hero 3bets and population...?

    • tolari
      tolari
      Black
      Joined: 02.11.2010 Posts: 761
      Hi!

      I've played with database and wondered what are the % for each outcome when we 3bet from each position against each position as UO pfr.

      I've runned this on 110k hands sample, nl50 zoom. My 3bet stats are (vs opens)
      vs EP 5%
      vs MP 8%
      vs CO 11%
      vs BU 15%
      vs SB 17%

      my 3bet % are:

      mp 4,6%
      Co 6,4%
      Bu 9,9%
      SB 10,2%
      BB 9,1%

      My fold to 4bet is 63,1%

      overall outcomes %:




      detailed outcomes:



      (saw flop calculated obviously with filter - hero faced 4bet after 3betting: false)

      I am sharing this with you to see what conclusions you will draw from this.

      For me it was rather surprising we see folds so little and I've completely stopped punishing ppl who seem to fold to 3bet to wide. I mean I usually 3betted for the sake of autoprofitable 3bett often with real garbage - since I thought hey he is folding enough to make this 3bet > ev than 0 -> folding. now I try to develope a solid range and stick to it, even after this my 3bet% didn't go that much down...

      also gives clues how to construct 3betting ranges, BU blockers heavy, MP CO SC,middle PP, SB&BB balance of both.

      Also I've runned similar tests on nl10 sample (really small - 60k hands) where my 3bet was 16%. And I've got similar folding equities while 3betting 16% at nl10 and average 9% at nl50.

      Dunno what to make of it but maybe there is a boarder to which the population will adapt regarding wide 3betting?? but probably ppl are so bad at nl10 that let you get away with it...
  • 7 replies
    • lnternet
      lnternet
      Bronze
      Joined: 19.06.2012 Posts: 782
      A few things are unclear, would nice if you could help me understand

      1) Did you filter for hands where you 3bet or for whenever anyone 3bet?
      2) 4bet filter: did you include cold 4bets or not?
      3) Saw flop filter: did you include coldcalls of the 3bet or not?
      4) What is your sizing plan? Both IP and OOP to various open sizes?

      Regarding 1) 110k hands might not be enough sample to look at these stats at least for the early positions if you only look at Hero hands. Just think about how often you 3bet a UT player from MP. You are MP once every 6 hands. UT opens maybe once every 6 hands of those. Then you 3bet once in 20 hands of those (5% MP 3bet). This means you 3bet MPvUT once very 720 hands, or about 150 times in a 110k hand sample (hand samples have less than 6 players on average also). So for 4bet UTvMP I see 11.38% in your chart.

      Assuming this is the true value, we have a binomial distribution of p=11.38%, n=150, approx. Standard deviation of bino is sqrt(np(1-p))=sqrt(150*11.38%*(1-11.38%))=3.9. So this means a 68% confidence interval would be around (17+-4) or (13 to 21) or (9% to 14%) as 4bet percent. 95% confidence interval gets you between 6% and 17% 4bet. So given your observations of 11.4% 4bet UTvMP, you can only be 95% sure the real value is between 6% and 17%. And those two numbers have vastly different meaning, so we can say your sample is surely too small to look at UTvMP 4bets.

      Lots of random math to say your sample size is a bit too small to look at some of these numbers.


      I would like to see your HM filters, or sql queries, for how you get all the percentages in your Excel. I don't mean to say you did them wrong, but since you did do them, why not post them along, so we can sure you actually have the numbers we think you have. Nothing is less useful then spending hours analyzing data and then finding out the data is biased or simply wrong.


      Another thing is you group regulars and fish together. It's difficult to filter out fish, but this would better the meaning of the results a lot. Of course we never 3bet bluff fish, and their presence of course skew all stats to being more loose/passive (more calls, less 4bets, less folds).

      Lastly, save images with text in them as .png instead of .jpg, there will be no blurr.



      Now I sound so negative, haha. Good work anyway, very good to spend time on this!
    • tolari
      tolari
      Black
      Joined: 02.11.2010 Posts: 761
      1. I filtered for my 3bet.

      2. for 4bet% I filtered: did 3bet; faced 4bet when hero 3bet: true; saw flop: false(to not mix them up with saw flop filters)

      3. I have filtered: did 3bet; saw flop: true; faced 4bet when hero 3bet: false

      4. IP: vs 3bb open - 3bet to 9bb
      vs 2bb open - 3bet to 6bb

      Sometimes changes, if vil 4bet bluffs alot I make it 3x+1 to make his bluffs more expensive and to have more profitable push, and some spots like this but i would say less than 5% of the time.

      OOP: vs minraise: 3bet to 7bb
      vs 3bb open - 3bet to 9bb/10bb

      I use HM2.

      filters on screen: (I obv. runned them for every open raiser position and each of mine positions)




      I don't mind negativity:D thanks for taking a look at this, I've done it because I have major problems with my 3bet bluff profitability.

      I know the sample is not perfect, but I don't mind running test again after 200k or 300k and comparing.
    • lnternet
      lnternet
      Bronze
      Joined: 19.06.2012 Posts: 782
      OK looks good.

      You have cold calls against 3bets in the saw flop and cold 4bets in the 4bets, but that's fine.

      One big thing I would change is that you add "preflop action facing hero: [x] 1 raiser". Right now you are also fetching 3bets where you squeezed, which I don't think you want.

      Not sure how the "Faced preflop 4-bet after Hero 3-bets" works in regards to 5bets, someone might shove before the action gets back to you. But it's only a rare thing to happen.

      To get to percentages you first gather all three filters #hands and then divide the respective filter number by the sum, right?
    • tolari
      tolari
      Black
      Joined: 02.11.2010 Posts: 761
      I divide each filter by the number of 3betts total from filtered position, just to check if all hands were calculated. So pretty much yes:D


      You are right about sqz, I thought HM2 didn't count them as 3betts since it has another filter specially for them. Doing this analysis for sqz won't make much sense now...

      Also what I think might influence the results is the fact that I play very often 200bb+ deep maybe thats the reason for such high "saw flop" %.

      would you draw any conclusions from this as it is now?
    • lnternet
      lnternet
      Bronze
      Joined: 19.06.2012 Posts: 782
      The last thing which is important. Since you did all these vsHero, we need to know how well your opponents know you. Do you play with the same people a lot? Or do most opponents have a max of 1000 hand sample on you so they don't know your stats and are playing default as opposed to Hero-adjusted?

      Ok so much for the methology. Now to the results.


      UT folds at 28-38, 34 average, are super low. 4bet high and call super high. Your cutoff 3bet bluffs from IP will still show a profit I think. That is Axs/KQo type hands from MP,CO,BT vs UT will be ~break even 3bets. But 3bet bluffs beyond that won't show a profit.

      MP folds 37 average, 4bet high. We can clearly see MP 4bets vs IP 3bets are much higher than MP 4bets vs OOP 3bets. This is quite cool too see, people know what they are doing at NL50 on average. I think the same I wrote for UT applies here. You can 3bet the correct 3bet bluff just around break even, maybe slightly profitable. But 3bet bluffing MP with 65s won't show a profit.

      CO folds 39%. Here we see a huge gap in 4bets. COvBT is at 28%! This is clearly an adjustment to Hero. There is no way in hell the COvBT average 4bet is that high. People are trying to exploit your high foldto4bet I am sure. (Your 63% is terribly high, you need to fix that. Good thing you pulled this info!) Same deal as other positions. CO folds a bit more but also 4bets a lot more, again keeping your solid bluffing hands at par.

      BT folds 48% and 4bets 16%. This is a bit tight. Given your OOP sizing you can probably 3bet bluff 72o profitably. You need like 66% folds for immediate profit, you got 50%, but a lot of postflop EV. So I think you can increase your blind vs button 3bet. Surely not down to 72o but add a few more bluffs.

      SB same as button. They defend slightly better, but you 3bet smaller IP, so still any two are a fine 3bet. Would actually recommend never folding BBvSB against a lot of people. Just 3bet bottom 30% trash, it's gonna show a profit vs the average player (until they have reads and adjust, of course)
    • tolari
      tolari
      Black
      Joined: 02.11.2010 Posts: 761
      Originally posted by lnternet
      The last thing which is important. Since you did all these vsHero, we need to know how well your opponents know you. Do you play with the same people a lot? Or do most opponents have a max of 1000 hand sample on you so they don't know your stats and are playing default as opposed to Hero-adjusted?
      I play zoom so if they don't buy database, they have max 2k hands against me and majority of field around 200hands.

      thanks a lot for your imput!

      I will work on establishing some 3betting ranges from position against some archetypes of vils and post it later.

      Regarding 4bet populationCo vs my 3bet - damn you are right... completely forgot to check this stat by position! I've got

      57% ft4bet on mp,
      57% ft4bet on co,
      61% ft4bet on sb,
      53% ft4bet on bb,
      aaaand 70,2 % ft4bet on BU ;d yep.. that explains a lot. gonna work on developing flat calling range vs 4bet, from blinds I 5bet bluff in quite ok spots I guess (ev of 5bet bluff > ev of 3bet folding) and for sure gonna pay more attention to players 4bet vs BU 3bet..

      To widen my 3bet% against BU I need to work more on postflop game I think, been playing lately with multibarrel bluffs in spots I didnt do before with ok results but still need bigger sample and probably more playing with flop zilla.

      Coz one and done doesnt seem like a perfect strategy since I already cbet a lot and get floated all the time:D I can play fairly well against cbet raises since ppl do it very often on boards where they don't rep anything just to "check" if you have "it" so I just get over the top if they raise on Khigh or some Ahigh boards.

      I like the idea of not folding BB vs SB against some players, but to do it I need to get rid of BU 4bet folding leak - so when I start 3betting 40% against sb open I won't bleed moneyz

      maybe develope some 5bet chip bluffs bb vs sb (not 5bet shoves) maybe making 3bet to 8bb and then have space for small 5bets ( 3bet to 8bb sb 4bets to 17-18bb, i 5bet to 35-37bb) hands like KQ,AJ have around 29% equity against most ppl jam range. I would obviously do the same sizings with stackoff range.
    • IronPumper
      IronPumper
      Coach
      Coach
      Joined: 03.01.2008 Posts: 14,838
      interesting thread