I tried to download it but I got "service unavailable" error.

I just tried downloading it and running it and it's working now. I expect you experienced a temporary service unavailablity, either due to a problem with your ISP, or a problem with my hosting company (iiNet).

The call branch showdowns are available too. The analysis is there for every showdown that RvR can see enough ranges to analyse (every bet range from one player where there is also a call range for the other player). So you get multiple showdowns, and so multiple showdown analyses each hand.

Also, your results for the hand are an appropriately weighted combination of all fold-equities and showdowns, based on fold ratios and showdown equities, each multiplied by the appropriate residual weight (which depends on how much of each player's range they have folded). So your results for the hand are accurate. A more accurate representation of your skill than... anything else I can think of.

I tried it verse myself. Improving my range thinking is certainly something I would love to work on. I left my email in the notification area. Feel free to pm if anyone wants someone to participate with.

Just played some hands with guyupstairs. Great program and a great guy. Come in and run some hands people. Its great practice.

Interesting idea. Is the main goal of the advanced training to find an approximate GTO strategy for the situation or to get better at exploiting your opponents?

It's both. It lets you find a nearly-unexploitable strategy, but it also lets you train making small adjustments to take advantage of the slightly exploitable play of your opponents, without letting yourself get more than a little exploitable.

The reason is: this software is about moving up in stakes. It starts from the assumption that you know how to beat fish, and now you want to know how to beat the regulars. Because that's what it takes to compete at higher stakes. If you're winning money from the fish and losing money to the regulars, you're not going anywhere. If you're a good bumhunter, you might make a profit, but you're not going to be able to move up. You ought to be winning money from the regulars too. And the regs certainly are exploitable. At all but the highest stakes, their defense is not good. And with this software, when you get to the highest stakes, and their defense is good, well your defense will be good too.

Thanks for the explanation. I am mainly interested in it as a GTO tool. I have read more about it now, including the above discussion with pasita, but I am still not sure how you would use it to get close to a GTO strategy.

My understanding is that two players would:
- Play through a hand in advanced training
- Analyze the results at the end of the hand
- Make a small adjustment towards exploiting their opponent
- Repeat the process

The problem that I see is that in the next hand:
- There will almost always be a different board
- Different player actions could be selected by the system
- Players could use different bet and raise sizes

This means that the next hand is almost guaranteed to be played in a different part of the game tree and as a result the players don't get a chance to work on their previous strategies. Each hand will have a different GTO strategy that the players are looking for.

As a completely separate question, have you considered changing the system to allow players to play hands even if they are not logged in at the same time? I am thinking something like: player 1 selects the scenarios that he is interested in playing and logs out, player 2 starts a game with player 1 and chooses the ranges for his first action, player 1 logs back in and continues the game, etc.

Yes, but two small corrections: there is now "play-on training" which is like advanced training except that you always get to play out the hand to a showdown (even multiple showdowns, believe it or not); and the analysis at the end of the hand is automated: the software tells you just about everything it is possible to say objectively about the play of the hand.
Yes, except for the bit about players' actions being selected by the system. You're playing against real live people, nothing is selected by the system (except the turn and river, obv.).
Yes, this is true. The solution is to start with very specific situations. For example, there is a training situation where MP has raised preflop and been called by BTN (everyone else folds), and the flop is Kh 8d 3c. Here, you can learn how to play optimally on dry boards. When you play against a given person, they will tend to use the same bet size, and so will you. There are very few possible betting lines. Even when betting lines diverge, you can see after the hand what your opponent's ranges were for the other actions they could have taken. Trust me, this is very effective for learning near-optimal play for dry boards.

Then there are other situations, such as a wet board situation, 9h 8h 6c, same preflop. Then once you're good at different board textures, you can try that situation with a random flop. Then once you're good at "MP opens, BTN calls"-type situations, you can try a "BTN-opens, BB calls"-type situation, or a 3bet pot.

The point is, even if there was software that could tell you straight up what GTO-play looked like in every situation, it wouldn't make a great training tool. You need to live it, to experience it, to keep adjusting to the small adjustments that your opponent makes, to learn what solid near-optimal play looks like. (Unless you're a genius at rote learning. I guess there are some people out there like that. But still: there is nothing out there that can tell you straight up, for any given situation, "This is how to play GTO here.")

That's also the reason you can only play "near-optimally": true GTO play would take infinitely long to establish. But it turns out that's okay. When you play near-optimally, even if you get maximally exploited, you are giving up only a tiny edge.
Yes, I've considered this. But it's not worth the effort. I have enough great features to add already. And when I do, there will be enough people logging in to play. Well, I won't say never, but it'll be far down my list. Once I've done all the stuff above it, I guess I'll get to it. As long as new, cooler things don't come up!

Thanks, I mainly just wanted to make sure that I understood the process. I have added some comments below but they are not that important.

I am not sure if this has changed due to the play-on training. What I was saying is that the players don't get to explore all of the possible betting lines in a single hand. The first player selects his ranges for betting and checking and then the system either tells the second player that the first player has bet or that they have checked. The system is selecting the action that the players are going to continue with. If in the next hand a different action is selected, the players will essentially be playing a new smaller game and their previous strategies don't necessarily apply. I am assuming that the new training still doesn't take the players through all the possible betting lines since it would take a long time to complete just one hand.

I didn't know that you already had the option to keep the flop fixed (I was going to suggest it).

I agree. There would have to be some tools built around it to simplify and analyze the solution. I don't think that memorizing the whole strategy would even be an option without understanding it.

This brings up another point. I don't think that saying that a strategy is near-optimal has any meaning unless you measure how far away it is from being optimal.

Good point. Increasing the number of players is probably a better way of solving the problem.

I tried the software today. It was very easy to play with. It worked flawless (it says still in alpha stage) and I think it is already a great tool if you truly want to improve your game against competent regs. With a friend it is a lot of fun too, played for an hour with a player I met there.

I would only suggest that there would be, lets say, a daily competition where all the players could play the very same game tree branch (fixed bet sizes and community cards, but random position) and then some additional analytics where you can compare how you did against every other players ranges and on average. That way you would soon to see whether your idea of GTO is doing often badly against some strategies, like for example aggressive strategies. And you could see whether some ppl are able to construct robust strategies over time and over different strategies, i.e. being always close to GTO. Could yield more activity too?

Thanks 'guyupstairs' for the software!

I tried your software. It looks interesting. The GUI is very stable (I am using a old vista laptop at the moment). One of the areas I had trouble with the interface was the ability to choose a range quickly.
I would have valued some shortcuts/buttons to highlight sub-ranges based on their equity or value. Say on a Q83r I could click on a Air button that will highlight all things such as J5, T6 ie air that missed the board or strongly made hands etc. With a half a dozen button to highlight value/drawing hands of various levels of strength I could then quickly assign a large number of hands to my check/bet or calling range. Rather that get bogged down with individual hands which I can still use to make fine adjustments.

I would like to see some connectivity with PT4 but I am no sure I can articulate exactly what that might be. I was thinking along the lines of evaluating alternative strategies observed in the DB vs. a pool of players with similar characteristics. Specifically I logged in twice to play against me but I couldn't help thinking that I could have retrieved a hand from the database to rep my opponents actions. But better still I could search for all similar boards e.g. JLLr where L=2-5 and derive the action from the DB. I can then evaluate my action vs. the population's observed plays. I appreciate your software doesn't do that but it feels that the engine is there. I think such a tool would gain a lot interest. Anyway just some thoughts. Good job overall and I hope you find a way to market this.

Hi, guys.

Sorry for my English, I am just learning to speak in it.

I like the idea of your program that the study your own strategy. This is a very interesting approach.

I have a few questions for you:

1. Could you add more situations against 40bb stackers, especially in the blinds, including preflop situations (4-bet minraises etc). I am having big problems with it.

2. What can you tell us about betsizing? How it can be improved with your program?

3. If preflop ranges are defined, is it possible to create an automatic player who will make the best decisions against you? Or is it not possible at the moment?

4. Can I add a selection of the flop and the opportunity to leave it for the next nand?

Thanks again for your work!

5. Sorry, but I didn't find a donk options.

Your English seems pretty good so far

I'm happy to add more situations, but I need you to give me the precise actions and the precise ranges for each player left in the hand. This needs to come from consensus, from the community, because if I do this unilaterally, nobody will be happy with the ranges.
This software lets you experiment with various bet sizes, in a way that is much more efficient than doing so at the tables. It also lets you practice flop situations that represent a particular opening-bet size, or 3-bet size preflop. But this software (nor any other method I have heard of) will directly tell you how to improve your bet size. The truth is that no one has an objective answer to the question of what size to bet.
That's not possible at the moment. I can see how that would be convenient for people (single-player mode), and useful for training to exploit the population tendencies. But Range vs. Range is about training against real people, real strategies, and making real adjustments over time to make an objective improvement to your overall strategy.
I don't think this makes sense because the flop will be different each hand. Preflop, you can do this – that is what the "direct entry" button is for. It lets you copy-and-paste in a previous range.
That's just betting, right? For example, if you are the BB in the "CO opens, BB defends" situation, you will be first to act. Whatever your betting range is there, those are the hands you are donking. Make sense? You only ever choose your action for right now, when it is your turn to act. You don't specify donking, checkraising etc. as such. It's the same as what happens at the poker table.

I really like your program and i tried it a bit vs a buddy and this was very funny

But i really miss the opportunity to create own scenarios or at least edit the ranges you used for different scenarios. Using own ranges will increase the level of training A LOT because you could f.e. put in perceived preflop ranges of an opponent you struggle with at the tables and can train vs his specific range. Pretty sure this function would be used by a lot of people and increase the popularity of your tool a lot!

You might be right about this. Can you give an example of the kind of range you find it hard to play against, and how you get to that point? Like give me an example of a situation you'd want to change, and what exact (weighted, I guess) ranges you'd use?

That way, I'll be able to see the kind of info you want to be able to input, and I'll be able to figure out how to go about it.

Well, first situation that comes to my mind is CO/BU vs BB 3bets.
At Midstakes NL Cashgames there are huge differences in peoples frequencies. F.e. i ran a db analysis a while ago for ~50 regs i have more than 100k hands on. Was able to create 3-4 different clusters (determined by stats like 3bet/fold to 4bet/call steal, etc). f.e it would be awesome if it would be possible to enter different ranges with weights like i can do in flopzilla:



and in the same case work with different 3bet-calling ranges to train specific spots vs specific opponents. Its like a gym for poker with different "workouts".

I like the preset ranges and situations in the tool so far, but some ranges do not really apply for the games i play in and so the training effect isn very big and i can see a much bigger effect if i would be able to input own scenarios.

As for software ergonomics i think i dont need very much.
Small wizard with description of situation and a range-matrix for both players with possible weights and then we can start from there.

That's some pretty neat analysis you did. But when you know someone's statistics, how does that tell you their range? If you know someone 3bets 15%, how do you decide how much of that is bluffs, and what weights to put on their bluffs?

For example, how do you know the guy who 3bets 9% and the guy who 3bets 15% don't have the same value range and differ only by bluffs? Or have the same bluff range and differ only by value hands?

Well, as everything in poker you have to work with assumptions. You also had something in mind while designing the implemented scenarios. Same with the ranges. If someone is 3betting 15% and 5betting 80% of this range you can assume that his 3betting range will include hands that are good for shoving (assuming villian is no big ******) which namely are pockets and AQ/AJs-type hands. Along with notes of hands that got to showdown and some filters in HM (Preflop Cards tab in HM1 will do wonders if you have a big enough sample on someone) you are able to come up with some good assumptions about how the population is 3betting a 15% range. A 8% 3betting range in a .1/.25NL game will also look way different than in a 5/10NL game assuming people have different theoretical background and different skill levels. So i'd really love to be able to create my own "poker workouts"/scenarios.
Of course, in the end you'll never know for sure and it is always a big guesstimate, but that doesnt keep you away from experimenting with ranges imo.

Hi again!

guyupstairs, can you make the program more visible? For example, replace all this QsTs or Kd8c etc. by images suits? It will be more clear and have the best training effect I think.

And can you record some video tutorial about your program (it can be a hand-by-hand against yourself in the advanced mode or something like that)?

I can't understand how to move the necessary cards in the range of a raise, when we forgot to do it. For example, I choose all suited Hearts hands, but for some reason I can't them move them to the range of a raise. Is it a bug?





and yet, can you reduce the limits to 0.5 / $ 1 to make it easier to count?

This is what happened: the hands you had selected were already in your raise range. That's why when you clicked "Raise" nothing happened. It was the other suit combos for those hands that were still unassigned, but because you had not selected those, they were not moved: nothing happened.

You can see which combos of a particular hand (e.g. A6s) are in which range by hovering the mouse over the button.

The other option is click "Select All" and then hit "Raise". This will move all unassigned hands to your raise range.
But what if you want to move up to $1/$2, or have to drop down to $0.25/$0.5? They're not dollars, they're chips. It's this way so that all the bets (including the small blind) are whole numbers.

Edit: I'll put in a warning message for when all the hands you move were already in the target range (with an explanation of mouse-over behavior), to avoid this confusion in future.

You're right.

Do you think it would be sufficient for there to be (for example) tight 3bet range, normal 3bet range, and loose 3bet range situations? I.e. standard situations? This has the advantage that the statistics work better (you can compare your results to everyone else's).

Hm, i actually dont see merit for such Statistics. Im fine with the statistics for the implemented scenarios which can work as kind of benchmark workouts. Why not just try it with the "full version". If people dont use it its fine and if they use it is also fine