Another bright example for how "useful" the term "fish" is! :-(
What do you consider as fish?! Anybody that significantly deviates from "optimal" play is a "fish". That said, there are gazillion types of weak player types. Against some it might be +EV, against others, you'll bleed money. So, what does this question actually aims at?
Sorry for ranting. :-)
Sept. 21, 2020 | 9:15 a.m.
Do you have links to resources on the topic of multi-street bluffing?
I've made videos on that, but they are not free of charge (as far as I know).
Typically am going to call here having roughly 38% against his range.
I's about equity realization (EQR), namely the fraction of your EQ that you will realize postflop. Depends on stack depth, position, ranges. Being oop with an easily dominated (offsuit) holding your EQR might easily drop to 50 - 55%, which means, your 38% (hot-cold-) equity will shrink to 19%. Not sure about this special situation though, I'm on the jump, so just as a rough example.
Sept. 13, 2020 | 8:08 a.m.
First of all, if you are sure to correctly predict the next street, you can - or must! - combine these. Like here, and I 100% follow your assumption, you can treat the raise as an all-in-shove. Otherwise you will get burnt EV-wise (topic is "multistreet-bluffing" which is a pretty nice topic to explore as it shows how and why we can significantly increase our bluffing range with more streets to come).
Second, how to react entirely depends on your assessment. I'd rarely ever expect a "tight passive" to minraise w/ AK on a dry K722 board in a 3-bet pot. In position! Against the preflop aggressor - which still is perceived to have AA in his range.
Against a range of all AK, 77, 22 you get a merely break-even call on the turn!! That means, folding does not lose anything!
Now compare the chances that you make a mistake by folding, that means, his range is significantly wider. Somewhat likely? You know better - but doubtful, right?
What about calling (remember, we are break-even), how about the probability that he's tighter - i.e. only has AKs, or only 50% AK or even AA in his range? I'd say somewhat higher than 0%. That means, on average, calling is close to folding, folding will almost never be wrong, but calling might be a big mistake.
That means, a tight (almost crazy looking) fold might be the best alternative in this special scenario.
Sept. 11, 2020 | 7:01 a.m.
There are three "simple" tricks:
1) GTO -> Define your optimal defense. Means, consider your betting-range and then define the optimal defense-range. The defense part not necessarily is the best part of your range in terms of hand rankings, but the part with the highest EQ. For example, AJ is likely a better hand to defend than AA.
2) Pseudo-GTO -> You are talking about being close to a "break-even-call". What does that help? Breakeven is losing - which means you can as well fold. Now, work with different probabilities. For example, make three baskets: B1 is the likelihood for him having the nuts like always (EV = -130), B2 is the likelihood for having a break-even-call (mixture between nuts and very weak valuebets, EV = 0) and B3 is the likelihood for splashing around (EV = say 200).
We can ignore the middle part (BE-call) as it changes nothing. When we are on the bad end, we lose 130, when we are at the good end we win 200. The latter one is approximately 1.5x in value of the bad end. That means, the ratio between nuts and splash should be 1.5 : 1, in other words, the likelihood of him splasing should roughly be 2/3 or the likelihood of him nutting. In numbers, when you assign him a 40% probability of having the nuts, you need 25% splashing (and remaining 35% of having a break-even-call). And then AGAIN, you're just break-even, so better plan with a risk buffer).
3) Error margin -> a somewhat simplified version of 2): consider only two situations, one where you're right (profitable call) and the other where you're wrong (bad call). The bad situation will lead to a significant loss. The good situation will lead to a marginal win. Now, as you can't be 100% sure of either, think about an error margin. When you fold (and were wrong - in your assumption, not the single call), you lost SOME EV, but the probability of being wrong is likely not that big, which reduces the loss again, likely close to zero. When you call (and were right in your assumptions), it will likely be a very small win, which might get reduced by the error margin close to zero again.
Summarized: in a live game against an unknown, with the pool tendency to be more nutheavy in that situation, there's likely NO situation that makes folding a mistake. Like ever. So, just fold and gather more information before playing the hero.
Aug. 28, 2020 | 8:20 a.m.
I can't really follow your math ... if you are sure that Villain only does it with Ax with redraw, it's irrelevant if he had one combo or 100.
Our EQ is ~40% (20% we lose, 80% we tie for a total of 50% of the pot, which makes 40% EQ share total - given a HU pot). After we call, there's 780 in the pot, our share is 312. We invest 350, so we lose 38.
Aug. 28, 2020 | 7:54 a.m.
Who bets? BB after 3bet or BTN vs. cbet after calling 3bet?
Aug. 8, 2020 | 6 p.m.
The "general" recipe for deep database analysis is this:
1) Look for "potential" leaks. -> DONE.
2) Identify situations that lead to those stats (i.e. "fold vs. riverbet", "b/f river", "x/f river", "xb river" and so on).
3) Define the reference point.
4) Compare situations to reference point and prove, if potential leaks are true leaks.
3) needs explanation: say for example, we want to analyze the WTSD. We start with situations where we have a non-nuttish hand with SD-value (say medium pair or better, but less than TPTK or even 2-pair) and face a riverbet (no matter how it came to it). So we set all necessary filters (i.e. "saw river" + "faced river bet") and add a filter "folded vs. riverbet". Then we note the EV. Obviously it's negative (not zero!) as we already have invested money up to that point. That is our reference point. Now, we change the filter "folded" to "called" (or more broadly "not folded") and again look at the EV. If it's higher than our reference value we know that we give up too easily. Then we can go into single hands and continue our exploration.
That is a very simple example - we could (and should) go as much into detail as we can. For example, add "PFR", add "cbet made on flop", "checked turn", "OOP" and such - so we can really pinpoint the bleeding points.
But obviously - and here's the hook - that demands huge sample sizes! That is the reason why I regularly only accept students (asking for database analysis in such a structured way) with at least 150k hands. Otherwise it's just impossible to setup filters that really show the ground. If we only get three hands after setting the filters, we can't really use the results - because variation is too high.
Hope that helps?
Aug. 5, 2020 | 2:34 p.m.
What immediately pops out is the combination of WTSD (pretty low), W$SD (too low in combination with low WTSD), W$WSF and aggression factors. Does not really fit ... but might be a matter of sample size (in which case you were running somewhat badly ...).
Aug. 5, 2020 | 11:36 a.m.
I give the UTG open range some respect [...]
Not enough, obviously. ;-) Fold preflop. Nothing good comes from playing a highly dominated hand with little to no cooler potential. Three remaining active players add up to the negative list.
Rest is a subsequent fault.
Aug. 2, 2020 | 7:40 p.m.
So, it's 6-way and you got to call 9bb for a pot of 57bb (expecting BTN to call 100%)?
Seems close but doable. You'll likely double your stack when you hit set (and give up anytime else) - which makes for positive EV. Times were you lose against higher set are compensated by the (good times) where you even triple.
July 29, 2020 | 2:55 p.m.
ad 1) I'd suggest reading the books of Will Tipton:
They are great to understand the topic. Aside from that, coaching would be a convenient way, obv.
ad 2) CREV solver can solve flops and tell you what range to take what action with. It CANNOT handle 3-way scenarios though.
ad 3) At the very basic level, CREV-solver and Pio are identical (and will deliver the same result). On the more advanced level, there are differences: CREV shows the entire tree (visually), Pio doesn't. Pio is way faster, as it's optimized for the solving task, CREV solver is "just" an add-on-product. Furthermore, Pio can be scripted and be used to perform huge automated tasks, i.e. solving multiple scenarios and then create aggregated statistical reports.
July 26, 2020 | 1:01 p.m.
I repeat myself and I know that some might disagree: a solver is not a LEARNING tool!
Compare it with a kid, learning math in the first school year (my son is too old now to still take him as a reference :D). Does he learn with pocket calculator? No!! Because a pocket calculator does not teach or help to learn, it simply gives a naked result. And the kid had to believe it. If it wanted to know what 2+2 is - and accidentally typed in 2+3, the calc would say 5. And the kid had would "learn" that 2+2 were 5.
And Pio (like any solver) IS a pocket calculator. It's not a tool to learn about strategies. It's about solving scenarios and get "the" answer. Obviously you can play around, lock nodes (fix certain strategies), analyze ranges, vary betsizes and such - but it's for many reasons not the most efficient way to learn:
- solvers take time, especially for big trees (i.e. flop scenarios)
- solvers work with GTO ranges
- node locking is quite laborious
- results (with mixed ranges) are too complex to ingrain into human brains
Just look at the various posts around here where people getting desperate while trying to understand the solver result in a certain scenario.
All that said, when one of my coachees asks me for what the right way is to attach that field, I recommend the following:
- DEEPLY understand what GTO means (know all about optimal frequencies for bluffs and defense, valuebetting oop, multi-street-bluffing and son on).
- Work with CardrunnersEV to deeply understand ranges and impacts on EV when deviating from strategies (CREV has an implemented solver by the way!!).
- Use PioSolver (in combination with CREV!!) to learn about EQR, preflop ranges, exploitation etc.
That is the way I would recommend. :-)
July 26, 2020 | 7:19 a.m.
CREV might not be en vogue anymore, but as a study tool it's still superior to any other tool / solver. Just my 2 cents. ;-)
July 23, 2020 | 6:11 p.m.
CREV is the perfect tool to accomplish what you're looking for. :-) I will PM you.
July 20, 2020 | 8:07 p.m.
-Are there better tools for EQR? Can PIO provide EQR values for specific combos?
CardrunnersEV is - imho - the best tool to study EQR and the impact on range construction (once you got the numbers). Pio can help determining the EQR - by using the script functionality and then analyzing multiple flops to create aggregation reports.
-Are there good resources/videos on how to determine EQR for various hand combos preflop and postflop?
I've made videos on that - but they are not publicly available. I know, this is not helpful, but it's kind of a specific topic actually. :)
-Is there a way to do DB analysis to get actual EQR numbers from my sample of hands for specific combos?
For sure. You can filter for a specific combo, i.e. AJo, filter for a certain scenario, i.e. BTN call vs. SB 3-bet and then look at the result. Put it in relation to the "original hot/cold-EQ" against the presumed SB-3b-range - and you get the EQR.
Obviously that's quite an effort - and it demands a huge sample size - but technically it will work.
July 17, 2020 | 3:14 p.m.
Threads are just separate tasks to be calculated. The PC decides when and in which order processor power will be used for which thread. Regularly one thread will be handled by one core. So, if you got 8 cores and run only one thread your PC will getting bored. If you run 8 threads the PC might have the chance to distribute it evenly to the cores, but still it might happen that single threads are ready faster and then the PC again will not be fully utilized anymore. Hence 2x the times of the cores is a good rule-of-thumb.
Your PC can handle unlimited threads. :) There will be a number where it's getting inefficient for the program (!) to handle it, but this will be very, very high.
July 17, 2020 | 5:40 a.m.
Tool might be technically great, but it turns you into a robot, following AI hints. Question is, up to what stakes will that function? I'm the oldschool-guy that suggests deep learning, using tools as tools, not as guiding rods.
July 15, 2020 | 8:07 a.m.
May I reply? :-)
I do not really get your thought process behind the EV-formula, so I can't really tell you where or why you made the mistake. Here's the corrected formula (I did use the combo fractions instead of 0.35 to make more clear what the numbers mean and I used more brackets than necessary to exactly show the different terms):
EV (shove) = (11/31 x ((0,76 x -217) + (0,24 x 384))) + (20/31 x 167)
EV (shove) = 81.92
Here's the formula in words:
EV (shove) = (C% x ((VEQ x B) + (HEQ x [P+B]))) + (F% x P)
C% = percentage Villain calls
F% = percentage Villain folds (1-C%)
VEQ = Villain's EQ
HEQ = Hero's EQ (1-VEQ)
B = betsize (Hero's shove)
P = potsize
Formula for checking looks good.
July 9, 2020 | 5:55 p.m.
I still do not really understand just exactly how it would be bad to bet small and get called by a weak Ahigh on the flop and turn
It's not bad in isolation - but it's bad due to the fact that you don't bet against the A-high fraction of your opp's range, but you bet against the entire range. Once again - you need 50% EQ (against the calling-range) to bet for value (assuming that otherwise it would get checked down). That is because you want to get more money back (or at the very least the same) than you invest.
When you bet with 44% EQ, you only get back $0.88 for every $1 you invest (44% of $2 after bet and call). That means, you either need some FE (which you don't have against the given range with the small bet) or you want to avoid losing EQ by getting bluffed on later streets - which again is unlikely.
That is why I showed the extreme example with TT+, AhKs. It's exactly the same - only that you lose even more money by betting, but the principle remains the same. As long as you're an underdog, it doesn't matter how much - that simply impacts the amount you lose, but you'll lose (EV) in ANY case.
July 9, 2020 | 4:13 p.m.
0.9 is Villain's EQ with overpairs - 0.1 is our EQ against.
Against a 50/50 range with 50% overpairs and 50% overcards, you got 44% EQ overall. That said, below 50% equity and w/o any FE at all, ANY bet in position will lose money. It does not matter that against one part of Villain's range you could valuebet - if overall you just run into better hands.
Imagine you're against a range of TT+, AhKs. Still you're 75% ahead of ONE combo - but lose to 30 combos. Still worth betting? :)
July 9, 2020 | 2:16 p.m.
Whatever size you choose, xb has the highest EV (assuming V. calls overpairs, folds AK and ignoring our marginal EQ when called):
EV(shove) = (0.5 x -100) + (0.5 x 90) = -5
EV(minbet) = (0.5 x -20) + (0.5 x 90) = +35
EV(check) = (0.5 x 0.25 x 0) + (0.5 x 0.75 x 90) + (0.5 x 0.9 x 0) + (0.5 x 0.1 x 90) = +38.25
So, when we take +38.25 as a baseline, we can calculate the break-even-point for betsize - assuming that Villain will ALWAYS fold AK to our bet:
(0.5 x -Z) + (0.5 x 90) = +38.25
-0.5Z + 45 = 38.25
-0.5Z = -6.75
Z = 13.5
That means, if you could bet 13.5 and Villain folds half of his range that you beat, the EQ denial would compensate for the loss. Anything > 13.5 is worse than checking.
July 9, 2020 | noon
Flop play is okay, but turn is unnecessary. With "overcard" you mean "overpairs"? Unlikely. It's less than PSB left, Villain won't fold an overpair. Same for NFD (incl. gutshot), I'd prefer to just xb and rely on your SD-value, along with some additional EQ.