Welcome.

Take a tour. Enjoy some free sample content.

How it works

Free Video: CLP Video No. 287: Home Game Bart Reviews His Splashy At $1-$3 Deep Part 2

Free Podcast: CLP Podcast No. 54: Time Warp And Turn Value
New to Crush Live Poker?

Properly Using a Solver to Review a Hand

dpbuckdpbuck Posts: 1,977SubscriberProfessional
One of my goals for 2019 is to get a better understanding of all the tools and technologies available for wannabe high-stakes crushers, and really figure out how to use them to improve my game. One of the primary tools I dabbled with in 2018 was GTO+ (a poor man's Pio). I know I am not getting the most out of it, and am seeking to start a discussion with others on how to utilize these solvers to further my game.

To start, I've documented how I am currently using GTO+ to analyze a hand and make conclusions I can apply to similar situations. I took a very simple hand I played last weekend against a weak and exploitable opponent and walked step by step through the process I'm currently using. I'm interested to see what others think of my process, and potentially point out errors in the way I'm using the software, as well as the conclusions I am drawing.

The walkthrough ended up being longer than expected, so I put it into Evernote and will just paste the link here. I'll also share my GTO+ files in case others are using GTO+ and want to load them up.

I'm hoping to really learn HOW to study this year, and I figure this is as good a first step. Anybody interested in reviewing this and letting me know your thoughts?


Process Review Link: https://www.evernote.com/l/AIADIdOQjodG_4jwqak4_bxg7CkKysZobfs/

GTO+ Files:
Optimal: https://www.dropbox.com/s/r49ql4xba2382k6/01 12 19 - QJdd optimal.gto?dl=0
Exploitative: https://www.dropbox.com/s/8euhl7je2t0edqu/01 12 19 - QJdd exploitative.gto?dl=0
Tagged:

Comments

  • thnkpositive Posts: 23SubscriberProfessional
    I'd like to type a longer comment and will come back to this.
    I think you're using the software overall very well, but I would caution with this conclusion in particular.
    "Once we've potted the turn, though, we've run out of thick value on brick rivers. If villain will bluff frequently, middle-strength top pair hands are perfect bluff catchers. The strongest top pairs should still be bet for value."
    These solvers are not meant to calculate maximally exploitative lines. So if you're using them to try and guesstimate maximally exploitative lines you could quickly start suffering from garbage in garbage out.

    A couple of things that I would do
    1) use the solver to calculate the GTO strategy vs a good villain so you know what it looks like
    2) try some primitive node locking but avoid going overboard on this, because this is where the real garbage in garbage out danger lies
    3) analyze the hand for maximally exploitative lines in an old school way. Plugging in villains range estimation to the best of your ability on poker cruncher, and just calculating if you have enough equity to make the bluff catch, and also guesstimate if check calling river there or check folding or just outright betting has the higher ev.

    There's a real danger when using solvers to start to discount the value of your human intuition. Don't do it, your brain is basically a poker computer specifically designed to calculate max exploit strategies. The solver is just for fine tuning what "GTO" looks like so you can improve your intuition even more.

    Also, I would always recommend you analyze the crap out of river scenarios. It works just like golf or chess, you always start with putting and/or end games. Rivers carry the highest reward per hour spent since the pots are so big. It's also the easiest street to get right when you're calculating everything in your head at the table.
  • dpbuckdpbuck Posts: 1,977SubscriberProfessional
    @thnkpositive Thanks for taking the time for your response. I was really hoping to spark some sort of discussion, and your feedback is much appreciated.

    As for a couple of your points:
    thnkpositive said:

    I think you're using the software overall very well, but I would caution with this conclusion in particular.
    "Once we've potted the turn, though, we've run out of thick value on brick rivers. If villain will bluff frequently, middle-strength top pair hands are perfect bluff catchers. The strongest top pairs should still be bet for value."
    These solvers are not meant to calculate maximally exploitative lines. So if you're using them to try and guesstimate maximally exploitative lines you could quickly start suffering from garbage in garbage out.
    I readily admit my actual conclusion in this hand may very well be wrong. This is me attempting to interpret the output that indicates that I am checking QdJd on the river against this villain in this runout based on the conditions I gave him. Yes, I definitely run into garbage in/garbage out problems. However, the best anyone can do is guesstimate opponents range, and what he does with that range.

    Realistically, isn't this what a solver is supposed to do? I mean, I indicated what I thought villain's range would be preflop, and then what he would do with different parts of his range depending on different lines through the hand. That yielded this result. Aside from whether my assumptions on villain are correct or not, this is an optimal line against this "garbage in", correct? Or am I completely missing the boat on the point of putting in his range, node locking, etc?
    thnkpositive said:

    A couple of things that I would do
    1) use the solver to calculate the GTO strategy vs a good villain so you know what it looks like
    2) try some primitive node locking but avoid going overboard on this, because this is where the real garbage in garbage out danger lies
    3) analyze the hand for maximally exploitative lines in an old school way. Plugging in villains range estimation to the best of your ability on poker cruncher, and just calculating if you have enough equity to make the bluff catch, and also guesstimate if check calling river there or check folding or just outright betting has the higher ev.
    I agree 100% percent with these, and I started with #1. Effectively I took a three step process where I looked at a proper strategy against a villain playing a proper range, then a proper strategy against a villain playing this wide range, and then finally a proper strategy against a villain playing a wide range in an unbalanced fashion.

    I am very green with node locking, and likely made some mistakes, but we almost have to go "overboard" in node locking against a villain that is playing miles from optimal. Realistically, the better the villain, the less node locking required. Maybe this was a bad example hand to use because villain was so bad and playing so far from proper poker.
    thnkpositive said:

    There's a real danger when using solvers to start to discount the value of your human intuition. Don't do it, your brain is basically a poker computer specifically designed to calculate max exploit strategies. The solver is just for fine tuning what "GTO" looks like so you can improve your intuition even more.
    That is my ultimate goal. In essence I'm trying to doublecheck my intuition. Instinctively, I knew to bet small on the flop. This confirmed it. Instinctively, I thought I needed to continue to bet small on the turn. This showed that that is an option, but I should have also considered potting/overbetting the turn, which my intuition didn't even consider. Now the goal is to figure out why that should be considered, and in what future situations I can utilize this.

    Hopefully that makes sense. One thing I didn't do through the process is look at my current EV and Equity street by street as I made different modifications to my bet sizing and his reactions to them. I think that was a piece of information that would have been helpful as I tried to draw rough conclusions.

    Hopefully that all makes sense. Anybody else want to chime in?
  • Superfly Posts: 145Subscriber
    Hi Dana, I don’t currently have a solver or much experience using them. But I’d be very interested in learning more. Maybe we could form a CLP solver study group? Meet online once a month to run a specific hand through a solver to 1) learn how to use the software, and 2) learn how to analyze and interpret the results. Maybe one of the instructors could even host?
    @bart, @KiLee
    by 2dpbuck MrO
  • KiLeeKiLee Posts: 249Pro
    Superfly said:
    Hi Dana, I don’t currently have a solver or much experience using them. But I’d be very interested in learning more. Maybe we could form a CLP solver study group? Meet online once a month to run a specific hand through a solver to 1) learn how to use the software, and 2) learn how to analyze and interpret the results. Maybe one of the instructors could even host?
    @bart, @KiLee
    It’s something to consider. Ultimately, it’s up to Bart. We will chat about it.

  • KiLeeKiLee Posts: 249Pro
    Here is a teaser-
    Solvers are very relevant in-
    HU pots
    Small, single raised pots
    More relevant as the PFR, but to a slightly lesser degree, as the PF caller
    More relevant in earlier streets
    Slightly more relevant vs reasonable villains

    They are less suitable in-
    Most multi-way pots
    When the pot is 4bet or even 3bet
    When there are postflop raises
    When the board is very scary and the villain is passive.
    When we get XRed
    On later streets
    When the V is very predictable as in tendencies and Live tells.
  • maphacks Posts: 1,985Subscriber
    short: they are quite useless in low stakes live poker^^

    also the problem is, looking at some solver outcomes here and there won't help you that much but will rather harm your live game. if you want to get a better player from "technical/GTO" standpoint, you have to put in hundreds of hours with solvers and implement the stuff by playing online where the stuff is applicable.
    I mainly play online these days and even I have quite some trouble in the first 2-3 sessions when I go at a live trip. the situations you face on a live poker table, even at higher stakes, but more so at the 5-10 level and below are so unique that most of the spots are basically unsolveable.
    sure you will get some good fundamentals once you start working with solver a lot, however, be aware that most of the stuff won't really help you too much when you play 200bbs in 4way pot vs 2 limp callers and one shorstacker^^

    I think that even preflop GTO charts are quite useless because you can flat way more live since ppl don't squeeze enough, there is usually one rec player in the pot and your postflop edge will be tremendous postflop.
    also very few solutions you can buy online are for 9max games AFAIK.
    and let me tell when you follow these charts you will end up with like vpip5 (assuming you just go with the charts and don't adapt to peoples super loose opening ranges in most live games).
  • maphacks Posts: 1,985Subscriber
    edited January 23
    PS: @OP, you should also nodelock villains response when it comes to calling your bets. I highly doubt that check/folding river even after potbetting turn, will yield the highest EV vs a typical rec. but as you can see you have to do a whole lot of nodelocking. this takes forever and the outcome is still not very reliable because you basically make a million assumptions on every street.
  • dpbuckdpbuck Posts: 1,977SubscriberProfessional
    Thanks for the response @maphacks. I definitely have a lot of respect for your posts and your thoughts.

    Realistically, what I'm trying to do is really understand what adjustments I should be making against villains who aren't playing optimally. It's great knowing that I have a villain with a super-wide range. But how do I maximize profits and/or minimize losses against these villains in certain situations. A solver is just another tool to use WHEN USED PROPERLY. I'm almost undoubtedly not using it properly, hence the point of this thread. All that being said, you make very good points.
    maphacks said:
    PS: @OP, you should also nodelock villains response when it comes to calling your bets. I highly doubt that check/folding river even after potbetting turn, will yield the highest EV vs a typical rec. but as you can see you have to do a whole lot of nodelocking. this takes forever and the outcome is still not very reliable because you basically make a million assumptions on every street.
    I thought about this river spot a lot, and it actually makes sense if we think about how I assigned the different parts of his range.
    - The only Qx that calls a flop bet are QdXd (which I block), and Q6s/Q2s (which I lose to)
    - The only Tx that calls a flop bet is TdXd, and I do not have villain calling three streets
    - Villain doesn't call three streets with 77-99, nor 6x.

    All that leaves 2x and 66 (which I lose to) and busted flush draws (which I don't think he turns into a bluff). So in that case, specifically QdJd is a check/fold on the river.

    More than anything, it seems like I should revisit the range/node-locking I did and look at different variances:
    - If I bet super-small and he heroes with Tx some percentage of the time.
    - If I check and he bluffs some of his missed flush draws.
    - If he never slowplays 2x past the turn.
    - etc.

  • MrSpecial Posts: 330Subscriber
    Hi,

    I took a look at your workflow and think it looks pretty good. Not a lot to add to the discussion that hasn't been already said.

    I believe nodelocking is awesome when you are really sure of something that villain is always doing. This is rare, but some people play a check range strategy on certain board types and it is really good to understand the gto way to counter this.

    I just wanted to point out that there is a danger in believing, that giving a solver x bet sizes and seeing which one he chooses most of the time, is the proper way to understanding which bet size is optimal. As the creator of GTO+ in the twoplustwo thread has several times pointed out (and the creators of PIO as well), a solver will show almost no difference in EV when taking a big bet line over a small bet line.
    It's possible to construct a credible GTO strategy for any bet size that will perform similarly to any other bet size. Now, although it may be argued that the tiny difference would still mean that a certain size is preferred, the problem is that a lot of detail is lost when translating the GTO solution to something that can be used in practical play. When deciding on sizes, it's better to just pick a size that you feel comfortable within your own play, or where you feel your opponent will make the biggest mistakes.
    (Scylla in the twoplustwothread).
    So finding the best bet size isn't necessarily what we can achieve with solvers, but it is rather how to play different sizes with which part of our range.

    Nonetheless, parts of our workflow look similar and I am sure we and others here could benefit from maybe forming a small study group. Maybe somehow on slack or discord (without the trolls). Pm me if anyone is interested.

    Cheers
    by 1dpbuck
  • dpbuckdpbuck Posts: 1,977SubscriberProfessional
    Thanks for the response @MrSpecial. I'm definitely intrigued by that comment of betsizes from the GTO+ creator, as that seems counterintuitive to me. I mean, isn't that why we can take smaller sizes with our entire range in some spots, but in others it makes more sense to polarize? I need to think about this a bit more...

    I know I am interested in a study group, and @superfly is as well. Not sure if there are others.
  • MrSpecial Posts: 330Subscriber
    If you want to challenge his quote, the easiest is to try out completely different sizes and then look at the EV which the different sizes have for the overall tree.
    There will likely be only a tiny difference in EV.

    I actually had asked him myself about it, because I also found it counterintuitive, but the more time I spent with solvers the more sense it made to me.

    Here is my question and his answer from a while ago:

    Re: CardRunnersEV
    Quote:
    Have stumbled up on some opinions of users that suggested solving for multiple bet sizes first, then interpreting the results and looking for which size the solver prefers. Then solving only for that preferred bet size. Would that work as well to find out optimal size?


    Trying out the different single sizes would work best, given that if you want to find the optimal size, then just trying out different sizes is the most direct approach. Using multiple bet sizes may result in the solver landing on a solution where for example a 50% size and a 90% size is used most frequently for two different ranges, whereas if you were forced to use a single size for the entire range, it would have landed on 60%. In the end though, bet sizing has only a very limited influence on the overall performance, and it's far more important to focus your efforts on the quality of play.
  • kaboojiekaboojie Posts: 360Subscriber
    I’d be interested in this study group. Initially, I’d probably just be a spectator until I purchase Pio and get up to speed with some of you who are more well versed at solver work than a currently am.
  • Superfly Posts: 145Subscriber
    I’d be interested, but like kaboojie I’d be starting from scratch.
Sign In or Register to comment.