Thanks for the shoutout and thanks to RIO for the complimentary one month elite membership that made my comments possible. I'm just trying to give some value back.
I agree with 528632484 that this vid's popularity is because it is "more applicable in practice than most theory vids". It also shows how to use Pio for exploitative analysis. I think many players are struggling with Pio, confused by all the numbers and confused by optimal play.
Also, OOP play is harder so the spot you chose to analyse is one we all recognise as needing improvement in our own play but are unsure how to go about it. Similarly, IMO the flop is deceptively difficult in that small flop mistakes in range composition compound and get geometrically exposed on turns and rivers but most players have trouble diagnosing the turn/river problem as flop error. Traditionally the flop has been presented as the easiest street but I think Pio is showing it is the hardest street.
I would like to see more exploitative theory vids:
Nick Howard and Tyler Forrester seem to advocate pure rather than mixed strategies including substantial OOP range checking but Pio ranges are highly mixed. How do you exploit an opponent seeking to simplify by adopting pure strats? (Besides polar, linear and tricky reg, range check is an emerging fourth flop strat).
I think many of us have less than optimal flop betting ranges but struggle trying to go from a polar or linear range to optimal without it becoming overly random and illogical. A related struggle is how to quickly adjust that pseudo-optimal range to exploit opponent tendencies. I would like to see a vid (or series) where you manually build a pseudo-optimal flop betting range (as you quickly did here for polar, linear and tricky ranges) but take it slower, break it down and justify why each category of hand at what frequency is bet and explain the exploitative holes left open if this category is missing or under or overrepresented. At each iterative refinement category step compare the pseudo range to Pio's optimal range by asking Pio to exploit the manually created range (and by using Pio's range arithmetic to add/subtract the pseudo-range with Pio's optimal range). Iterate and discuss. As part of building the optimal range or as a second or series of vids explain how you would adjust the optimal range to be exploitative of certain opponent reads and demonstrate its effectiveness by using Pio to adjust and nodelock opponent tendencies.
Dec. 12, 2016 | 11:38 p.m.
A forced OOP Check costs 8bb/100 at an inaccuracy tolerance of 3.2bb/100 assuming the modified IP range above with slightly lower 55,44,54s in contrast to Ben's original IP range where a forced OOP check costs nothing.
In case IP does not cold call small PPs (and optimal play would be close to bet whole OOP range) instead an OOP Check costs 28bb/100 EV.
(A typical winrate in MP is about 15bb/100 across whole MP range across all board runouts so the voluntary loss of 8+ on too many flops is probably a big deal).
Dec. 11, 2016 | 9:06 p.m.
Barely reducing the weighting on 55 (75% down to 50%),44 & 54s (down to 25%) (11 combos reduced by only 5.5 combos to 5.5 combos) and solving we get OOP Bet just under 50%. In this sim I made no other changes to IP range.
The critical factor affecting whether OOP optimally bets a reasonable frequency is not simply whether 44,55,54s are in IP range but rather their relative proportion. Ben's range triggers near OOP check because 44,55,54s are relatively overrepresented in IP range. You could instead engineer a similar max exploit increase in OOP betting without changing the weight on 44,55,54s by simply increasing the weighting on other hands in IP range or adding other hands such that 55,44,54s become relatively less in the whole IP range.
Such subtle IP range changes producing such large theoretical OOP betting swings are problematic in game for default strategy since of course in many spots we can never have such exhaustive reads on our opponents and even if we did our opponent can ruin that read by "accidently" humanly adding/subtracting a few combos.
Dec. 11, 2016 | 7:27 a.m.
OK, I've now run the sim to test how unstable OOP's strat is versus IP's calling range not including small pairs. Many regs will not call small PP's versus a solid reg so to not include these is a reasonable assumption.
If I change IP calling range to
and then maximally exploit we get OOP Betting pretty much his whole range
So OOP strat is extremely sensitive in this spot to assumptions made about IP range and strategies.
Footnote: OOP Bet 79% solely by changing IP range to not include 44,55,54s
Dec. 10, 2016 | 11:42 p.m.
Ben, besides the practicality of an OOP X versus a tough opponent do you think the other takeaway is the need to somehow understand our opponent's range and strategy a lot better?
I suggest this because the Pio strat in this spot seems unstable and highly sensitive to changes in assumptions about IP's range and strategy. Default OOP X against some reg opponents could waste a lot of EV. (I'm assuming exploitation of non-regs anyway but it seems our strat against regs needs to be quite variable, too).
Change IP's range, say to not include 44,55, 54 (as you indeed deliberately included for flop coverage in this exact spot) and Pio's OOP strat changes to Bet quite a bit (80+% see next post). There are a lot of reg opponents who will have such holes on this flop. Whenever IP board coverage has holes PIO's clairvoyance means it will exploit maximally, dependent on flop attributes piercing those IP holes. It can be problematic to do that in game.
Secondly, when I experimented with IP strategy, keeping the ranges and other parameters the same as yours Pio still adjusted strats quite a bit. If IP tends to stab flop and then X back Turn and river, then Pio changed OOP strat to flop X 99% and X/R 20% (double the XR of initial sim). On the other hand, if I locked IP to passive strat of X back on all streets, now Pio wants to OOP Bet flop 26% with a range seemingly quite different to your first sim (much more polar I think but I could be wrong as I haven't studied the range in detail) and IP now only RR's 2% (down from 10%).
Thirdly, OOPs response to IP's flops stabs depend on how IP constructs the stabbing range as linear, polar, optimal which suggests a bunch of other Pio sims to validate.
So, in the real world as reg ranges and strats change, OOP Bet in this spot could reasonably vary from X always to Bet optimally (~7%) to Bet polar with frequency varying from zero to quite a bit (80%+). And OOP X/R composition needs to vary depending on IP's range composition and street-by-street tendencies.
Dec. 10, 2016 | 10:27 p.m.
I ran the Pio sim for forced OOP X using otherwise same params as Ben's sims and the EV is the same as the initial sim that allowed an OOP bet (actually tiny amount better but within error tolerance). I suspect the small frequency bet in the initial sim is an artefact of not running the sim to higher tolerance. The optimal OOP bet % in this spot probably converges to zero given long enough sim runtime.
Dec. 10, 2016 | 9:39 p.m.
Note that the effect of SB XR range % on BB flop CBet % is NOT linear. A somewhat slight error in the XR percentage away from the sweet spot (or poor range design amounting to the same thing) and BB becomes incentivised to the appropriate extrema of Betting either close to 100% or betting only 20ish%.
That is why I say that the SB needs to understand the dynamics and construct a mixed range well. Anything else can be brutally punished.
But back to the real world, and, as Nick says, for the moment, BB's seem mostly oblivious to the SB OOP range check. However, if one Nick disciple plays another Nick disciple it is the disciple that understands the BB counter that will crush harder. And, in the internet age, the window for that information to become well known and limit the leak will happen rapidly.
Nov. 13, 2016 | 8:38 a.m.
2) Even if they understand it, they rationalize a more comfortable approach by deciding that they'll exploit you by just widening their turn/river calling range after defending vs your XR with the same (tight) range. If your blueprint plan is to just 1-and-done XR the flop, you're no longer forcing turn/river aggression with bluffs, so they just keep offering you +EV flop spots and never capitalizing on later streets.
The correct counter to a SB who is XR bluffing too much is to re-raise their bluffy range and put their arse to the blowtorch. OOP play is hard if IP abuses its position.
I'll be back with the Pio sim where I force SB to only XR and it can't XC and let's see what incentive BB now max exploits.
Ok, I'm back:
When SB is forced to check its entire range and cannot respond by XC but only by XR and then BB max exploits:
BB Cbets flop Q75r 21%, SB XR 51% (Fold 49%), BB 3B flop 42% (C 6%)
So the correct counter to a SB that checks entire range and then bluffy XR's is to somewhat tighten up BB's flop Cbet (extent dependent on SB XR range and polarise BB's flop Cbet range to match SB's folding percentage and then flop 3B the arse off the out-of-line SB.
^Garfield's bodyguard. I love this game.
Nov. 13, 2016 | 7:57 a.m.
That brings up another point. If you do it right you can just rip through your limit in like 100k hands before anyone understood what even really happened. Then move up and do it all over again. Once you get to high stakes you can just put money on every soft site that offers 10/20 and never really have to account for your shitty image. There's a whole nother level to the career once you expand your vision to include diversifying across sites, to idea being to spread your volume out with super-exploitative strategies. They're also soft a f.
I like this. This is similar to years ago when the standard was to 3B QQ+,AK and a few pioneers tore through the limits by light 3B'ing and it literally took years for the population to recognise what was happening and evolve an effective counter.
Nov. 13, 2016 | 7:52 a.m.
This sounds wrong if you're referring to a situation where IP is stabbing too much vs an OOP PFR who checks his entire range. OOP's range will be stronger than it would be had he played a balanced c-bet range. IP will be exploited here for stabbing too wide. Am I misunderstanding what you meant by "the stab percent"?
In my Pio Sims above the OOP was forced to check entire range. BB betting 80 percent on flop is not betting too much if SB is XR too little. Entire range check does not protect, rather the XR does. As a test, if I sim the extrema with SB not allowed to XR but only allowed to XC or XF and still forced to OOP check entire range, BB's max exploit is to Bet 98% of its range, proving that the XR limits the BB flop Cbet %.
(In any Pio sim the IP betting range is limited by the OOP defensive action. Initial raise is limited by 3B's. 3B are limited by 4B's etc. Flop Bet percentage is limited by XR percentage etc).
Unsurprisingly, in BvB, even when SB checks entire original flop range, BB is still incentivised to Bet flop at a wide percentage if small blind defends that BB flop bet poorly. As I reported in my second post above if I ask Pio to round the OOP flop actions to 1/1 then the maximal exploitation by BB is to Bet flop nearly 80% because the SB defense is insufficiently aggressive to make BB's flop Bet indifferent. Here the total SB defend percentage was the same but it failed to be aggressive enough to make BB indifferent and start checking back.
Against real players who mostly construct their ranges poorly compared with solvers, it would be exploitatively "correct" for BB to stab with close to 100% versus SB check of entire range. So, in mass DB analysis that shows BB's Betting flop faced with SB range check, it is not incorrect for them to do so. Rather BB's in mass DB analysis are exploitively stabbing too little at 50ish% given that the average SB is incapable of XR (and to a lesser extent XC) close to theoretically correct percentage and theoretically correct range construction (which hands check and which hands XR and at what mixed percentage). I further assert that not mixing this percentage and instead looking for pure strat ie hand only in XR or only in XC leads to a max exploit with BB correctly Betting flop at a high percentage.
It is an incorrect approach to take action numbers from raw Pio sims and apply them in comparison to mass DB analysis unless one nodelocks the Pio sims to match mass DB tendencies and then max exploits those tendencies to elicit the proper exploitative response.
The easy way to test what I am asserting is to test the extrema. Ask Pio to prevent SB from XR (either by remove line Check,Bet,Raise or by nodelocking the XR fixed to no hands). When you max exploit the BB bet flop versus SB's entire range will be close to 80+%.
Success with SB OOP range checking is not because BB's are stabbing too much but rather because they design their ranges poorly compared with Pio's betting range.
The advice to check SB OOP entire range is arguably human correct but for different reasons than "to exploit BB high stab percentage" which is actually too low not too high. The actual reason is to exploit BB's poor range construction in this unfamiliar territory in whatever percentage they choose to Bet flop and to exploit their poor responses to that XR.
Nov. 13, 2016 | 7:29 a.m.
Pio sims of Q75r nodelocking a rounded OOP strat. If I round to 1/1 to create a recommended pure strat then XR drops to 43 combos from 61.5 and max exploit IP Flop stab rises to near 80%.
On the other hand, rounding to 1/2, XR only drops 2.5 combos, and IP Flop stab only rises to 37% so a rounding 1/2 strategy seems both more practical than full pio complexity but without sacrificing the ease of playability of pure strat. Rounding 1/2 only cost 0.2BB/100 EV. Rounding pure 2.5BB/100. WDYT?
Nov. 13, 2016 | 3:10 a.m.
I like FR. I'd watch a video on differences.
Nov. 13, 2016 | 2:51 a.m.
More on checking OOP.
Point 1. Many times you say that the solvers stab only 20 to 30 percent after PFR OOP check and then claim that population pools are stabbing too much in these spots (~50 percent). And that by checking IP excess stabbing can be exploited.
But the stab percent is only made indifferent by the OOP defending range, especially the XR range and if the OOP player poorly constructs these ranges he has induced the IP player to stab more ie make the correct adjustment to the poor XR range construction.
I ran some quick Pio sims BvB on Q75r and pure max exploit to equilibrium, yes, the BB flop stab was only 26% but, by simply nodelocking the XR range by stepwise iteration deleting the solver artefacts (cells with frequencies less than practically playable amounts) the max exploit stab jumps to 39% (with cells less than 10% deleted), then to 56% (with cells less than 20%). (The cells deleted from XR are shifted by the solver to either call or fold).
So by reducing the XR range by only 9 combos (out of 64), IP stab max exploited from 26% to 56%, about the population norm. So is the population really overstabbing given the likely poor construction of XR range in the pools?
(By nodelocking IP to underdefend the XR the max exploit stabbing was still over 50%).
(I did not yet try a pure XR strat where each cell is wholly either XR or XC but not mixed in both but I expect it will cause the IP stab max exploit to jump way over 50%).
Point 2. Doesn't checking merely push the mixed strat over the immediate decision horizon, delaying the inevitable mixed strat to XR or to turn decisions?
Checking eliminates the flop CBet range construction decision but the XR needs to be a mixed strat or else the max exploit IP stab will be incentivised to be quite high. It's pretty easy to stuff up the XR range so IP max exploit becomes as high as 90%. It is very sensitive to XR strat.
This seems to take the range construction pressure that was on our flop Cbetting range and move it to the XR range, albeit with more information because we get to see if the BB checks or bets.
While we have eliminated one range, the flop Cbetting range but the XR and XC range absorb the complexity, becoming a greater proportion of the total lines and needing to be better constructed.
Nov. 12, 2016 | 10:16 p.m.
I'd like to ask about your suggested strat of checking range OOP.
Do you intend to always check OOP or was this a misclick or just a spot where people underdefend?
Nick cbets flop OOP at 1 minute (BvB, SB A9 BDFD B 1/3 K72r)
Nov. 12, 2016 | 9:45 p.m.
EV 5BB/100 despite downswing is excellent progress.
I actually found a new mental leak in my game: whenever I'm a big upswing, I feel bored. Everything just goes right no matter what I do so I suddenly start to feel bored and lose focus.
Boredom is a side effect of getting better at playing poker. The more hands you play, the more skill automation you achieve, the higher the likelihood of boredom.
Psychologically, every skill starts out requiring more cognitive attention and then in chunks becomes progressively more automated until in an elite player it is fully automated. So the boredom problem only gets worse with more hands played.
Boredom is a state that occurs when challenges are low, but one's skill level exceeds those challenges causing one to seek higher challenges.
(Edited for brevity) :)
Nov. 11, 2016 | 11:06 p.m.
FWIW, I did a quick sim nodelocking only Q9+ on flop to 100% CBet, meaning that only an extra 13 combos (+2.5%) bet the flop. It appears the Turn play is very sensitive to flop bet of TPGK+.
On the turn after Pio max exploits, instead of a Turn check of 69.63% it dropped to a turn check of 57.3% (-12%) with overbet turn doubling from 17.8% (84.9 combos) to 36.16% (176.6 combos).
So flop betting TPGK too frequently (just an extra 13 combos) seems to dramatically expose the flop check back range to turn stab exploitation (doubling overbet frequency).
But the EV lost was only 1BB/100 which seems insignificant.
Nov. 11, 2016 | 7:58 a.m.
@26:00 QT3hh, Turn 5d, after Flop XX, and Pio Bets Turn OOP only ~30% versus your own estimate of ~50% with 1/2 overbets.
What is it that Pio's IP opponent is doing that is limiting Pio's OOP max exploit to only 30% Turn stab?
How does that compare with what real opponents are doing?
My sense is that real opponents are neither checking as much top pair with a wide variety of kickers as Pio is on the flop (but rather tending to bet TPGK and check TPWK leading to poor board coverage and poor hand strength on Turn) nor betting as much 2nd and third pair on the flop (to fill out the missing TP in flop bet range while setting up unusual two pair combos on both bet flop and check flop lines for turn) and that this information hides a fair bit of good TP after a check/check that Pio is happy to call Turn with while maintaining a diversity of board coverage to hit hard-to-see two pairs on the turn that Pio is happy to continue with and as a result limits OOP Turn stab. WDYT?
Nov. 11, 2016 | 5:04 a.m.
Well no prizes for guessing part of my background is as a research scientist. I have a trained, heightened BS meter for "Truth" and a keen desire to uncover hidden assumptions that might allow the extraction of a grain of usable Truth.
Nov. 11, 2016 | 2:42 a.m.
There are logical patterns in poker as revealed in Pio sims and written about by Will Tipton in HUNL V1 and V2. Logic is maths. (I have studied pure maths and maths stats at postgrad level and the patterns in Pio sims are a lot easier to discern than most postgrad maths). I disagree with those who fashionably claim that Pio is all mindless numbers, especially when they have a vested commercial conflict of interest.
Of course we adjust against "unknowns" based on data being our population reads and consequent expectations such as your read that X back ranges are weak.
No need to get sarcastic mate. I'm actually seeking to help.
At least try to look for Pio patterns. You might surprise yourself.
Last year there was a small group of highstakes HUSNG regs who sought approval for a new third party software tool. At first they received approval but after complaints by competitors their tool was banned.
Their tool was a voice activated Excel spreadsheet with each page containing ranges for every action for every single BB of stack! They had solved preflop HUSNGs and automated it and were crushing the games. I'm not sure but I suspect they also had postflop spreadsheets.
There must be people doing that at 6max with a separate undetectable PC containing all that they need to play close to their understanding of perfectly.
For those of us who wish to play ethically we have to find the patterns to stay competitive.
Nov. 11, 2016 | 2:31 a.m.
Hands that are pure strats at equillibrium are the pretty obvious ones, which us humans can see without a solver that one strategic option is higher EV than the other. Its exactly the mixed strats that matter, since they compose the hands that give us more confusion in game. Saying that a hand is a mixed strategy doesnt say much because if its a mixed strategy both options have the same EV, so if your opponent doesnt adjust you can just pick one and play a pure strategy without losing any EV. Our goal when studying solver is to figure out which direction to push the equillibrium mixed strategies to pure strategies (based on opponents tendencies and range construction) since we are not capable of implementing those mixed strategies accurately.
I don't agree. In Pio sims, there are logical patterns to mixed strats and their percentages relative to the alternative and relative to percentages of other hands matter. Pure strats are highly exploitable. And using the wrong percentages is also highly exploitable. And IMO people are capable of implementing the patterns of mixed strats accurately enough to make an in-game difference. Not the 67% detail but a close enough approximation such as 70% is often surprisingly easy to deduce from the logical Pio pattern. IMO, it is no harder than pot odds or betsizing or fold equity.
It is a common misconception that varying your percentage does not change the EV (predating solvers when we didn't know any better). Actually, when you vary the percentage in a range you change the EV of that hand and of the range. Try it in Pio and watch the opponent exploit the change mercilessly.
The fallacy that a pure strat is required to maximally exploit has its refutation on different board runouts. If you always Bet a particular hand on a particular flop let's say then there exists a set of turn and river runouts after checking where your board coverage is exposed to counter-exploitation. The frequency of correctly checking back that hand is partly related to the proportion of turn/rivers where that hand's board coverage will be needed to maximally exploit the opponent in the checking subtree. When you always Bet some particular hand that should be a mixed strat you may be exploiting an opponent's tendency to overfold, for example, but you check some other hands sometimes and the EV of those hands suffer because you did not check some of the time with the hand you always bet. You also expose yourself to being exploited by being RR'ed at a higher freq. Checking the mostly bet hand some of the time you also exploit him the times when you check. Both betting and checking lines are important to maximally exploit and we do that with calculated patterns of mixed strats not by random play.
Often I find merely playing a mixed strat default exploits a reg. Your ATcc hand is a great example because merely checking back some QQ resulted in you paying off a full stack with ATcc because you wrongly assumed your opponent had no strong hands in his checkback range. To always Bet QQ is to reduce the EV of betting QQ! To always Check QQ is to reduce the EV of checking QQ. To get the percentages wrong is to reduce the EV of QQ!
FWIW, one of the two key changes to turn ATcc into a check instead of a Bet I made in the sim of your SB C 4B range was to remove AA from your range (to assume you instead shoved 100% AA which is bad strat). The other was to add a weaker hand so that it would be preferred as the bluff instead of ATcc because it had less SDV. The weaker overall range reduces your theoretical betting freq making ATcc a SDV check and the added weaker hand a replacement bluff. This reveals something about ATcc in your range when it is bet: it is being bet partly because it is one of the weakest hands in your range not so much because it has a GS and an overcard.
Nov. 11, 2016 | 1:01 a.m.
Collective wisdom is an extraordinary source of infos to deviate from the standard and start to explore new ways of thinking about the game. As Soto said, you'll be past by the game really soon if you don't try to find new ways of playing this game.
How do you outcompete regs if you follow the herd and merely do what other regs are reputed to do by collective wisdom? You don't. Your statement is contradictory: collective wisdom is old ways not new ways.
You outcompete by having an edge: by definition something you do that is different. This cannot come from collective wisdom as an edge and collective wisdom are mutually exclusive.
We should differentiate between the targets of collective "wisdom. There is collective wisdom about how to play against recreational players which for the most part is more reliable and then there is collective wisdom about how to play against regs which for the most part is a mixed bag of solid advice and plausible misinformation.
IMO, there are a lot more "secrets" and misinformation in the collective public "wisdom" about how to play against regs because inherent competition in the same player pool incentivises cliques to share "secrets" only with one another but spread misinformation more widely.
I contend that solvers (and a growing number of other tools) and disciplined experimental play followed by subsequent DB filters allow Hero to gain confidence in or to refute, in other words to distill collective "wisdom" from collective misinformation.
As a follow on, I believe that anyone offering advice should be called upon to provide an appropriate "proof" for that advice. Something that can be independently analysed and challenged for validity and generalisation.
Nov. 11, 2016 | 12:27 a.m.
A small suggestion for the sims:
I suggest multiplying potsize and starting stacks by 10 because solver only has resolution of 1 so betsizes eg 55%,66% on potsize of 5 causes rounding errors (both resolve to 3) where as on 50 is fine (28 & 33).
Nov. 10, 2016 | 10:47 p.m.
Why do your flop X/raise sizes differ (I think 60% on two tone board, 70% on 3 to straight+2-tone)? Is this to size the odds for calling opponent to the number of outs they most likely have for their draw?
For standard raise sizes, do you advocate 50ish% for dry boards, 60% for two tone, 70% for combo draws? How much for monotone no strt drw?
Have you validated these as Pio's preference or are they based on player experience or both?