Videothread "The Nemesis"

  • 26 replies
    • GingerKid
      GingerKid
      Black
      Joined: 05.08.2007 Posts: 5,536
      Hi,

      In video 2, min 24, you say that we defend 1-alpha, e.g. if villian bets 2/3 pots, we should defend 67% of our range. Then you said, because BU bluffs have equity, then we should defend wider, around 70%. How did you come to this number (70%)?
      If BU bluffs on so dry boards are mostly GS, they have 15% equity vs BB turn defend range (hand like JT), and if BB calls T9 then GS has around 20% equity.

      To find required FE for BUs GS bluffs, we can approximate it if BU bets 2/3 pot on turn and is all in. I will put 100bb pot on turn, and bet size 66bb, to make calculation easier.

      EV_bet = fe * pot_turn + (1-fe) * (equity * pot_river - bet_size)
      EV_bet = fe * 100bb + (1-fe) * (0.15 * 232bb - 66bb)
      EV_bet = fe * 100bb + (1-fe) * (-31bb) = 0
      fe * 131bb = 31bb
      fe = 0.236


      So if BU needs 23.6% FE when going all in on turn with 2/3 pot bet size, then he obviously needs even less FE if he is not all in on the turn and he can make another bet on the river. So I tink 70% defending is not enough, it should be around 80% if we want to make his weakest bluffs EV=0.
      If board would have fd, then we would need to defend even more.

      So my question is, are we trying to make his worst bluffs in turn bet range EV=0, or we try to make his worst possible bluff in turn whole range EV=0 (vs hands like 22, T2s it is enough to defend 70%) ?

      Sorry if going a bit off topic, but I got confused when you said we need to defend 70% instead of 67%.
    • Tackleberry
      Tackleberry
      Coach
      Coach
      Joined: 05.05.2006 Posts: 1,668
      Hi GingerKid, that´s a very interesting and intelligent question.

      I have to admit that I didn´t put too much work yet into calculating / estimating "optimal" defense frequencies against "semi"-draws. Reason being that there are numerous factors to consider.

      One thing that immediately strikes me when reading your post (which is great!!) is one (minor) mistake:

      [...] if we want to make his weakest bluffs EV=0. [...]
      We don´t try to make his bluffs EV=0, what we try to accomplish is that his bluffs don´t gain any additional EV!

      Like if he has a gutshot to the nuts, you can say that a check-behind has a positive EV. How much - is hard to quantify, but it´s definitely +EV. So, if he´s bluffing with a gutshot (that had a positive EV by xb), it would be a mistake to make his bluffs 0EV, because then he had no incentive to bluff at all, but he would simply check behind and realize his equity. OK?

      That means, you´re defending at a frequency that keeps his EV when bluffing similar to the EV of checking back. But the unsolvable question obviously is what that EV would be. You already made some smart assumptions about this EV, but you went with the wrong premise, namely trying to get him to 0EV.

      That said, the 70% purely were my gutfeeling, it might well be that the "correct" answer is somewhere between the two of us, as 80% definitely seems too high (for the reasons given above).

      Is that understandable?
    • GingerKid
      GingerKid
      Black
      Joined: 05.08.2007 Posts: 5,536
      min 36.

      You said, if we raise all our strong hands on turn (like 2 pair+), then our call range is too weak and villian can exploit us with shoving any two on river. Do you assume here, that
      BU knows exactly what range BB has, and that BB is completely blind and will not notice that BU is shoving any 2 on river? Imo, even if BB raises any 2 pair+ on turn, he can still snap call river with any Qx, since it greatly blocks villian value shove range, especially if BB knows that BU is not value heavy.

      I think, the reason for not raising always 2 pair+ on turn shouldnt be because villian can shove any 2 on river due to BB cappt range, but because BU can shove for value more hands, like KQ+, meaning that he can use many combos as bluff in balanced range, which means his cbet % on river would be very high and our bluff catch range on turn will have -EV call.
    • Tackleberry
      Tackleberry
      Coach
      Coach
      Joined: 05.05.2006 Posts: 1,668
      Originally posted by GingerKid
      I think, the reason for not raising always 2 pair+ on turn shouldnt be because villian can shove any 2 on river due to BB cappt range, but because BU can shove for value more hands, like KQ+, meaning that he can use many combos as bluff in balanced range, which means his cbet % on river would be very high and our bluff catch range on turn will have -EV call.
      Totally agree, and that was essentially what I meant. If our range is severely capped so we´re forced to call "superlight", Villain can extend his valuerange significantly, ultimately resulting in him almost betting his entire range (consisting of a thin, but large valuepart and a corresponding large bluffpart).
    • GingerKid
      GingerKid
      Black
      Joined: 05.08.2007 Posts: 5,536
      Originally posted by Tackleberry
      Hi GingerKid, that´s a very interesting and intelligent question.

      I have to admit that I didn´t put too much work yet into calculating / estimating "optimal" defense frequencies against "semi"-draws. Reason being that there are numerous factors to consider.

      One thing that immediately strikes me when reading your post (which is great!!) is one (minor) mistake:

      [...] if we want to make his weakest bluffs EV=0. [...]
      We don´t try to make his bluffs EV=0, what we try to accomplish is that his bluffs don´t gain any additional EV!

      Like if he has a gutshot to the nuts, you can say that a check-behind has a positive EV. How much - is hard to quantify, but it´s definitely +EV. So, if he´s bluffing with a gutshot (that had a positive EV by xb), it would be a mistake to make his bluffs 0EV, because then he had no incentive to bluff at all, but he would simply check behind and realize his equity. OK?

      That means, you´re defending at a frequency that keeps his EV when bluffing similar to the EV of checking back. But the unsolvable question obviously is what that EV would be. You already made some smart assumptions about this EV, but you went with the wrong premise, namely trying to get him to 0EV.

      That said, the 70% purely were my gutfeeling, it might well be that the "correct" answer is somewhere between the two of us, as 80% definitely seems too high (for the reasons given above).

      Is that understandable?
      hi, it is great explaination, makes sense. I have read book from Janda "applications of no limit holdem", and couldnt conclude this fact from the book, even though I was searching for answer. I was also asking many coatches for this specific question and nobody explained it so that I could understand it (or maybe they also dont understand it really).
    • GingerKid
      GingerKid
      Black
      Joined: 05.08.2007 Posts: 5,536
      Originally posted by Tackleberry
      Originally posted by GingerKid
      I think, the reason for not raising always 2 pair+ on turn shouldnt be because villian can shove any 2 on river due to BB cappt range, but because BU can shove for value more hands, like KQ+, meaning that he can use many combos as bluff in balanced range, which means his cbet % on river would be very high and our bluff catch range on turn will have -EV call.
      Totally agree, and that was essentially what I meant. If our range is severely capped so we´re forced to call "superlight", Villain can extend his valuerange significantly, ultimately resulting in him almost betting his entire range (consisting of a thin, but large valuepart and a corresponding large bluffpart).

      Again I will ask question not related to the video, but related to this post.

      I analysed my ranges, where I open raise CO and villian calls BU all fold.

      Flop is 8 :spade: 4 :spade: 2 :heart:


      I open raise 28% CO. Not sure how to construct flop range, I have 2 versions:

      1) cbet around 70% range, where 30% check range is mostly overcards with no bdfd. I cbet all 8x, and lower pairs, for one street thin value / protection and plan to cc it on good turn, if overcard comes then I can fold them because my blufffs will improve.

      2) I cbet 30% range, 99-KK, and draws, and check 70%. I c/r all sets and AA without bdfd (to lower his bet EV because on low boards he has big positional advantage and so many draws), and cc all 8x , and lower pairs, AT+.


      Imo pros for 1) is that I cbet weaker pairs for protection, and I have equity advantage so I shoul mostly bet. cons is that once I check, I fold 100% time (is it a problem since villian can exploit it only rarely when I check?). And this is the spot where our range is very weak and villian can make huge bets with big part of his range, so we shouldnt try to defend it but fold 100%? If I bet 70%, my cbet range becomes vulnerable to raises (is it a problem?).

      pros for 2) is that I have both ranges balanced, but villian is able to play IP so often and realize his equity.

      Since i liked your last explanation about EV=0, I think you can explained me also this one, hope it is not a problem.
    • Tackleberry
      Tackleberry
      Coach
      Coach
      Joined: 05.05.2006 Posts: 1,668
      GingerKid, would it be okay for you if I took your question as the basis for a complex range analysis either in 3rd or 4th part of my Nemesis Video Series?
    • GingerKid
      GingerKid
      Black
      Joined: 05.08.2007 Posts: 5,536
      Originally posted by Tackleberry
      GingerKid, would it be okay for you if I took your question as the basis for a complex range analysis either in 3rd or 4th part of my Nemesis Video Series?
      Sure, great!
    • Tackleberry
      Tackleberry
      Coach
      Coach
      Joined: 05.05.2006 Posts: 1,668
      Originally posted by GingerKid
      hi, it is great explaination, makes sense. I have read book from Janda "applications of no limit holdem", and couldnt conclude this fact from the book, even though I was searching for answer. I was also asking many coatches for this specific question and nobody explained it so that I could understand it (or maybe they also dont understand it really).
      Thanks for the nice words by the way! :)
    • GingerKid
      GingerKid
      Black
      Joined: 05.08.2007 Posts: 5,536
      Again me :)
      I am wondering about nemesis approach, is it something invented by you / your friends, or is it well known concept described in some book? I would like to read it if there is something.

      What I didnt like in your video is that you just briefly explained idea of the approach and then showed one detailed and long example. What I mean is, there is no theoretical approach in your video, just purely practical.
      Since it is called "approach", then I would expect to see some algorithm (steps) how to optimize our range. So I would expect something like:
      -------------------------------------
      1) In step 1, do that
      2) step 2 do that
      ...
      n) step n do that

      iteratate until specific condidions are satisfied (e.g. we cbet balanced, c/f < 50%, defend vs raise 40% etc)
      --------------------------------------


      For example, If my current range on 2s 4s 8c board CO vs BU would have bluff heavy cbet range, and I fold 80% when I check, then I would start by:

      1) removing weakest bluffs from my cbet range and puting into check range (this would improve my cbet range, but would make check range even weaker)
      2) in order to check/fold e.g. only 50% I would have to move some strong hands from cbet range in check/raise range and some pairs which will c/c multi streets. (this way I improve check range, but now I am again bluff heavy in cbet range)

      Now my algorithm (approach) would have 2 base steps, and I need to iterate now, go to step 1), and already after 1st iteration leaks are improved. I would stop iterating until I cbet balanced and I check/fold 50%.
      This is off course simplification, I would have to care about defending enough vs raises, and to defend my check call range on mutli streets, but it is going in right direction.
    • Tackleberry
      Tackleberry
      Coach
      Coach
      Joined: 05.05.2006 Posts: 1,668
      Originally posted by GingerKid
      Again me :)
      Hi. :)

      Originally posted by GingerKid
      I am wondering about nemesis approach, is it something invented by you / your friends, or is it well known concept described in some book? I would like to read it if there is something.
      Eventually it´s just a name for adjusting your range against an opponent, which we expect to play / adjust perfectly in any situation (afaik, the term "nemesis" is commonly used in poker for describing this situation).

      Originally posted by GingerKid
      What I didnt like in your video is that you just briefly explained idea of the approach and then showed one detailed and long example.
      Did you watch the first part of the video? There I explained what I understand by the "nemesis approach" ... but from what follows I´m afraid your expectation was different.

      Originally posted by GingerKid
      What I mean is, there is no theoretical approach in your video, just purely practical.
      Yeah, that´s right. :)

      Originally posted by GingerKid
      Since it is called "approach", then I would expect to see some algorithm (steps) how to optimize our range.
      If such an "approach" really existed (and it was good), the game would´ve solved, right? ;) But seriously, there is no method that can be learned step by step. The whole video series is about showing you how to train your range construction skills, namely based on the idea that you adjust against an unknown opponent, from which you know nothing more than that he plays "optimally in any given situation" (= nemesis).

      The term "approach" was meant to describe the way of not judging our hand "in the vacuum", considering the situational circumstances, but judging our hand in the context of our entire range against an optimally playing opponent.

      I´m afraid I can´t deliver you the goose that lays golden eggs. :( I can just show you a starting point from where to begin your journey.

      Does that make sense for you?
    • GingerKid
      GingerKid
      Black
      Joined: 05.08.2007 Posts: 5,536

      If such an "approach" really existed (and it was good), the game would´ve solved, right? wink But seriously, there is no method that can be learned step by step.

      I didnt mean, to have approach step by step, where in the end you get perfect GTO ranges. You showed one example, how you improve your range. So you didnt get to the GTO range at the end, but you improved it. So if improving the range in the way you suggested in the video is possible and you try to teach other people to do it, then there must be a way to define the approach using finite number of steps.
      If you cant define it using finite number of steps as a general solution to improve range, then you probably dont understand what you are doing. Dont get me wrong, I am not saying you dont understand it, I am just saying if you know how to improve any "bad" constructed range (or most of them), then you must be following some algorithm (otherwise you would do it randomly and wouldnt improve always the range).
      Understand what i mean? I used wrong word "optimize", I should say improve because it is not possible to optimize it using current knowledge.


      The term "approach" was meant to describe the way of not judging our hand "in the vacuum", considering the situational circumstances, but judging our hand in the context of our entire range against an optimally playing opponent.
      I watched both of your videos, but I was sick at that time (fever) so maybe I missed something, couldnt watch it fully concentrated. I think you explained it nicely in the first viedo what are the pros and cons of playing the hand in the vacuum and in the context of our entire range. So as I unederstand now, the video series is trying to teach people that they should start playing hands in the context of entire range against an optimally playing opponent. You showed one example in CREV how to do it, but that example is more like motivation, and you dont expect people to learn from that example how to play in the context of entire range? If that is the case, then it makes sense. I missunderstood that you are trying to teach allmighty approach how to improve our range, to make it more balanced. The reason why I missunderstood it, is because in 2nd video, the font in CREV was too small for me (I watched it on 14'' laptop monitor) so I couldnt see the details, and was just listening to hear idea, and planned to see it again on big monitor. So basically from first video I understood that we should play a hand in context of entire range vs opponent who will find all our leak and exploit them, so we should construct ranges in the way that he can minimally possible exploit us.
      And in second video I saw how you are clicking on buttons in CREV 1 hour, and correcting multiple times same ranges, but couldnt see what exactly you are doing with ranges, and I concluded that you iteratively improve ranges until you are satisfied with all conditions. Sorry for missunderstanding, and commenting without really watching it, but idea was very interesting to me and I wanted to start discoussion.
      In the end the idea is not anymore so interesting to me, because you are not bringing "allmighty approach" :) but it means that you presented video in a very interesting and professional way, that it made me pay so much attention and think about this topic. Most coacthing vidoes I watch and say, I didnt learn much new, it was boring and forget about it fast.
    • GingerKid
      GingerKid
      Black
      Joined: 05.08.2007 Posts: 5,536
      Since I started thinking about "allmighty algorithm" motivate by your video, I started making step by step approach how to construct any given range balanced (off course not perfectly balanced, but the goal is to be decent).

      Assumptions:
      1) current ranges of both opponents are known.
      2) hero open raises, villian calls IP all other fold (HU)
      3) villian is very good player, who is going to spot all leaks and exploit them

      Goal:
      Construct as much as possible balanced range, try to minimize our leaks using step by step approach for any given input ranges of hero and villian using above assumptions


      I will start explaining approach using example, and after example I will formulate step by step approach to make it general.

      Example is, preflop hero open raises from CO and BU calls, all other fold.
      Hero opens 28% range (352 combos), BU calls with stanard pokerstrategy range from charts.

      Flop: 8 :spade: 4 :spade: 2 :heart:

      Phase 1: Splitting into categories

      step1:
      Find hands that are able to bet 3 streets for value:
      99+, 22, 44, 88 (45 combos)

      step2:
      Find hands that are able to value bet 2 streets, or to cbet flop and cc turn, or to cc flop and turn:
      8x hands, 55, 66, 77, A2s, A4s (65 combos)

      step3:
      Find hands that are able to cc flop and give up vs 2nd barrel:
      AJ+ (32 combos). (here fd and bdfd combos
      are excluded)

      step4:
      Find hands that are good bluffs for cbet (that retain equity) like backdoor draws and draws with >= 4 outs:
      (125 combos) here I dont want to listt them because it is longer list and it is unimportant


      step5:
      Find hands that are not having SD value able to call or bet even flop, and not having good playability. It is simply the rest of range, which is (82 combos).



      Now when we splitted all ranges, we can assign them to different ranges

      Phase 2: Assigning combos to ranges

      step1:
      Determine maximum possible hands that are able to cbet at least flop and to either cbet turn or c/c turn. It is simply range from phase1, step1 and step2. Put them initially all in cbet range
      (110 combos)

      step2:
      Determine number of bluffs for the value combos from step1, choosing bluffs that retain equity well. For each 3 street value combo (phase1, step 1 range) use 2 bluff combos, and for each combo from phase1, step2 use 1 bluff combo. These are 125 combos of range in phase1 step4 and 30 strongest hands from phase1 step5.
      (155 combos)

      step3:
      Because the hands from phase1 step3 (weak SD hands) and remaining of bluffs cant be in current cbet range (current cbet range is using max value combos and max bluffs in balanced range), they HAVE to be in the check range (if they would be in cbet range we would have to be bluff heavy, or to cbet hands that dont ratain equity well, and to c/f hands that retain equity well which are both bad).
      So, put all 32 combos from phase1 step3 in check range and also all remaining of 52 combos from phase1 step5.
      (87 combos)

      step4:
      Deterimine c/f range. The combos from phase1 step5 that are not added to cbet range HAVE to be in cf range. (hands bad for bluffing and very low SD value) (52 combo)

      step5:
      Determine number of combos that are in check/defend range. Since we want to fold 40-50% of time once we check (assuming 3/4 bet size), I will choose 40% fold. It means we need to defend (52/0.4 - 52) (78 combos)


      step6:
      Determine range from phase2, step5. We need to start moving value hands and draws from cbet range in check/call and check/raise range in order to defend 78 combos on flop after checking. The more drawy the board is, i add more strong combos in check/raise range.
      Here I put all sets in check/raise range to reduce villians big positional advantage. 9 combos set, and we can use for each value raise combo 2 bluffs, so in total 27 raise combos.
      This means that from 91 combos in check/defend range, we should call 91-27= 64 combos.
      As summary, we c/f flop 61 combo, c/r 27 combo, c/c 64 combos.
      Since I moved only 9 combo sets from cbet range to check range, and 18 combos bluff (simply buttom bluffs for cbet range, to keep it simple initially) cbet range still remains balanced.


      step7:
      Determine number of combos that c/c flop and c/f turn unimproved. We want to fold turn around 35% vs 3/4 pot size bet, so we fold on turn 64 * 0.35 = 22 combos and we c/c turn with remaining of
      42 combos

      step8:
      On river vs 3/4 pot size bet, we want to fold 42% range and to call 58% range.
      So we fold 42 * 0.4 = 17 combos and
      we call 25 combos

      step9:
      Move value hands from cbet range, into range from step8. On river we usually want to defend at least top pairs, otherwise our range would be too weak and villian could use overbets. We dont want to defend it with too many strong hands because they would loose too much value by playing passive, but we want to put some strong hands. For few strong hands, add hands that dont need much protection and are blocking villians defend range. I would add in this example AsAx, because it needs least protection and it can call down on any runout.
      The remaining of 22 combos, I would add 22 weakest top pair combos because hands like A8 can value bet (have more value than 78), same is for any top pair.
      Now since we moved 3 combos AsAx from cbet range that can value bet 3 streets, we need to move 6 bluffs to check range so that cbet range remains ballanced. We also need to move 22 combos bluff from cbet range to check range for all 8x combos. In total we will currently remember that we need to move 28 combos bluffs to check range from cbet range.

      step10:
      Now when we have river defend range, we need to determine completely turn defend range. On turn we defend 22 combos 8x and 3 combos AsAx (which we defend also on river), and we need now to determine additional 17 combos that call turn and fold river (unimproved). Since we want to be able to defend on river if flush completes, I would put As9s-AsKs, KsQs, KsJs (8 combos fd) which can also improve to top pair on river and have SD value so they can win even unimproved some % of time.
      So we need 9 additional combos, and I would add 6 combos 77 and 3 combos 66.
      So since we moved 8 combos fd from cbet range to check range, we need to move 20 bluff combos more so that cbet range remains balanced.

      step11:
      Determine range that c/c flop and c/f turn unimproved (22 combos). For that, imo some of AT-AK combos are the best because they have SD value and can check down on later streets, and also we will have some TPTK hands on turn if T-A cards come. Also I would add to that range some GS (maybe also some bdfd) because we want also some floats that have little SD and can improve to nuts, because when villian checks back turn we have combos to bluff on river unimproved. So I would add 67s 56s, 57s (12 combos) and 10 combos from AT-AK (from each 2 combos).

      ... to be continued no time now.
    • Tackleberry
      Tackleberry
      Coach
      Coach
      Joined: 05.05.2006 Posts: 1,668
      Hello GingerKid, awesome work. I was already aware of what you expected to see - but I think your "approach" is not universally applicable, but I´d like to await the "final ranges" you're coming up with and then I´ll analyze in detail.

      At the very least it seems as if I´ve sparked something in you which is valuable in and of itself, right? ;)
    • Shakaflaka
      Shakaflaka
      Bronze
      Joined: 18.01.2010 Posts: 512
      Nice discussion!

      I'm also currently doing the same kind of analysis, ginger. What I find it useful when analysing oop ranges as pfr is to use the mathew janda's formula to check the Ev of opponent's worst hands on the flop. It is a good way to"check" that our frequencies are somewhat decent. For example, if you cbet 40% and give up 80% when you check, the opponent's worst hands will be highly profitable.

      Janda's recomendation is to keep the EV of opponent's bluffs significantly below to money they invested preflop. For example, if you open raise to 3bb, their Ev should be way below that becasue they invested 3bb to see the flop. What do you think about that trackleberry?

      Now a couple of questions regarding the video and stuff:

      1) The conclusion was that we need to call with our oed because otherwise we will either check raise bluff too much or have to call with weaker hands like gshots. But calling with an oed with no showdown vlaue on the turn is questionable I suppose (following the vaccum approach), since we can also loose to some of our oppoennt's bluffs that give up on the river (unless we develop a donking range in the river). Do we sometimes need to include this kind of Ev- calls to balance our range? I thought all our moves should be Ev+.

      2) Regarding the min defence frequency, you are saying the we souldn't keep our opponent's bluffs Ev=0 but just indifferent between checking and betting. And since checking always has some positive EV, then we should almost always defend less than the min defence frequency right? But I often hear that we have to defend more becasue opponent's bluffs have equity...

      Thanks!
    • GingerKid
      GingerKid
      Black
      Joined: 05.08.2007 Posts: 5,536
      Originally posted by Tackleberry
      Hello GingerKid, awesome work. I was already aware of what you expected to see - but I think your "approach" is not universally applicable, but I´d like to await the "final ranges" you're coming up with and then I´ll analyze in detail.

      At the very least it seems as if I´ve sparked something in you which is valuable in and of itself, right? ;)
      Yes you inspired me. I think I have general decent step by step approach. I did it in my computer, I will post it here tonight or tomorrow when I finish it.
    • GingerKid
      GingerKid
      Black
      Joined: 05.08.2007 Posts: 5,536
      Originally posted by Shakaflaka
      Janda's recomendation is to keep the EV of opponent's bluffs significantly below to money they invested preflop. For example, if you open raise to 3bb, their Ev should be way below that becasue they invested 3bb to see the flop. What do you think about that trackleberry?
      It makes sense what you wrote here, and it is to show that even if villian can have +EV bets once we check with all bluffs, he still plays "bad bluffs" -EV for the whole Hand, since he invested money preflop. It is the case because he doesnt get to this profitable spot often enough. It is similar if villian cbets turn, and if hero defended flop
      more agressive than in average, he can let villian have +EV bluffs on turn because he had -EV bet on flop, so we should make sure that his

      EV_bet_turn + EV_bet_flop = EV_check_turn

      Which would be indifference principle. As and example, if villian cbets pot on flop, and we defended e.g. 80% of range because we hit good, then on turn we can afford to fold vs pot size bet e.g. 55%.


      2) Regarding the min defence frequency, you are saying the we souldn't keep our opponent's bluffs Ev=0 but just indifferent between checking and betting. And since checking always has some positive EV, then we should almost always defend less than the min defence frequency right? But I often hear that we have to defend more becasue opponent's bluffs have equity...
      If his bluff has 0 equity vs our range, then EV of check is zero. In that case we should defend 1-alpha which holds only for 100% polarised ranges and in that case he has EV of bet zero.

      If his bluffs have equity, and we defend exactly 1-alpha, then if he has EV=0 for betting with 0% equity, then he obviously has EV>0 with equity>0 for betting.

      It means that if his EV_check = 10bb, then we should defend in the way to make his EV_bet = 10bb.

      One problem is, that it is very difficult to estimate his EV_check and EV_bet, so even if all of his bluffs would have same equity which is >0% it would be super hard to make EV_check = EV_bet. Another problem is that villian has many different bluffs having different equity and playability and thus different EV_check and EV_bet.

      To show that it is impossible to make villians bluffs indifferent, lets say that villian has only 2 bluffs in range. 1 bluff has 0% equity, and thus EV_check = 0. Other bluff has e.g. 20% equity and thus EV_check > 0. if we defend exactly 1-alpha, we make his 0% equity bluff indifferent, but other bluff will not be indifferent. If we try to make his 20% equity indifferent, then the 0% equity will not be indifferent. So it means that it is impossible to make all of villian bluffs indifferent, not even most of them.

      One thing which is not clear to me. If we defend more than (1-alpha), can it be that villians bluffs gain even more EV with bet? I think it can be, if we have to overfold river due to having too weak range. Or if he has a draw, if he hits it on river, he can get more EV with bet than check if we give him good enough implied odds. We can say, GTO player will not make so stupid things to overfold river or to give too much implied odds, but we are not gto players, and who knows if defending ((1-alpha)-X) on flop and turn is GTO play?

      One more thing that confuses me. GTO exists only for zero sum games. Poker is not a zero sum game because of the rake. When we analyse EV, we ignore rake all the time. Rake is not like statistical error (totally unimportant) but it is actually influencieng EV a lot! So it means there is no GTO in poker, because rake cant be excluded from EV calculations! It means, just keep on defending (1-alpha) vs "balanced villian" and you are garanteed to loose money due to rake (unless you have insane rakeback). Also, if you follow nash equilibrium and construct barreling ranges, you are balanced but you gave very wrong pot odds to villian because he cant call it EV>= 0 with his bluff catchers. So I can conclude, it is the best to always fold vs balanced players on flop, and continue playing only very good hands which are sure to play +EV considering rake. It is the best for both gto players if they agree on only checking always postflop and waiting for fish to come. So that would be max EV of two excelent players.
    • Tackleberry
      Tackleberry
      Coach
      Coach
      Joined: 05.05.2006 Posts: 1,668
      So many good points have been made! :) Let me add some thoughts - and answer the questions you put on me:

      Originally posted by Shakaflaka
      Janda's recomendation is to keep the EV of opponent's bluffs significantly below to money they invested preflop. For example, if you open raise to 3bb, their Ev should be way below that becasue they invested 3bb to see the flop. What do you think about that trackleberry?
      Trackleberry is nice. :D Yeah, what Janda refers to is that if Villain would make 3bb profit on the flop (against our entire strategy), he could literally call any2 preflop (because he gets his money back and something on top). But obviously it´s not that easy. First of all, the EV of opponent´s worst hands should be < 3bb BEFORE we do anything on the flop, not once we checked (and Villain starts bluffing). Like Ginger said, if we (profitably) cbet 90%, it´s almost irrelevant what we do when we check, Villain still can´t call pre, just hoping that we check and he has an auto-profit spot. That way it´s technically correct, but almost impossible to translate into actual betting / defense strategies.

      Originally posted by Shakaflaka
      1) The conclusion was that we need to call with our oed because otherwise we will either check raise bluff too much or have to call with weaker hands like gshots. But calling with an oed with no showdown vlaue on the turn is questionable [...] Do we sometimes need to include this kind of Ev- calls to balance our range? I thought all our moves should be Ev+.
      No, we actually "NEVER" do anything that´s -EV, just to balance our range. In fact all our moves have to be at least 0EV. The reason why we still can call with our OESD (at least do I think so, in fact I might be wrong), is that we have at least some non-zero implied odds and we can use those combos as x/r-bluffs on the river (what supports our river-bluffing-range because it keeps some "zero-EV-hands" in our river-range).

      Originally posted by Shakaflaka
      2) Regarding the min defence frequency, you are saying the we souldn't keep our opponent's bluffs Ev=0 but just indifferent between checking and betting. And since checking always has some positive EV, then we should almost always defend less than the min defence frequency right? But I often hear that we have to defend more becasue opponent's bluffs have equity...
      Both is correct! You defend less than 1-alpha, if Villain´s bluffs would have +EV when checked behind - but you defend more than 1-alpha, when you´re calling instead of raising (and let Villain realize his equity).

      Like imagine a simple scenario: Villain has 20% equity and bluffs. If you shove (1-alpha) and fold the rest, Villain´s EV is zero, so he´d be better off checking. If you called (1-alpha), Villain had a zero-EV bluff and now has 20% equity on a way bigger pot (!) - which means his EV increased. So, the golden defense% is impossible to calculate, we can just estimate it. Bluffs with SD-equity => defense goes down, defend passively => defense goes up.

      OK?

      Originally posted by GingerKid
      One thing which is not clear to me. If we defend more than (1-alpha), can it be that villians bluffs gain even more EV with bet?
      Sure! The EV of the turnbet is the result of EV(turn) + EV(river). If we overfold the river or offer massive implieds to Villain in a way that his positive EV(river) is way higher than his negative EV(turn) absolute, then his overall EV on the turnbet is +EV.

      Like if he makes -2bb on the turn, but +8bb on the river (and he only comes into that spot by betting the turn), then taking the "negative" EV on the turn is the right decision. Does that counter the initial statement that we shouldn´t take any -EV-decisions? No! The turn-decision is only "-EV" if we open-folded the river (as Villain), but as that doesn´t happen, the total EV on the turn is +EV (as shown: -2 + 8 = +6bb).

      Originally posted by GingerKid
      and who knows if defending ((1-alpha)-X) on flop and turn is GTO play?
      Nobody knows! :D

      Did you ever hear about the methods to determine the exploitability of GTO-bots? They run simulations against their GTO-bots to look for "leaks" and then relentlessly exploit them. ... In the end they get to a "bb/100"-estimation of how exploitable their bots are. Why don´t they simply stop being exploitable then, when they know they are? Because it´s impossible - given current knowledge! Knowing that I´m exploitable is not the same as knowing how to play optimally!

      And that´s exactly the target of the nemesis approach: we don´t get to optimal ranges in a finite number of steps (as frustrating as that may sound first), but what we do is that we watch out for leaks in our strategy, small things that allow our Villain to profit from. And then we look for possibilities to fix them (which is not always easy, it´s oftentimes just try-and-error, to look what works best). Once we think we found a fix, we have to test that new strategy (and most likely recognize that we decreased our exploitability but we didn´t eliminate it - which is the reason why I think your approach might certainly not be the "finite" solution you´re looking for).

      Agree?
    • GingerKid
      GingerKid
      Black
      Joined: 05.08.2007 Posts: 5,536

      No, we actually "NEVER" do anything that´s -EV, just to balance our range.
      Hi,

      Can you please explain why is this the case? Imo, the only thing which matters is maximizing our EV of whole range. Why is it impossible that making -EV action with one hand can increase EV of same action of other hand in range?
      Or do we actually consider all hands in range to be statistically independent?


      Like imagine a simple scenario: Villain has 20% equity and bluffs. If you shove (1-alpha) and fold the rest, Villain´s EV is zero, so he´d be better off checking. If you called (1-alpha), Villain had a zero-EV bluff and now has 20% equity on a way bigger pot (!) - which means his EV increased. So, the golden defense% is impossible to calculate, we can just estimate it. Bluffs with SD-equity => defense goes down, defend passively => defense goes up.
      Another great explaination, thanks for this! you dont write much, but when you write it explains so much :)



      Did you ever hear about the methods to determine the exploitability of GTO-bots? They run simulations against their GTO-bots to look for "leaks" and then relentlessly exploit them. ... In the end they get to a "bb/100"-estimation of how exploitable their bots are. Why don´t they simply stop being exploitable then, when they know they are? Because it´s impossible - given current knowledge! Knowing that I´m exploitable is not the same as knowing how to play optimally!
      I heard about gto bots but not about such simulations. Also they play HU which simplifies a lot their strategy? I didnt get it, how exploitable the bots are? I can imagine that they have so many small leaks, but someone who plays vs such bots and knows all leaks has to play huge amount of hands to gain profit vs it. And with rake, I am not sure if there is profit at all vs such bots?
      With approach I try to make, I dont expect to have optimal ranges. Just to have good enough ranges, which doesnt have big leaks and to be able to construct them in a fast way (not to need 1 hour for it). Also, we dont have to play perfect poker in order to win in poker, we just have to kill the fish and be better than average reg.
    • 1
    • 2