Its looking to have the proper ratio of hand classes across all possible runouts (or something like that). It also considers which hands make the most efficient raises to fold out better (raise K6s to fold out KJo, K8/9s, A6s …)
When you look through the sims and see the strategies, you get a better feel for what’s going on. I went all the way down the rabbit hole trying to chase perfect play before I came to my senses again. Now I’m looking for simplifications that don’t lose more EV than the mistakes I make with more complex strategies. When I look at the hand matrix above, I know that the strategy to play it will be far more complex than I can handle (and this is a pretty tight range). I want strategies I can understand and execute well. Should I run into someone like Limitless on the tables who can tear apart this strategy, I’ll cash out and find another game
Poker is like throwing knives… it all comes down to range estimation.
Obviously, we would play differently against someone opening 6% of his hands than against someone opening 20%. But that’s only a +/- 7% spread.
At what error point does this all fall apart?
How do you cope with someone who switches gears?
If you have history on someone (which we used to call “having book”), what makes you think they will still be playing the same as when that data was collected? (I have run into this problem more than once)
I’m not bashing anyone or anything, I am genuinely wondering… has anyone done a serious study on how errors in range estimation impact “optimal” ranges and lines?
Great questions but they are more appropriately addressed to exploitative strategies than to GTO-based ones. The simple answer is that the optimal strategy doesn’t change or need to change or really care what the other player is doing. All strategies other than optimal will lose to it. The farther the opponent strays from optimal, the more he loses to optimal play. Go too loose preflop and you lose in one part of the game tree. Too tight c-betting frequencies lose in another part. Fold more or less than MDF or choose the wrong hands to do either with and you lose in every remaining branch of the tree. Every error, no matter where it occurs in the game tree, loses EV to optimal. That’s the beauty of building a solid GTO-based default strategy - it captures EV vs opponents errors, no matter what those errors are or how frequently they are made. It captures them even if you aren’t noticing they are being made.
Yes, we don’t know what “optimal” actually is. There is EV left out there to be found in new lines and understandings. What we have is close enough to capture the bulk of EV from other strategies though. Also, optimal play won’t maximize winnings vs exploitable opponents. However, in order to exploit someone, you open yourself up to being counter-exploited.
What GTO-based play does is give a player a solid baseline to build exploitative strategies from. If I have no reads, I use my baseline game and collect data until I have enough information to adjust. These adjustments aren’t going to be huge one way or the other unless someone is consistently way out of line or unaware of the mistakes they are making. If I notice someone adjust back, I can retreat to baseline.
Anyway, errors in estimations of ranges impact exploitative play orders of magnitude more than they impact strategies based on GTO, as admittedly flawed and incomplete as that is.
I do think 99.9% of the players on Replay at least are significantly off of equilibrium (myself included). And it tends not to be a small thing. Looking mostly at players in the top few thousand:
open raising (RFI) frequencies are an order of magnitude low in most cases
cold calling ranges are 2 to 20 times wider than they should be
people over fold post flop
bets are heavily skewed toward value
I’d note too that you probably won’t want to just keep marching down the GTO tree after you opponent has jumped off earlier. If you’ve made big bets on the flop and the turn, and your opponent isn’t defending at anything remotely close to MDF, your opponent’s range is much stronger when they wind up still around at the river, and the value part of your river betting range has to contract, which then also forces a reduction in combinations of bluffs. If you don’t account for this, while you’ve won a lot of money with all the folds generated on the flop and turn, your river bluffs and weaker value will fair quite poorly on the river.
Correct and that only increases the amount of EV available to be passively captured. I’m not sure if you were around when he was here but the top player on this site used to be someone named “unranked”. He played a GTO-based game and freely admitted that he wasn’t actively exploiting the population very much. He just collected the EV of their mistakes over and over and over again. Then another player came into the mix named “idi0tplayer”. She was purely exploitative and rose quickly to #1 or #2. She stated clearly that she was not balanced nor did she want to be vs this pool.
They had entirely different strategies and yet they both wound up in the same place here. The amount you could actively exploit the pool was enormous. The amount of EV you could passively capture from the pool was somewhat smaller but still substantial. I found it interesting to see how well both styles worked against the same opponents.
Agreed that you can and should deviate from optimal lines when opponents are making the same mistakes over and over again. As I said earlier, the GTO based style will not capture as much EV as the well executed exploitative one. That’s something that holds true to one degree or another in almost every pool. You make minor adjustments to account for player/population tendencies but your core strategy remains the same. In your example of players reaching the river with far stronger ranges than they should, you could adjust by dumping just a few bluff candidates or by sizing up your strongest value hands.
I guess I should just watch you guys debate the best way to climb a tree in order to get that last cherry on top, while I move to a different tree and feast on the low hanging fruit.
And I thought only me and miniature horses liked low hanging fruit. . I’m still of the mindset of starting at the bottom and picking my way to the top… poker players that is …
I thought I’d quickly add 2 more cases defending as the big blind, with one case addressing a defend against someone that really opens wide (and there are certainly some players at the top that do this). Assuming in both cases that the small blind had folded to the button raise, and that there were no early limps.
Case 6: as BB defending against tight button that only opens 16%
Even though this is a button open, I’m assuming here that we have enough data to be reasonably confident that they have an unusually tight opening range. We probably can’t really pin it down to 16%… it might be 14% or 19%, but we are pretty confident it’s kind of in that range. Here, I’ll probably fold about 70% of my hands.
Always 3 bet AA - QQ, AKs and KQs
Mostly 3 bet AKo, AQs, KJs and QJs (flatting when not 3 betting)
Mostly call AJs - A2s, with AJs, ATs, A5s and A4s having reasonably high 3 bet frequencies, and the others being 3 bet rarely
Mostly call all Kx suited from KTs down, 3 betting KTs, K6s and K5s a small amount
Mostly call with all other pairs, 3 betting JJ and TT a fair amount, and others down to 55 very rarely
Mostly Call with QTs through Q5s, with some rare 3 bets with the first three (QTs to Q8s)
Call with suited connectors and suited single and double gappers, raising occasionally with some of the connectors and single gappers
Always call with AQo - ATo
Mostly call with QJo, QTo, JTo, folding when not calling
Mostly fold with KTo, T9o, Q4s and 32s
Case 7: defending as BB against button with 50%+ open
3 bet AA - 99, AK - AQ, AJs - ATs, A5s, suited connectors KQs - 54s, suited gappers KJs - T8s
Mix mostly 3 betting with some calls with A4s, KTs, Q9s, 97s and 88
Mix mostly calling with some 3 bets with A8s-A6s, A3s, K9s-K6s, Q8s, AJo, KQo, KJo, KTo, 77-55, J7s, T7s, T6s and 96s
call with A2s, ATo-A3o, K9o-K6o, QJo-Q8o, JTo-J8o, T9o-T8o, 98o-65o, and all other suited hands except 92s, 83s, 82s and 72s
call some fraction of the time (and fold the rest) with A2o, 97o, 54o
You can see that latter one in particular gets pretty messy, 3 betting a lot of hands that don’t feel that strong, and also calling with some junk like 97o or T2s. But your opponents raising range is littered with junk, also.
Lets look at the same situations, button raise and you are the big blind, but there was one or more early limps. What now?
In general, some players will limp 3! or limp 4! more than others. If the limpers are in that category, I’ll normally dump my flatting range, expand by 3 betting range quite a bit, and raise or fold. I’m no longer last to act, and a call can incentivize one of the early limpers to squeeze, leaving me in a bad spot.
If the limpers are really passive, and they rarely limp raise, then I’ll mostly continue to split my range between 3 bets, calls and folds, but will shift a few more hands to 3 bets to take advantage of the dead money. I personally like to add high off suit hands here to the 3 betting range, since they are especially unhappy multi way out of position, but do better heads up with lower SPR. There is a down side risk, of course, in that many of the hands that will continue can dominate you.
OK, I had some time to get a bit over 3,000 hands of $.05/.10 zoom in to test the strategy out and see what happened. I know it isn’t a big sample but I broke it down into 3 sessions and all the results were similar. I wasn’t actively exploiting anyone since I didn’t know anyone in the pool. I played without a HUD because it was zoom format (fast fold).
These results are from executing a baseline strategy involving a 3-bet or fold from every position other than the BB. Just simple mechanical play using base ranges and standard bet sizes/frequencies postflop. RFI was to 2.5x LJ, HJ, CO: 2.75x BTN: 3x SB. SB played as raise or fold.
Yes, the stats are a bit tight but fast-fold tables play this way, especially in high-rake environments. Its a position game to a large degree. Solid, tight ranges, lots of 3-betting rather than flatting and a simplified postflop strategy.
Here’s a screenshot of a solved BB range vs a ~42% BTN open of various sizes. Sorry, I don’t have one for a 50% range - just expand both value and bluff 3! at the margins if you face a range this wide.
Red is 3-bet vs any of these sizes
I love how this shows the expanding call range versus smaller open sizes. I suspect at least some players wouldn’t expect to modify their calling ranges quite this much in response to seemingly small changes in the size of the raise they are facing.
In this particular case, I think you are holding the open size to roughly 42% no matter the sizing, but in general, I’d expect the smaller raises to also represent wider ranges, which would further amplify this (and would also then impact the shape of the range you are 3 betting).
Man, the rake there is nasty. I don’t imagine it is a very large percentage of players that are good enough to overcome that.
I love what a high 3 betting frequency you’ve generated with this strategy. In general, I wouldn’t recommend that here at Replay, just because there are so many players with raising ranges in the 2% to 8% ballpark, and then you’d be 3 betting a weaker range than they are initially raising with, but assuming you are facing players that are open raising at more normal frequencies, well… I think the results speak for themselves. That’s a great win rate at a real cash site (though I’d guess post flop skill is a large factor here also).
Rake is the primary reason 90%+ of players aren’t profitable. At microstakes online or 1/2 live its almost obscene. These pools are extremely soft because anyone with a decent winrate moves up as soon as they can. No one wants to grind in a punitive rake environment if they don’t have to.
IMO it is 100% replicable for anyone willing to do just a little work. Maybe not 20bb+ but certainly 10bb (net of rake) within a few months of study. People willing/able to exploit the pool can probably do significantly better. My postflop was almost autopilot - using a simplified strategy and not polarizing until the turn. Limited bet sizes on all streets. This is the power of playing a GTO-based strategy even vs a weaker pool. Passively collecting EV from their mistakes without taxing yourself on each decision point.
It’s actually higher on regular tables for the games I’m playing normally. The coach I’m working with wants to see me between 10 and 12%. I need to find more 3-bets from the BB and stop being so lazy about taking flops. Its easy to pull the trigger with the value but its tough to reach deep enough into the garbage for me to achieve the rates I want. Part of going to the 3! or fold strategy was to force me to find more of them.
If people are only opening 2-8%, its like playing against a pure value 3! range. Vs a pool like this I suppose you could make a standard 3-bet range your isolating range. Maybe treat the limp like an open? What I saw on the tables played for this experiment was the opposite. People were opening ranges far too wide for them to defend properly. They can get away with it if no one is punishing them but they cannot withstand a wide 3! range. They open too wide and don’t 4! or fold enough OOP. They let you play heads-up, in position, with a stronger range than them. You don’t have to be very talented to make money if they cede these advantages to you over and over again.
Its a good way to visualize how ranges expand and contract depending on the preceding action. I wouldn’t implement this in practice myself though. I can’t find a way to make many of the hands in grey profitable if your opponent is c-betting nearly enough. I won’t defend K2-K5o, Q6-Q7o, J7o or the weaker unsuited gappers most of the time.
Anyway, I find visuals to be helpful when trying to understand concepts. Seeing how hands are added/subtracted to a range is more important than getting the exact ranges correct.
I think this gets to what I suspect is the single biggest source of EV for me over the years. I’ll have multiple bullets here, but somehow this is all just one idea for me.
I raise over limpers fairly frequently
I 3 bet over raises fairly frequently
I 4 bet over 3 bets fairly frequently
I they over fold, I profit
If they call too wide, especially out of position, I profit
Especially as players face aggression they aren’t used to, they have difficulty finding the right calling and raising ranges. If they call too much or too little, or raise too much or too little, you usually wind up profiting.
On a side note, this is exactly the predicament El-Jog and gamergirl typically put me in: all of a sudden I’m facing levels aggression I’m not used to, and I’m too slow to find the right defense and re-raising frequencies.
Oh, and one other related point: if the raising range is stronger than the calling range, then another key point is just how this means that you are repeatedly playing post flop with a statistical advantage, for a larger starting pot. Let’s say you call instead with most of that range; you might still have nearly the same post flop advantage, but by starting with a smaller pot, you’re going to win less on average.
What would you consider “frequently”? In the games I play, a 3! is almost always a super premium, so I can’t usually 4! with hands I’m not willing to get it in with preflop.
. I’m still of the mindset of starting at the bottom and picking my way to the top… poker players that is …