ZenFish's avatar

ZenFish

4310 points

Sept. 24, 2017 | 9:03 p.m.

Sept. 21, 2017 | 8:30 p.m.

I'm working a lot with MonkerSolver at the moment and I want somebody to discuss with, dammit. So I've started a MonkerSolver Skype group for those who use the program and for those who are curious about what it can do:

The MonkerSolver Hangout

Nothing formal, and not a study group, just a place for discussing technicalities and multiway strategy.

Sept. 21, 2017 | 2:36 p.m.

If you want to exploit the pool hard, you have to make yourself exploitable in the process. If your line is very profitable, and the pool doesn't counter you hard, keep doing it. Those that go after you on the river can be marked as special cases and handled separately.

Sept. 16, 2017 | 7:50 p.m.

Comment | ZenFish commented on Snowie Preflop ranges

Hey guys. I just found the Snowie Preflop ranges, and I want to know if anybody had success with learning from Snowie's preflop tendencies? I can see, Snowie doesn't like opening SC's from EP as an example.

What does the Snowie EP range look like?

Sept. 13, 2017 | 5:28 p.m.

Comment | ZenFish commented on river shove too thin?

When we are overbetting with a medium strength hand in a 3-way pot, I don't think we'll get looked up by worse often enough.

Sept. 11, 2017 | 3:43 p.m.

Comment | ZenFish commented on river shove too thin?

Are you bluffing or shoving for value?

Sept. 11, 2017 | 1:53 p.m.

Comment | ZenFish commented on poker solvers

Zer0:

I am trying to decide between buying MonkerSolver and Pio. Which would you recommend for NLHE?

It depends a lot on whether or not you're interested in studying multiway play. If you're content studying only HU situations, Pio will serve you perfectly. If you would like the option to work on multiway strategy, MonkerSolver is the only option on the market right now.

You can't go wrong with Pio, and there's a lot of information out there about how to use it well. But MonkerSolver is unique in the things it can solve. Pio will release a PLO solver soon, but multiway solving is not planned.

What kind of computer specs would be required to calculate preflop ranges for 6-max NLHE with MonkerSolver?

Approximately same as with Pio, but you will need more RAM to do the heaviest solves. To solve the 4-handed pre flop game (CO v BTN v SB v BB) with decent accuracy, You'll want a 240 GB RAM setup, same as for solving pre flop HU play with Pio, using a big flop set.

To solve for 6-handed ranges with good accuracy, you will need at least 500GB RAM.

Does MonkerSolver have all the necessary features to study the game?

It solves NLHE, PLO, PLO8, pre flop and postflop, HU and multiway. It lacks one very nifty Pio tool, aggregation reports.

Sept. 6, 2017 | 11:22 a.m.

Comment | ZenFish commented on poker solvers

PLOCalc is an advanced equity calculator/range tool, but not a solver (yet). From what I understand, the upcoming PLO solver will use the PLOCalc interface.

Sept. 4, 2017 | 1:20 p.m.

Comment | ZenFish commented on poker solvers

MonkerSolver is the most versatile solver out there right now. It solves pre flop and post flop, multiway and HU, and NLHE/PLO/PLO8. It's around half the price of Pio Edge, and a little more expensive than Pio Pro (which is the same as Pio Edge, minus pre flop solving).

Pio is the standard for HU solving pre flop and post flop, but after working a lot with Monker, I think Monker is overall more useful.

Also, check out CREV for a cheap user friendly solver. They have just renamed it GTO+ btw.

Sept. 4, 2017 | 12:52 p.m.

i see where your coming from but wouldnt that mean id seriously have to tighten my ranges
ie utg raise i cant play 3b or fold with my entire calling range there

You need to construct a linear 3BF range that contains the best, but not all, of your 3B + flat VPIP range.

Sept. 3, 2017 | 10:54 p.m.

You can improve your pre flop strategy (+ make it simpler) by playing 3B-or-fold from all positions except BTN and BB in high rake games. And even on BTN, there's not much to gain from having a flatting range with this rake.

Sept. 1, 2017 | 3:34 p.m.

The comments above are valid for all pre flop solvers, but there is an additional issue with MonkerSolver:

It does not measure how far from equilibrium we actually are, and the solver does not yet have any convergence criteria implemented.

So we have to be pragmatic. If we trust that the MonkerSolver algorithm produces strong multiway strategy (and as far as I know, the Poker Research Group from the University of Alberta assures us that it does), we can figure out good convergence criteria from experience.

From what I have seen, the deeper into the tree we go, the slower ranges converge. Open and 3B ranges are fast, multiway flatting ranges go slower and slower the more players are involved. For example, have not yet seen an open + call + call + call branch converge in reasonable time, even when running the sim 2x longer than the time it took for all the HU and 3-way branches to converge.

So my pragmatic stance on MonkerSolving at the moment is to run the simulations until I have clear convergence for all the ranges I'm interested in.

--

For those who are curious about the quality of the MonkerSolver pre flop simulations, I can report that I have benchmarked it against Pio for the SB v BB equilibrium and it reproduced an accurate Pio simulation (184 flops + tight convergence) very well when I cranked up the accuracy settings a bit.

We have nothing to benchmark it against for multiway simulations, but it's encouraging that the algorithm gets the HU case totally right.

Aug. 31, 2017 | 9:24 p.m.

Feel free to correct, but as I understand pre-flop trees include errors in the turn actions and worse errors in the river actions; i.e. checking nut flush back and other such things and other less egregious clear mistakes. In what proportion I don't know. And without Nash distance we can't know how bad the aggregated impact is, but I seriously doubt all hands are affected equally.

Preflop strategy crunching is concerned with preflop EV and frequencies. We don't need the post flop tree to be perfectly solved in order to arrive at near-perfect pre flop ranges.

Two reasons for this:

1) Preflop EVs are not very sensitive to variations in post flop strategy

For those who have used Pio, this is similar to the observation that flop EV is not very sensitive to flop bet sizing. In the sense that when we change to a different sizing, the solver chooses a different range for it, our opponent adjusts to this, and we end up with a flop EV that is almost the same.

The relationship between pre flop EVs and post flop strategy is like that. For example, if you run a pre flop solve in Pio and build a complex post flop game tree to get it perfect, you will not get much different results than from a simple tree with one size per street (but you will blow up the size of the simulation tremendously).

2) Error cancellation is our friend

For a simulation where we have myriads of error sources in the model, many of them will cancel each other, and the results we are after (pre flop EVs and frequencies) can get very accurate, even if we can't solve to infinite precision and we still have many small errors left in the post flop part of the tree.

--

It's important to distinguish between a pre flop simulation and the corresponding post flop game tree. Usually we're not interested in keeping the post flop part of the simulation since it will be rather crude. This is because we need to keep the post flop tree simple to keep the size of the pre flop simulation down, and we will often not solve to the point where all the post flop strategy (like river play) is well converged.

But for reasons 1) and 2) outlined above, this does not matter all that much for computing accurate pre flop strategy.

So we extract the pre flop strategy from the simulation, we save it, and for future post flop analysis we do separate post flop simulations using these ranges as input.

Aug. 31, 2017 | 9:19 p.m.

I think this is one of those moments where there won't be much future regret about throwing away some work now and redoing things.

The PLO solvers have just arrived on the scene, and they will change the way the game is studied (hopefully in a smart way). The PJ ranges will become standard tools for those who use the program, so why not use a solver to make them as good as they can be.

That way equilibrium pre flop ranges will become available also for those who can't afford to run a solver in the cloud (crunching PLO ranges is damn hard), and it should add good value to PJ.

Aug. 31, 2017 | 4:05 p.m.

1) Nick has been vetted as a coach by RIO, Jonna102 has not

2) Nick has always been open about his mediocre "MadAgenda" results, Jonna102 has been working very hard to hide the fact that he's always been a losing micro rec.

3) For both cases, price is irrelevant, since it's supply and demand, and afaik, both have offered a money-back option for dissatisfied customers.

What is on trial in this thread is integrity, not PokerStars results or price of packages. Jonna102 has misled his customers, Nick has not.

Aug. 31, 2017 | 1:26 p.m.

From the Pio blog:

Our first Omaha tool - PLOCalc

Today we are happy to release our first Omaha tool: PLOCalc. PLOCalc is an equity calculator and range explorer for PLO. What is novel about it is that it calculates range vs range equity which included equity for every single hand in the range. This works on preflop, flop, turn and river. It's done without any approximations - the results are exact.

About the solver:

PLOCalc is a fantastic tool for analyzing your PLO game. We anticipate the natural follow-up question though: "What about the solver?". Short answer is: PLO solver is going to happen in the near future although it will require very powerful hardware to run.

The solver and PLOCalc are going to use the same interface for showing and analyzing ranges. PLOCalc might be more useful for some real life scenarios as it allows you to analyze ranges based on reads and assumptions while GTO solutions are going to be very hard to make sense of in the game as complicated as PLO. The future of those tools is going to be influenced by the feedback we got from first batch of our users.

Aug. 31, 2017 | 11:03 a.m.

It's as easy as setting up the AWS account. Once you're in, you will have a virtual Windows desktop that you handle just like any other Windows machine. Use Dropbox to move files in/out.

How do I create and activate a new Amazon Web Services account?

Aug. 25, 2017 | 1:45 p.m.

Here we just have to use experience. Pick a few flops and experiment. Does picking the most used bet size always give higher EV than the other sizes for the single-size calculation? When all sizes are reasonable, I'm assuming this will be the case, but I haven't tested that assumption and we can't prove it with theory. So go and see.

You should probably do rake-free simulations so that we don't get noise from rake effects.

Aug. 3, 2017 | 12:44 p.m.

If Y>X,Does X situation is really GTO?

Any solution that you compute will be an approximation to GTO (if GTO is defined as the solution to the game where both players can use all possible options). The more options you give the players, the closer you will get.

Aug. 2, 2017 | 7:36 p.m.

Post | ZenFish posted in Chatter: Is Pio releasing a PLO solver?

From the 2+2 Pio support thread (some question about an upcoming update is answered and developer Piotrek hints at something):

Anyone know anything about it?

July 26, 2017 | 5:48 p.m.

Small Stakes PLO Research Blog - Day 653
Fold equity still remarkably low. How very odd.

July 26, 2017 | 2:50 p.m.

For inspiration:

Interview with limitless on Pokerstrategy.com (Oct 2017)
limitless on Joey's Podcast Jan (2017)

Limitless also started playing 4 years ago and built his way to Stars high stakes, starting out with free rolls. Then $2nl, playing $100nl after a year, $5-$10 after 2 years and high stakes after 3 years.

So during the time OP was struggling to get out of the micros, this guy started at the very bottom and worked his way to the very top. And for the first couple of years he was a professional handball player and had to find time for poker in between daily training.

What you describe in this thread is not happening (at least not yet):

Poker is crumbling due to GTO and finite skill gap.

Psychological research also suggests that the human species has tremendous potential for learning and growth, but it needs to be done in the right way (read Peak by Swedish researcher Anders Ericsson for more about that). So you have probably not reached your skill ceiling either.

If I may speculate, I believe you have been working in unproductive ways for the last 4 years and that you have hit the ceiling for what your current method can achieve. If that's true, you should change your method or quit.

Declaring failure may hurt your ego less than admitting that you have been wasting time working in wrong way. Or maybe you have too rigid ideas about what good poker is all about, so that it will be uncomfortable to start over in a new direction.

Quitting is maybe the right thing to do, or maybe not. But you should not blame it on games being too tough, or that the community is approaching a skill ceiling in the near future. Nothing points to this, quite the opposite.

Just an example of things happening:

Recently a new solver appeared (Monkersolver). It can solve multiway scenarios and PLO as well. The smart and driven now go "Multiway NLHE/PLO solver, you say? Hmm ..." and get to work with it. Then there's those who ignore it and continue to bitch and moan that the games are getting tougher by the day.

Which group do you think is largest? Can you see that the advanced software tools are not killing the games, but creating tremendous opportunities for those who can use them best? The average poker player isn't using them much (and if he is, he isn't using them well) and he never will.

(Sure, everybody and their grandmother is using Pio, but running sims and staring at grids is not how you learn with Pio. I conjecture that the average Pio license holder is using the program at something like 20% of capacity).

July 18, 2017 | 11:11 a.m.

That's the technique used by the University of Alberta group that solved HU LHE. Before finally solving the game, they made a HU bot that could beat elite human players already in 2008 (the "Polaris" bot), based on bucketing similar hands.

The Alberta group have published their techniques/algorithms in peer-reviewed journals, so it's proper science.

July 18, 2017 | 9:03 a.m.

In this light, I find this thread very odd

Do you think it's odd that people question the credibility of someone who sells coaching packages for $10k without posting a convincing lifetime graph?

It's standard procedure to inform potential customers about your lifetime results. Jnandez has been a professional for years, has a crushing win rate at $2000PLO (double-digit win rate in 2017!) and sells his product for a fraction of your price. So you should expect some scrutiny when you put a product on the market for $10k.

I'm not actually in poker for money. I'm in it for the love of the game, and because I enjoy the intellectual challenge.
All I can say is that I've never pretended to be anything I'm not,

You're not claiming to be a successful mid/high stakes player, but you're not mentioning anywhere in your marketing that you're a recreational micro/small stakes player either.

When you're not upfront about what stakes you've had success at and what you have accomplished lifetime, it's only natural that people ask questions when you enter the market with high-priced products.

The value of a coaching pack doesn't have to be strongly correlated with the performance of the coach (especially when the pack is theoretical), but transparency is still necessary. Otherwise you get threads like this, which is totally fair.

(Aside from also being against the rules of conduct of runitonce.com.)

You just made that up.

July 16, 2017 | 10:58 a.m.

You can try out a lower version and upgrade later (paying the difference). Preflop solving is a rabbit hole of sorts, but it can be very illuminating to crunch ranges and see what is what. You would want to rent a cloud server for that.

There's a new software out called Monkersolver. It uses some abstraction techniques (whatever that means) to solve multiplayer games (pre flop and post flop). So if you want to see how the BTN/SB/BB pre flop game should look, you can produce solutions there. You can solve the full 6-player pre flop game as well, but I doubt that is practical at the moment (would mean a humongous cloud computing job).

Not sure how accurate Monkersolver's solutions are, or how demanding the multiway sims are, but I'm keeping an eye on the program to see how it develops. They sell simulations too and they are not very expensive.

July 15, 2017 | 11:18 p.m.

TOC or it didn't happen!

July 14, 2017 | 10:18 a.m.

July 13, 2017 | 10:25 p.m.

Comment | ZenFish commented on 4b pot with weak TP

Did a quick sim with Hero on an optimal 4B range vs SB on optimal 4B calling range after playing 3B-or-fold vs your open.

Assuming SB doesn't donk anywhere:

  • CB range on on flop (if you use 1/3 size)
  • If called, check back range on turn
  • Then call 2P+ if Villain jams river. You'll then call ~60% of your range.

When you check flop and Villain leads turn, it becomes:

  • Call turn with AQ+, around half of your worse TP, and all flush draws
  • Call river jam with 2P+. You'll then call 76% of your range.

The 4B range used here is heavy in KK+,AK with small chunks of AJs-A2s/KJs-K9s as bluffs. So we hit this board rather hard. Interesting that we can CB range, but when we get called, we maximize EV by checking range on turn.

I suppose that is because we don't need to bet turn to get stacks in (only one moderate bet left), and solver prefers to delay all turn/river action to the river and get the money in there.

July 7, 2017 | 12:08 p.m.

Comment | ZenFish commented on ExPro comeback.

Lots of new software for you to explore. Should be interesting for you, since I bet you had many questions back in the day that were impossible to answer with the tools we had then. These days we have solvers.

Here's a very good Sauce video that demonstrates practical solver use:

Responding to Various C-Betting Strategies

July 4, 2017 | 3:52 p.m.

Load more