Hem Ev

    • KillerFishes
      KillerFishes
      Bronze
      Joined: 11.07.2010 Posts: 2,313
      Hey guys ... I've read all posts here about EV in HEM... However, I don't think it's somehow reliable ..

      It still takes only ICM hand vs hand .. not hand vs range, therefore you can have -10000000000EV$ for 100 moves but still the moves could be well played.

      For instance:
      - I will demonstrate it on very easy one.

      You're calling in situation BvB, you are pretty sure that SB will push ATC because you've seen him doing so before and have note.


      I won't count riskpremium / chipodds ... Just think about it in this way:

      You think that he has some pushing range, you call him with +EV hand vs that range, but he showes you 3× AA, 5× AK×, 10× 88+ etc. Does it mean, it was a bad call? No, it doesn't! It just means you are unlucky and you run into the best hands of his range.

      What will HEM's conclusion about this play? Your EV line will go straight down.

      This reasoning means for me that EV in HEM says nothing. It needs really a HUGE sample to have balanced that range you are counting with. And I mean really huge sample 1k hands isn't just enough.


      I recommend not making conclusions based on HEM ev, rather work with SNG wiz :)

      I'd like to hear any opinions ... Maybe I'm wrong, who knows :)

      Thx cya,
      KillerFiishes
  • 6 replies
    • goldchess
      goldchess
      Bronze
      Joined: 17.02.2010 Posts: 641
      groundbreaking news
    • pzhon
      pzhon
      Bronze
      Joined: 17.06.2010 Posts: 1,151
      The luck adjustment in HEM (and other programs) correctly adjusts for some of the luck. It does not adjust for all forms of luck such as getting coolered. It is still an improvement which reduces the amount of data you need to get an accurate estimate of your ROI, and it does not introduce significant biases.

      If you play turbo 9-player SNGs, the luck adjustment might reduce your variance by 55%. That means instead of having to play 1000 tournaments to get a 95% confidence interval of +- 10% ROI, you get that level of accuracy after only about 450 tournaments.
    • pzhon
      pzhon
      Bronze
      Joined: 17.06.2010 Posts: 1,151
      By the way, it's a great idea to use SNG Wizard to study particular decisions. If you adjust the ranges well, SNG Wizard is very good at telling you whether your call was correct, and by how much. However, if you feed the wrong ranges into SNG Wizard, such as by assuming that your opponents are pushing ATC when they are actually pushing a 40% range, then you can lose money while you make the plays SNG Wizard says are right according to your assumptions. When you assume that your opponent has a range which differs from his actual range, you are introducing a bias.

      HEM's all-in luck adjustment is only supposed to have a negligible bias from card removal effects, that the deck is not uniformly random after the players outside the blinds fold. It may be incorrectly implemented (and might introduce a bias) in some multiway pots or in multitable SNGs, or perhaps some STTs with unusual prize structures.
    • KillerFishes
      KillerFishes
      Bronze
      Joined: 11.07.2010 Posts: 2,313
      Well ... I will just write down an example and please be so kind and tell me what's wrong. (I have plenty of these hands so I will just take a random one)


      Poker Stars $6.45+$0.55 No Limit Hold'em Tournament - t100/t200 Blinds - 6 players - View hand 1377644
      DeucesCracked Poker Videos Hand History Converter

      Hero (CO): t1615 8.07 BBs
      BTN: t4515 22.57 BBs
      SB: t2095 10.47 BBs
      BB: t2640 13.20 BBs
      UTG: t1605 8.03 BBs
      MP: t1030 5.15 BBs

      Pre Flop: (t300) Hero is CO with Q :club: A :spade:
      2 folds, Hero raises to t1615 all in, 1 fold, SB raises to t2095 all in, 1 fold

      Flop: (t3430) K :spade: K :diamond: T :club: (2 players - 2 are all in)

      Turn: (t3430) 4 :club: (2 players - 2 are all in)

      River: (t3430) J :diamond: (2 players - 2 are all in)



      The HEM's EV diff is:

      -$7,79

      How is it possible? I don't even have to open SNG wiz to know that this push is HIGHLY profitable... So then what's wrong?
    • akrammon
      akrammon
      Bronze
      Joined: 17.05.2009 Posts: 3,142
      You are correct, HEM's EV-line is not punctual.

      But it eliminates a certain percentage of the luck - if you don't even calculate with ICM, but only chipEV, then you will get even worse results.
      HEM's EV eliminates some of the luck. There is no point in looking at the EV-line on a 20, sometimes not even on a 100 SnG sample, but it get6s closer to your actual ROI faster than the "green line", your winnings. So if say you play 1000 SnGs and you are breakeven but your EV line is in +3% ROI then it is more likely that your ROI is positive even though there might be many flaws regarding the EV-line.

      Hope that made sense.

      Regards,
      akrammon
    • pzhon
      pzhon
      Bronze
      Joined: 17.06.2010 Posts: 1,151
      This is not a failure of HEM. This is just something you find counterintuitive.

      Suppose you get all-in with 50% equity so that winning is worth $10 and losing is worth $0. How lucky are you if you win? On that hand, you were lucky by $5. If you lose, you were unlucky by $5 worth. This has nothing do do with whether your push was correct, or whether you were unlucky to get called by that hand.

      There are many sources of luck in the game. Did you get dealt AA? Did someone have a hand good enough to give your AA action? Did your AA hold up after the money went in? HEM correctly estimates the third type of luck. It doesn't have to estimate the first or second type of luck for it to be helpful to remove the third type.