazs          2500      0.1798836265

------------ EV calculation: ------------
"kampiuceris" EV = (0.0238597140 * 0.9066912) + (0.0268889046 * 0.0715761) + (0.1068666667 * 0.0217327) = 0.0258804973

EV Diff = 0.0258804973 - 0.0238597140 = 0.0020207833 ($1.93712287138)

You can see the EV Diff calculation on paper (which I am sure is correct) is $1.93712287138 ~ $1.94 and in my case both hm1 and hm2 calculated it correctly. In your case hm1 is incorrect, hm2 is correct.




I assume hm2 (after this ev fix update) probably had most accurate algorithm HM developers had so far. Now what caught my attention is another hand AhJh (marked red in my picture) where my hm2 calculations were incorrect (hm1 were good). What's more surprising is that your hm1 calculations differs from mine (looking into your picture $5.88) I don't know if you have updated hm1 version or what, but the one I have (1.12.10b) shows -$28.27 which is actually a correct calculation. Here it is:

------------ Possible hand endings: ------------
Tc9d < AhJh      probability 65.21108% 
Player Name         Stack     Icm
metzi69             3778      0.2215677168
samy212             1538      0.1278082642
BigMadDrongo        2515      0.1735627389
craigbor18          3995      0.2293375098
kampiuceris         924       0.0853377732
Quadchrazs          2250      0.1623859970

Tc9d = AhJh      probability 0.39619%
Player Name         Stack     Icm
metzi69             3778      0.2232963795
samy212             1538      0.1359091203
BigMadDrongo        2515      0.1774210643
craigbor18          4457      0.2468162215
kampiuceris         462       0.0495124946
Quadchrazs          2250      0.1670447198

Tc9d > AhJh      probability 34.39272%   
Player Name         Stack     Icm
metzi69             3778      0.2259333333   
samy212             1538      0.1512666667 
BigMadDrongo        2515      0.1838333333       
craigbor18          4919      0.2639666667 
kampiuceris         0         0.0000000000
Quadchrazs          2250      0.1750000000

------------ EV of the hand: ------------
Player Name         Icm Ending   ($)
metzi69             0.2230760201 ($213.84)
samy212             0.1359083429 ($130.28)
BigMadDrongo        0.1771103625 ($169.78)
craigbor18          0.2413166695 ($231.33)
kampiuceris         0.0558458512 ($53.53)
Quadchrazs          0.1667427538 ($159.84)

------------ Result of the hand: ------------
Player Name         Stack     Icm
metzi69             3778      0.2215677168
samy212             1538      0.1278082642
BigMadDrongo        2515      0.1735627389
craigbor18          3995      0.2293375098
kampiuceris         924       0.0853377732
Quadchrazs          2250      0.1623859970

------------ EV calculation: ------------
"kampiuceris" EV = (0.0853377732 * 0.6521108) + (0.0495124946 * 0.0039619) + (0.0000000000 * 0.3439272) = 0.0558458512

EV Diff = 0.0558458512 - 0.0853377732 = -0.029491922 (-$28.2709564292)

EV Diff on paper is -$28.2709564292 ~= -$28.27 and my hm1 gives the same number (your hm1 gives also random number +$5.88). Where hm2 gives +$6.07 (yours hm2 too).

So if you are not HM EV algorithm developer, please pass these "on paper" calculations with the results I mentioned in this post to the developers so they can take a deeper look in the algorithm, because there is clearly a mistake somewhere.

Thanks

[Patvs]

I imported the hands into HM1.13.01 internal, and HM2 build 7758.


I'll add your last post to the main EV bugticket.

[kampiuceris]

I imported the hands into HM1.13.01 internal, and HM2 build 7758.


I'll add your last post to the main EV bugticket.

Should I take a look in 2.0.0.7778 if this is fixed? And I actually don't know where to watch these bugtickets, so I could have a clue of what's going on.

[udbrky]

We email you when they are resolved.

[Patvs]

The bugtickets (and feature request tickets) is our internal system to write up bugs.
A ticket will be verified first--> then assigned to a programmer--> once fixed the users attached to the bugticket are notified.

HoldemManager users are not able to access this ticket system.
From what I can tell there are no new EV fixes in build 7778. I'd just wait until some of the remaining EV bug tickets are also fixed.