Applications of No-Limit Hold em

Have you tried running it through a solver? I haven't, but just from my experience of lolSnowie, and seeing a few solutions from GTOrb/Pio, I believe that we should rarely be c-betting >50% of the time when we're OOP, especially with "standard sizing" of half pot or more. This board might be one where we can bet at a higher frequency than average (I doubt it, as we have no frontdoor draws to balance our value hands), but only if we choose a very small sizing.
Range advantage doesn't automatically mean "bet more often", and I think the 'positional situation' is a bigger factor. GTO poker apparently features a LOT of checking when OOP as the PFR, and whereas newbies are told "Always c-bet the dry Axx flops" (because of the fold equity), GTO bots/solvers appear to bet the driest boards at a lower frequency than one might expect.

Some results I got:

I have been running pio sims to study this chapter. The results I got using a UTG/MP open of 20.5% and a BTN CC range of 9.1% was for MP/UTG to c-bet this flop 12% of the time.

The only results I got with a c-bet above 50% was when I studied SB vs BB SRP

Using a sb steal range of 39.97% and a BB call range of 63.27% I got small blind c-bet this flop 61.83% of the time.


All sims were ran using 8bb pot on flop with 75% psb and whatever else to match the example on page 281. Next I will use some of the frequencies I got to study the equation on page 282.


Another thing I can do is lock a c-bet of around %50-%60 on that flop for the UTG/MP vs BTN example. Because that is exactly what most players do. I can then find how to exploit that using a solver. Or the chapter.

I mean its possible ranges for optimal play are close enough that a positional advantage trumps a small range advantage, but from what I recall of seeing solutions for toy games inherent positional advantage is small. Position is very powerful when its you in it and you're facing a lesser player, but it shrinks considerably vs tough players, let alone optimal ones. But anyway my 80% number wasn't specifically referring to optimal play, more just of guess in general as to what our overall cbet% should be on that board vs typical players.

We can break it down by what type of player we might be up against if we raise around 20% and they flat from from the button:

1) tough/taggy opponent (7%-15% calling range):
kinda close but, since top 7% usually gets 3bet (yes there are still some strong flats here and there), the OR's range should be significantly stronger and highly suspect its correct to cbet more than 50% on dry flops.

2) average opponent (5%-25% calling range):
Our range is just significantly stronger than 1) so cbetting starts to become obvious

3) fishy opponent (5%-25% or more and plays ABC poker):
Our range is stronger but more importantly we're facing a lot less floats and bluff-raises so a simple cbet just becomes mandatory

Its entirely possible that vs a truly optimal player we should be cbetting less than 50% and maybe today's "solvers" are hinting at that, but its absolute crime if a new player reads that chapter, concludes that they shouldn't be cbetting dry boards, and then applies that to their game in real life.

There is a passage in the preface that warns new players of the strategies taught in the book. The book is about optimal play. It is not a book about how to play low stakes games. That said, after reading those 5 pages it should be clear how to destroy someone who c-bets that board OOP 80%.

Nah cbetting dry boards isn't just for low stakes its for any stakes games. Anyone with any experience in mid/high games knows it, but again, don't take my word for it, ask in any of the NL forums.

If the author wants to go full GTO that's fine, but then there should at least be an actual solution somewhere, otherwise its just guesswork.

I suggest rereading the section of the book posted. The entire section and not just the page.

When you come up with a point, I'll respond to it.

The checking range being too weak is basically nonsense as long as you defend it and do some exploit stuffs.

I know for a fact that this situation is awful for me vs. regs. Sometimes it is so hard to check/fold instead of bet/fold even when you know it is right.

Here is a Sutherland article about a similar spot.
UTG vs BTN AT3r
He solved for a UTG c-bet of 15%

http://blog.gtorangebuilder.com/2016...rning.html?m=1

Probably the single best way for bad regs to increase their winrates would be to stop c-betting so often when they are OOP. It's not 2012 any more.

Here are some of my tests. I am not done yet, but if you look through the org mode file, you will get some good info.

http://pastebin.com/vEdjdNiZ

This stuff makes me want to quit poker all together lol.

I mean its interesting stuff, but I feel there should be more focus on what the initial ranges are (or should be) if we act like the solutions should be applied to actual games.

I'd be curious as to what the results are for more realistic (looser) ranges.

You probably only saw the snowie study
Snowies CC ranges are so tight.
LOL! at Snowie...
Snowie is alright though.

Looser is there, but i really only have UTG vs IP CC done, so ranges are not very loose.

The entire study has a lot to do with just that. But i am not done yet.
If you check for BTN vs UTG ( i have done the most here so far) the btn cc is 9.10% (from a Sutherland video, best CC range I have found, he even teaches you how to derive it) and i have ran that against an utg range of 20.5 , 16.5 and extreme nit of 3.77%

I already can see the change in strategy from the 3.7% range to the 20.5% range. Any tighter or looser it should be clear what is happening.

Next I will vary the BTN cc range from tighter to looser....
I do this for every posistion. Run it on the same 74 flop weighted subset...
I also test different stack sizes.
About anything you can think of I vary. I usually start with extreme cases like above.

But yeah, The whole point is exactly that. Is varying one thing at a time.
I left out ranges I get in the games I play, studying population tendencies. That is my personal crap. Also locking strats...
Like locking OOP c-bet of 80%

here is the video of how the BTN cold calling range was developed:
http://blog.gtorangebuilder.com/2015...ns-and-ev.html
Awesome Video!!!!

Arty, I use pokersnowie and recall you being active in the snowie thread. What turned you off from it? I want to know what I need to be wary of when using snowie.

Snowie is alright. Just be wary of everything. Always question why it does something. I don't use it though. I would if someone coded something to view each nodes actions in a 13x13 grid where each hands strategy is color coded and sized by frequency. (Like the usual CFR solution) It would be fun comparing its neural networks 'solution' to a CFR solution.

I bet you can learn a lot from snowie but I don't really know cause I only used the trial and the pie charts and lists freaked me out. I use the preflop advisor all the time though.

Snowie knows the answer to this thread so I guess that is good.

Snowie is fun to play heads up with. It does a lot of weird **** when short stacked though. Especially at like 9BBs I've seen it open 3x. Very unusual sizing.

I think they should add a shove feature might make the calling ranges a little looser preflop which would be good.

Snowie is a strong AI, but meh though. MEH!

Nothing really turned me off. I really like how quick it is (instantaneous!) at providing results that gave me a general idea of what the ranges and strategies look like in any scenario that I threw at it. (I don't have time - or the computer power - to run Pio/GTORB overnight just to look at one solution, and I'm not particularly concerned with finding "perfect" results. I just wanted to learn the answers to some general thematic questions, like "Which spots are good for overbets?" or "What kind of boards should be check-raised (or check-folded) at a high frequency?"
The main reason I didn't renew my subscription (after studying cashgames for two years) is that I switched to playing tournaments. Snowie hasn't got a clue how to play a 6bb stack in a satellite, so it was pointless for me to continue paying for the sub, when ICMizer (or similar) is the tool I need now. I might re-sub in the future if I start grinding cash again.

Wow, this seems so counter intuitive. If you only cbet AT3r 15% what about really wet flops that hit flatting ranges hard like 987ss?

In theory, check about 98% of the time and use mixed frequencies for the few combos you bet. In practical terms, you can just check your entire range on the flops where the IP player's range has more than 50% equity and where you have almost no fold equity.
It's a lot easier to design continuance ranges if you eliminate bet-folds, bet-calls and bet-raises from your strat. On those "tricky" boards, you can cut your decision tree in half by always starting with a check.

EDIT: Just look at the OOP checking frequencies for these flops in Outfit's analysis:

My notes are really confusing if you do not use emacs as an editor with org mode.

The spreadsheet above are the hands that are checked the most.

The spot is UTG (snowie 16.52%) vs BTN (Sutherland 9.10%)
I ran a sim over a 74 flop 'weighted' subset, a pretty good estimation of all 1755 strategically different flops. Snowie's UTG range checked an average of 78.03% of the time over this subset.

Here is the average for the subset:
http://pastebin.com/C9TCXMg5

Here are the flops with the largest 60% PSB (Bet of 51) frequency:
http://pastebin.com/Rnvu5jpT

Here are the flops with the largest 33% PSB (Bet of 28) frequency:
http://pastebin.com/ZTL7vrXW

As you already saw, here are the flops where you check the most:
http://pastebin.com/1meP2TaZ

Next I analyzed what happens when you check and are facing a 33% PSB
http://pastebin.com/NScNSEL4
note: notice you defend about 64% of the time, you raise (check raise) about 21% of the time and you check call about 43% of the time

The types of flops you are most likely to check raise on:
http://pastebin.com/JLmwTasQ

The types of flops where you raise the least:
http://pastebin.com/jz4bjvfn
note: I am not sure why I did this. this is not very interesting without further analysis, do you call alot? do you fold a lot... check raise the least is kinda lame analysis... whatever.

For snowie users:
The UTG vs CO SRP analysis was done with snowies UTG range (16.52%) and snowies CO cold call range (2.97%). It would be interesting for someone that uses snowie to compare how snowies ranges perform in equilibrium to how they perform while you are using snowie.
The frequencies are here for those interested.
I used a 49 flop subset and snowie ranges:
UTG vs CO SRP
http://pastebin.com/xpS5Z55u

A spot where you can pretty much check your entire range OOP is BB vs BTN SRP.
I ran a 74 flop subset where BB can donk bet the flop and one where BB does not have the option. The EV for BB was almost identical with or without the donk bet. That was a waste of electricity. Donk betting in that spot is lame. I like doing it though cause I think it is funny. There were some flops where it was done though, but like I said even on the ones where it was done checking did not change BB's EV an amount worth investigating.

One more thing. If the population or player folds to c-bets at a greater frequency than what the equilibrium strategy calls for, then c-betting at a greater frequency will maximize ev. So, c-bet away... Just know when doing so you open yourself up to being exploited. Ya know, obviously if someone folds all the time you bet...

I study equilibrium strategies so I know when someone deviates from them. I would not use an equilibrium strategy unless that was the optimal choice.