This article is brought to you from a member of our team who also happens to be a technical specialist (kostya1q1).
Hello everyone!
Not long ago I decided to return to the game (I haven't played for over a year because of our project) and started testing different options for working on the game. Not long ago, the another tool became the gtowizard. And I have some interesting info with examples for you that I want to show about this tool. As it turns out, this happen to be a long read, so make some coffee and enjoy reading!
Considering postflop, it became clear rather quickly that the basic 50$ subscription does not provide everything I want from the software. Bet sizes are not what I want them to be in many cases and in some spots, the critical lines that influence the strategy on a fundamental level and massively affect the logic of a specific hand, are simply not there. This was mostly noticeable in BvB spots; in other spots, it was a bit better. However, when I first saw a 150$ version with custom trees, everything changed! (Although it began to resemble your ordinary solver, but whatever). Please note that I myself haven’t yet bought a 150$ subscription.
So, I didn’t really like the $50 version for postflop solutions. However, the main reason for purchasing Wizard for me was the trainer. To be honest, GTOWizard has the best trainer in terms of functionality and UI of all that are currently on the market. I really liked the fact that you can look at how you should play your entire range in any spot in a separate window in real time. However, the depiction of the opponent's range is uninformative for some reason, which is really weird, but okay. Nevertheless, I began to often notice different preflop lines (not all-in 5-bets, strange cold-calls), preflop sizes and even hands that I would say are extremely weird for some spots. Well, we have already talked about postflop before. Why are they weird? Now that’s a great reason to write this whole article for you and post it on the forum.
Let’s go over preflop in more detail, because it’s the basis of any trainer and any postflop decision. Wizard has a number of preflop trees and among a large number of these, I would say, there are only a few that use reasonable sizes. What trees are offered to us for the trainer work, to be honest, it is not very clear to me. Nevertheless, maybe I’m not experienced enough, but finding a reasonable game tree (preflop only) to work with always takes a lot of time and some nerves as well.
Recently, while working on a certain spot in the trainer, I noticed a horrible preflop solution. No matter how cool, fast and advanced GTOWizard is, it still has problems with poor preflop solutions. First, some technical data. We turned to a foreign consultant for help, who worked with analysts who conducted research on the quality of Wizard’s preflop solutions. The algorithms and bucketing systems are the same as those build into HRC, which is not good enough for the modern cash Hold’em calculations.
Let’s first introduce the definition of the quality of the solution, which is based on two parameters.
1. The depth of the solution (calculation time; NashDist);
2. The bucketing system and the number of buckets (In other words, the abstraction on which the preflop is calculated. Abstractions are basically characterized by the number of buckets. The more buckets, the better the solution. Preflop calculations are not possible without abstractions, because it would require an immense amount of RAM).
Talking to developers of such software, I noticed something they have in common; they all pay too much attention to point #1, and not enough to point #2. In the case of Wizard, the developers decided not to bother at all. Our consultant showed us a lot of different data, and from what we have seen we can say that if you translate it into buckets, then the preflop for Wizard ranges approximately from 1k-1k-1k to 1k-5k-5k. While the classic solutions for GTOPacks have 10k-100k-100k buckets. Our approach to iteration/nodes (calculation depth) is considered hybrid. Could be a lot of them, could be not so much. In the process of calculations, we use the quality check approach. Often the same number of iteration/nodes can show different quality depending on different factors.
Now let’s take a look at an example of how it looks in GTOWizard
Here is the spot: 3bet pot MPvsCO nl500 (3bet 7.5bb)
This is how it looks in Wizard
There are 2 things we are interested in. You don’t need any special knowledge to see my point.
1. K6s in the 3-bet calling range, you must agree, is super weird. I can get behind 4-betting this hand, but calling it looks random.
2. The number of small pocket pairs 22-44. It seems we have too many of those in the preflop opening range to begin with (especially considering the number of pocket pairs that dominate us in villain’s calling range). And the open-raising frequency distribution is weird. You can see on the right, 22 opens 41%, 33 – 36% and then 44 again 41%.
3. The third point is less significant and not that noticeable, however it also looks strange to me. I am talking about the absence of A9o in the MP opening range. Of course, it could be caused by a SB’s cold-calling range, but the frequency should not be reduced to zero still. Upd: While I was writing this article, I put together a calculation on the fly regarding SB’s cold calling. 22 are not in the range anymore, while A9o opening remains the same 75%.
Now this is the solution for the same spot (except for the SB’s cold calling range and shoves, which do not affect the problems described above) with same 7.5bb 3bet from GTOPacks:
1. As we can see, K6s is not in the defending range.
2. Pocket pairs, especially lower ones, are opened much less often and the frequency increases linearly, the higher the pair, the higher the frequency. 18.5% opening frequency for 22 looks much more reasonable!
3. We see 75% open with A9o.
While I was going over this Wizard sim, I also noticed some weird stuff in its SB defending range vs RFI. The notorious K6s is here again, more 22-33 than higher pairs and AKo is partly called:
You might say it’s a small thing. And I would agree, those errors are probably not that significant. But what about the tree itself? Well, let’s say we 3-bet 10bb SBvsMP, why then the squeeze on BB is again 10bb after SB calls? Out of position it does not seem reasonable. At the same time, when BU and SB both call, then BB squeeze size is 16bb.
We can finish it with noticing the fact that in this preflop solution there are non-all-in 5-bets, including those that put 40% of the stack into the pot.
All these shortcomings add up and create a domino effect, where small inaccuracies change the outcome significantly. Of course, your average 50-100NL regular probably would not care at all. Although, there are still many questions about postflop branches, which are not played in real 50-100NL games and are mostly of no use to those players.
So, answering the most frequently asked question, how are we better than GTOWizard? The answer is quite simple. Preflop. The quality of solutions. At least of such simple solutions as described above.
Does this mean that you have to relearn and force yourself to study postflop solvers? Of course not! There is a universal solution! Wizard still has some great qualities. GTO trainer is great, the postflop tree creation interface is easy to handle and it still has the fastest solver algorithm. So why not combine GTOPack preflop solutions with GTOWizard’s functionality? You can import GTOPack ranges into the Elite version of GTOWizard without any problems. And the software for browsing our solution packs is very simple and intuitive, if you want to study preflop separately. Not to mention that the solution packs are quite cheap. We also provide full support for everyone who purchased our solutions and are happy to answer any questions. We also often refine our packs and add new calculations without changing the price. Very soon, we will present add-ons to the standard packs. We already have ideas on what we can add.
-------------
We didn’t stop at this tree and decided to dig into the preflop solutions called “Research”, where there are many different sizes for each spot. We took 100NL calculations as a basis. Now please pay attention if you study these solutions for small stakes. The rake calculations in those are way off the norm. The standard rake cap for 100NL solutions is considered to be around 3bb. After some calculations, it was found that Wizard uses a rake cap of 6bb, which is twice above the norm. This is the equivalent of 50NL rake. So in postflop solutions for 50NL the rake may also be off.
So what did we do? We decided to recreate the tree starting from BU like in “Research” and see what the differences would be. The tree is quite large and to be honest with you, we decided not to waste a lot of precious server time, because we have already recalculated the tree with different rake, so the calculation depth is not that high. Because of that, the frequency distribution for some sizes may not be as accurate as possible. However, there are still some big differences even in choice of sizes, and there shouldn’t be any. This is an indicator of the low quality of GTOWizard’s solutions. I will attach the calculations for our tree below so that anyone can download it for free. Now I want to demonstrate the basic BBvsBU spot.
You can notice different things here, but the critical importance and a ‘leak’ is the presence of A4o combo in both wide 3-bet and calling action. In no simulations with such rake structure will you find such 3-bets or calls. Even with 500NL rake this combo is not a 100% call. And the 4-bet sizes in the tree are completely random. I will attach screenshots of Wizard’s and our solutions, for those who do not want to look for it themselves.
Now a screenshot of our solution:
You can download the calculations here
wizard vs gtopacks research.rar (2.3 megabytes)
I had to zip it because the site does not allow attaching other formats. To view the solution you need Simple PREflop Holdem. You can download it from simplepoker.com. License no need.
So,
let’s summarize everything we went over in this article. GTOWizard is a very cool and handy tool to work on your game, but with a number of amendments. Its main disadvantages are incorrect preflop solutions and frequent “inaccuracies” in the selection of postflop bet sizes. If inaccuracies in postflop sizes can be fixed through custom postflop trees, using the Elite version of the software, it wouldn’t work for poor preflop solutions. Therefore, we recommend importing preflop solutions from GTOPacks into the Elite version of GTOWizard in custom postflop trees.
That would be all for today, my dear readers!
I sincerely hope that someone have read it top to bottom. Share it with friends and don’t forget to stay cool and smart!
Our shop
Our contacts:
Telegram: @funpokerrr
Telegram: @kostya1q1
Discord: !funpokerrr#6788
Skype: funpokerrr
Last edited by Funpokerrr; 12-07-2024 at 10:01 AM.