I waffled back and forth on buying a beast of a machine and renting a server eventually deciding to continue renting a server. The nice thing about renting is that it's cheaper to upgrade when better specs become available and you don't pay electricity.
I think I have roughly double your ram, triple your storage and likely double to triple your computing power for roughly $250/month. Before electricity costs factored in that's roughly 2 years of use of your machine with cheaper upgrades to be made along the way. This was the real deciding factor for me (and that my electricity costs are very high)
Jan. 24, 2019 | 2:41 p.m.
Your high 3bet for the BB in your sim is, in part, due to the fact that you didn't allow the button to limp.
And before claiming a win on your bet, if I were on the other side of it, I'd want to have a look at the EV difference between a button capping strategy and non capping strategy. Even then I'd be suspicious of the results because of how bad the bucketing is for O8 in monker. (for example, there will be boards that a preflop sim will tell you to donk 30% on whereas when you run those same ranges on that board in a post flop solve the donk range will go to near 0)
You have to be very cautious about extracting any post flop strategy from your preflop sims in O8 in monker.
Jan. 24, 2019 | 2:35 p.m.
Thanks for all the good posts lately. I'm glad to be following you. And sorry I missed this post for so long.
I went ahead and ran some sims for this using what I believe to be accurate preflop ranges and depending on exact turn bet size, we should be folding in the neighborhood of 10-20% of our Tx on the turn if we work from the assumption he only has JTo-T6o, 50% of T8s and all of T7s-T2s as his Tx in preflop flat range.
Solver shows us folding some T9 and some T8 and a sliver of T7 though T9 has the worst EV of the three facing a turn bet, so preferring to fold T9 over T8 and T7 would be fine if we know villain won't figure out what we're doing and exploit us based on blockers.
Did you use the preflop ranges Zach used in his Pio sim? I think we will find a much tougher time finding a fold in that case.
April 24, 2017 | 9:37 p.m.
Fun video. I like checking hands vs Pio as the structure!
In general, I would be careful to draw specific conclusions about what is or isn't a mistake for Doug given that you've made a lot of assumptions about the availability of options in the game tree regarding bet size options and also some assumptions about his preflop ranges.
For example, I would think I could construct some reasonable assumptions and ranges where x/f w T9 on the QcTc2d6s double barrel isn't a mistake.
I think it would be interesting to make a video or two comparing how your strategy changes by changing bet size options and preflop ranges for villain in a couple of hands.
March 21, 2017 | 4:28 p.m.
Yeah. Perhaps it should have been something more like:
The death of the "Is this a +EV play?" and the birth of "Is this the highest EV play?"
As often times players were just asking "is this a +EV jam?" or "is this a +EV bet?" while just looking at their equity or fold equity or combination there of instead of considering that they need to compare that to the EV of all the alternatives.
Then the potshare concept is useful in comparing the EV of a shove or a bet to what your actual share of the pot should be. That is to know if something is likely the highest EV play by comparing it to your potshare.
Oct. 9, 2016 | 5:10 a.m.
I don't play much 6max so I did some work on this.
I think that if the BTN opens and peels wide preflop we should just bet everything at 33%.
If he's tight with his opens or his peels then your 9x are going to be about the same EV no matter what you do with them but they should probably mostly be checked (that is more than 50%.)
Can you lay out your 3B range as well as BTN's calling range pre?
Sept. 29, 2016 | 11:22 p.m.
I asked about this (or the inverse of this) in a prior video where I noticed Krab bet a middle or bottom pair.
IIRC he responded saying he's been selecting a some combos to do this with based on some solver work he'd been doing.
So the combos he has selected to be for mid/bottom pair probably just didn't appear in this video.
Sept. 29, 2016 | 4:15 a.m.
I'd probably x/c AA and then most of my Ax that can't go for three streets. Idk if you need a x/r range but it's certainly not a problem to have one. I'd prob just x/r mostly bluffs tbh. You can mix in a combo or 3 of 99 if you'd like but I don't think it's that important.
I still don't like betting pairs as bluffs. Sometimes it will get checked down and you will win. May as well bet some hands where that isn't the case. Seems better to want to bet hands that can turn more equity as well, which isn't really the case with 66-88.
If you appropriately protect you x/c's on each street I don't think you need to bet the 66-88 and 9x OTF. What is it about them that makes you want to bet?
Also, what are your sizings for preflop and on the flop?
Sept. 29, 2016 | 3:56 a.m.
I would "bluff" (as you've put it) with fewer pairs and instead bluff bet with way more unpaired hands. I'd add the anything with a bdsd first and see where that got me combo wise.
I'd probably have only one combo or a fraction of a combo of 9x I'm cbetting with but not sure what's best regarding that tbh.
Sept. 26, 2016 | 3:42 a.m.
@Zen regarding the LHE analogy.
Here's some context for you.
I doubt Polaris would have beaten Hoss at HULHE over a large sample, but he was the only HULHE specialist in that group. So it was a competition of the AI vs some good poker thinking opponents but not really the best at HULHE.
Compare this to the recent brains vs. AI match between Doug Polk et. al. and the Carnegie Melon HUNLHE AI. While human team won some bets the match was considered a draw from a statistical stand point. In this case the the team of human players was comprised of only HUNLHE specialists.
Sept. 5, 2016 | 9:58 p.m.
ram will determine the size of the trees you can build and solve. when i bought pio i also bought as much more ram as my computer could take and it was worth it. you may be able to replace all of your ram sticks and get even more, check your computer's capacity.
your cpu will just affect how quickly it will solve the trees. if you get pro version you can easily script what you'd like to run and let it do the computations over night or over a weekend so in this regard cpu wouldn't make any difference at all.
Aug. 31, 2016 | 6:19 p.m.
Aug. 9, 2016 | 3:43 a.m.
I'd bet 1/5th pot. Likely gets you paid by the same range that villain calls a shove with or more. Still gives him room to bluff. You can also have some bluffs here for this size if you have some AKQx hands in your 4b range.
Aug. 6, 2016 | 6:18 p.m.
Can you guys add a sort by date function after searching for posts. I was looking to comment on some hu posts yesterday so I searched for "hu" and all of the results at the top of the page were quite old and I couldn't sort for the most recent ones to make a usable contribution to the forum.
Aug. 6, 2016 | 4:22 a.m.
Obv you know the flop bet is very nonstandard, but you point out you sometimes play nonstandard. In doing so we assume you have a good reason(read) for doing so and expect villain to react very poorly to it so from this point the flop/turn play is solely dependent upon that feel you have.
Raising the river is hopeful (again you need a personal read that he's paying off lots worse or folding tons of chops for this size which are somewhat contradictory assumptions) as is calling it off. It's certainly not necessary to call it off from a range point of view, even if you check back most Ax combos on the flop.
k back flop, call or bet the turn, call the river without soul reads.
Aug. 6, 2016 | 4:19 a.m.
Do you have a flatting range?
The difference from 6max in the sb and from bb (both hu and 6max) for 3betting range construction is going to be different. By different I mean less polar and more linear.
If you're flatting some, you'll need to flat some traps as well. If you're not flatting just go straight linear perhaps only dropping out some off suit hands that will be dominated by his calling range since you're deep.
Aug. 5, 2016 | 8:51 p.m.
If it's a multiway pot and going to be a cluster* I'll usually insist to just run it once and get the game moving along again.
One interesting idea is to run it an odd number of times if you're accustomed to getting it it quite good. e.g. offer 1 or 3 times.
Aug. 5, 2016 | 4:11 a.m.
Pretty standard call call call here.
You also don't need to consider three betting this preflop. I'd just treat it as a sb vs bb situation with a little extra dead money in the pot.
In your position I wouldn't do much raising of the turn at all.
Aug. 5, 2016 | 3:57 a.m.
If you're looking for a practical approach to the problem there are some tools you can use to guide you.
I see you've mentioned Alex Sutherland. He's the guy behind Gtorb, right? I'm not too familiar with their software. I have played around with PioSolver a bit though.
With Pio there are subsets of flops that it can appropriately weight to "model" the full set of flops. You can run simulations on all those flops and then get an aggregate report and a preflop ordering from such. With this all preflop hands (in the possible range) will be ordered by their weighted-averaged ev across all the flops simulated (and if they've done their modeling correctly this should be close enough). You can then use their ordering for adding more or fewer hands to your ranges. (though iirc they may already include the weighted-averaged ev in the report so you can just pick all the +ev hands as your calling range.)
I could be misremembering something about the way this works as it's been a long time since I worked with it. But it works in some way along these lines.
For practical purposes, I'd say it's more important to play your range well (and have an easily implementable strategy) than it is to get the perfect range. Further, given that you're a smart player any range you're coming up with and solver is telling all hands in are +ev is probably quite good.
Aug. 4, 2016 | 3:43 a.m.
What I'm saying is that when you take a random collection of hands as your calling range, even if all of them turn out to be (relative) +EV calls, some of those hands may not be in the gto calling range.
Suppose, for instance, we selected the 10 best combos not in our 3bet range and then 7c2h as the only two hands in our calling range. I suspect that 72 is still going to be higher EV than folding as it's going to get to make up a bunch of equity by bluffing lots since your calling range is now fairly strong.
Aug. 4, 2016 | 3:14 a.m.
If you randomly select a calling range from the complement of your gto 3betting range and all of the hands in this random range have higher EV than folding preflop (a la your solver output) I do not believe you can conclude that they MUST be in the gto calling range.
Is this the kind of answer you're looking for?
I believe we could easily construct some examples to illustrate this. Just clarifying the question first.
Aug. 3, 2016 | 9:57 p.m.
14:00 you cbet Ts4s on AdTh8c and mention in it's a very, very small number of Tx you cbet on this board just to keep a non-zero amount of Tx in your cbet range.
I have 2 questions regarding this.
How imperative do you think it is that you have bare Tx in your cbet range given that you may cbet some 2pair combos containing Tx and also given that you'll have sets in your cbet range that also improve on Tx turns or rivers. (put another way, do you think you are missing out on significant or only minimal strat value if you don't cbet any Tx?)
Does the number of combos of bare Tx you want to cbet change based on board texture? e.g. if the flop is AT6r or AT5 where he will flop fewer draws to float (or even say some 9x on A94, e.g.) do you want to cbet still with some middle pair combos? And if so, is it fewer combos?