This is the support thread for GTORangeBuilder, a postflop GTO/Equilibrium strategy solver.

way to expensive to even consider it

Totally understand that for the turn license, and appreciate the feedback. That said...

There are some affordable options, eg there is a free trial for river calculations and unlimited river calculations are only $10 a month. I also I try and release videos / blog posts with example turn solutions that are free to browse so that people who are motivated can learn for free. An example is here:

http://gtorangebuilder.com/#share_sc...6012/root_v=22

Associated video: https://www.youtube.com/watch?v=iMlcwSPKcKM

Associated blog post: http://blog.gtorangebuilder.com/2014...h-gto-gto.html

Also anyone with a license can compute solutions and share them with non-license holders for free.

Of course, the turn solver (and the flop solver when it is released) are designed to be high end products for pro/semi pro players, training site coaches etc, which of course is where most of the existing user base comes from. The river solver is the part of the program that is really designed for non-professional players (plus the free blog / video content of course).

Thanks!

-swc

Right, but the product itself decreases a lot in value if you cant make those calcs for turn. So thats why the pricing for the full product is way to expensive for me to even consider it.

Buying the GTORangeBuilder River wouldnt make any sense. Would only be the GTORangeBuilder Turn that would be interesting. But with that pricing, no way!

GL tho, program seems interesting.

yeah crazy prices!

monthly price + activation fee + my wife and my car ...

also I m pretty sure a similar soft will come up soon way cheaper

I'm sure someday similar software will come along, but of course the value of computational GTO solutions are highest right now when they aren't widely available and thus they can give you a bigger edge against your opponents and a big head start on improving your game. The fact is that right now no existing available software offers anything close to what GTORB does. I've actually had several existing customers ask me to raise my prices for this reason, and I'm sure that if I made my software free it would make some players unhappy and hurt the profitability of certain games.

Anyways, of course its a free market, I would never expect anyone to purchase the software unless they thought that it would help them gain enough of an edge in their games to give them a positive return on investment, and similarly I'd never set a price that I didn't think maximized my own long term EV.

Obviously some people are going to feel that purchasing a turn license is +EV and some are not and it depends on many factors (eg what games / stakes you play). My goal is just to make the best software I possibly can so that those who have decided that purchasing a license is +EV continue to have a great experience.

River spot where I want OOP to only bet one sizing and I want IP to be able to call or shove, is this possible? Every time I try do it - it gives the option for OOP to shove as well

wud also prefer monthly turn option w no activation fee

This is possible in some cases depending on the stack size using the advanced "Include Shove" setting and putting it to "No". It requires that a raise of the same % pot as OOPs lead is a shove (or more). Example is here, hit the back arrow at the top left of the solution browser to back out and view the scenario setup:

http://gtorangebuilder.com/#share_sc...84/root_v=22.1

At the moment if you want the IP shove to be a bigger % of the pot than the % pot lead by OOP there isn't a way to do that. I am working on adding an interface to build any kind of betting tree you like but it will probably be a few months before all that is done.

Regarding the activation fee on the turn monthly, definitely appreciate where you're coming from and I'll think about reducing / removing it.

Thanks!

Interesting software. But if I understand that correctly the solution is highly dependent on the ranges, with which we arrive at the river (or turn if we start there). And in a nash equilibrium these ranges should depend on the way we play the river. A strategy calculated with this tool could be highly exploitable and very bad if our villain's range differs from the range we expect. We need strong reads to estimate a good starting range for our villain, but if we have strong reads it's probably not the best idea to play a balanced strategy.

Some good questions in here, thanks!

Regarding the turn solutions depending on how you play the river, of course a GTORangeBuilder turn solution will also show you how to play on every possible river outcome as part of the solution (just click the "to river" arrow after a turn call and select a river runout). Similarly, once flop solutions are available they will also include GTO play on every subsequent turn and river that might be reached. The software considers reaching the river with every possible range that can be constructed from your turn range, so in a turn solution, if your opponent goes to the river with a non-GTO range, he will be decreasing his EV if you stick to the GTO strategy listed.

In terms of how robust the solutions are to accurate starting ranges, the goal with the software is definitely not to memorize the solution to a specific situation and play it exactly. Instead the way I recommend using the software is to look at a solution to a situation of interest, tweak the starting parameters (stack to pot ratio, hand ranges, bet sizing, etc) and see how the solution changes, and use your observations in that spot to develop simple testable predictions that you can use to immediately improve your game, like "I should (or not) be check raising my combo draws oop on the turn on wet boards in single raised pots".

You can then enter a different but similar scenario with different but similar hand ranges to test your predictions outside of the model scenario you were originally examining to determine how robust these predictions are. I have a new CardRunners video that will come out in a month or so with some pretty interesting examples of this process, but the general goal is to use GTORangeBuilder to provide a way to be scientific and quantitative about hypothesis testing and measuring the quality of specific strategy advice, not to run a scenario, memorize it, and try and play that way at the tables.

You can browse an example turn solution and watch a video with a brief introduction into making testable predictions using the software here:

http://blog.gtorangebuilder.com/2014...h-gto-gto.html

I'd guarantee that even players who have spent a lot of time constructing ranges would be able to learn a number of surprising results just by applying this process to the river.

All that said, even if you don't follow the process above, GTO strategies, particularly with wide hand ranges are generally more robust than you might guess to reasonably small perturbations in ranges, far far more so than maximally exploitative play. You can of course also approximate fuzziness in your opponents range by giving him a small number of combos of "unexpected hands". Outside of toy examples like polarized vs merged clairvoyance games from the mathematics of poker I think if you poke around you will be surprised that even river strategies are reasonably stable to small changes in starting ranges in many cases.

My upcoming CR video has a bunch more analysis (Multistreet theory and practice pt 3), I'll update this thread when it is released in about a month.

EDIT:

Just to back up my claims a bit with examples, I just randomly ran 2 completely made up scenarios on the same board with hero IP with top 20.4 of hands for hero and the top 27.5% for villain. I then reran it after removing about 10% of villains range, mostly K/A hands.

http://gtorangebuilder.com/#share_sc...42/root_v=22.1
http://gtorangebuilder.com/#share_sc...65/root_v=22.1

As you can see, from the hero's perspective he strategy does not shift at all drastically in response to the villains range shift. Rather it is a pretty smooth shift about as you might expect as villains range strengthens. He folds to a bet a few tenths of a percent more often, checks back a few tenths of a percent more often, and his EV is a bit lower vs the tighter range. But for example the way the hero reacts to a villain check is almost identical in both spots and thinking that you couldn't learn something from the first scenario that would help you play the second well is definitely not true. You can of course play around yourself a bit to measure robustness, I plan to add features in the future that will let you measure robustness to starting hand ranges (and other settings) precisely and mathematically in the future over the coming months.

Hey guys, if anyone is interested in a sneak peak of a flop solution you can check out my latest blog post:

http://blog.gtorangebuilder.com/2014...-3-street.html

Hi,
Are you planning on adding different betting options for the SB and BB?

I mean if i set a scenario where I give BB option to bet 50%pot i also like the SB to not be restricted at making 50%pot raises only, id like to give him other options as well.


ps
on a second thought, i think it would be nice that every bet stage have their own options. ie: 1st bet, 2nd bet, 3bet and so on. IMO this will help a lot in recreating scenarios where human players actually have very few and fixed bet(raise)sizes in their game.

Yeah I think this is definitely a really good idea. I'm planning to add the ability to basically let you build any betting tree you want on each street (and even different ones for different run-outs) but the UI for the is reasonably complex so it may be a few months before its launched.

Thanks for the feedback, really appreciate it!

swc, I have a couple of questions:

1) What kind of algorithms do you use in finding equilibrium? Do you use No-Regret, Fictitious Play or others?

2) Does the software requires large amounts of computational power (so you are forced to run it on server side) or it can run on a high-end PC too?

Good questions

1) The software uses a highly optimized variant of fictions play
2) It could run on a high end PC for turn/river calcs, the memory required for flop calcs (and even larger turn calcs) is well beyond what a home computer would have. The solution files also end up very large so it would fill up your hard drive very quickly and be a pain to transfer solutions between different machines or share them with friends if they were not cloud hosted.

thanks

-swc

What do you think that the complexity is like in going from "optimizing" a given set of bet sizes to actually solving for the "optimal" bet size(s), even w/ out knowing if the starting ranges you use are sound?

Its mostly a question of resolution. If you truly want the exact optimal bet size (e.g. I have a blog post showing how to do this type of calculation manually here http://blog.gtorangebuilder.com/2014...-solution.html) then that is a very complex problem.

If you want to know "If I limit myself to considering betting 5% increments of the pot size" which is best then it is approximately on the order of complexity of going back a street.

I haven't made it a big development priority at the moment, mainly because in my early simulations (and in spots where I can mathematically work out the optimal bet size) the EV difference between an optimal bet size and something that is at all close is shockingly small, and my early tests did not indicate that that was likely to change as hand ranges got bigger.

As you can see in that blog post in a toy game, betting 50% of the pot is 0.02 chips (0.02% of the pot) lower EV than betting the optimal amount 52.69% of the pot. Of course 50% is very near 52.69%, but it turns out that the EV gain from betting 40% or 60% vs 52.69% isn't actually very large either.

After I get flop calcs commercially launched, depending on user feedback I may go back and investigate bet sizing in more depth, but in my limited early experience the potential EV gains weren't enormous. That said, if when I get time to look at this in more depth there are some spots where the EV difference of various bet sizes is very large it would definitely be exciting and could potentially be very practically valuable information.

After I get flop calcs commercially launched, depending on user feedback I may go back and investigate bet sizing in more depth, but in my limited early experience the potential EV gains weren't enormous.

Thanks for responding; two follow-up questions.

1) Is there a way to get some kind of estimate on the difference between a truly "optimal" bet-size and the best solution from a (given) limited range of options?

In other words, is it possible to take the salient features of the calculation and generalize them produce some kind of bounds on the "user error" value which can actually be displayed to the user?

2) Is it correct to assume that the problems you're currently working on are actually kind of elementary (but important) problems compared to actually "solving" poker?

In other words, is there some kind of huge leap in complexity when you want to start actually solving for pre-flop strategies? It seems like this is where the lion's share of the work is.

Also, I think that $49.99/month for your product is a steal, and I wouldn't think twice about paying double for the turn solver (or the river solver, for that matter).

Oh, a third (but unrelated) question:

What's your opinion of PokerSnowie?

On a scale of one to ten, their marketing is seems like an eleven on the idiot-meter, but what do you think of the actual engine?

Glad you like the software and think its a bargain for the price!

Regarding 1, you can compute epsilon equilibrium distance without knowing a GTO solution which is a bound on error. For more on that see my youtube videos:

https://www.youtube.com/watch?v=j7xSHC90_Og
https://www.youtube.com/watch?v=-UenhnsjiOY

This is the standard way to measure the accuracy of an approximate GTO solution and the "Nash Distance" number that GTORB shows you at the top left of every solution is exactly that number.

However, the calculation that GTORB shows is the "in model" epsilon equilibrium distance, that is it is the distance assuming that the only strategic options are the ones in the game tree that you specified.

In the future I plan to add options to do stuff like see "If I play the GTO solution to the game where people can only bet 50% pot but they are actually allowed to bet 40%, 50%, 60% what is the out of model equilibrium distance", i.e. how bad is it to treat 40% pot bets as though they were 50% pot bets.

Unfortunately I don't think there is only way to do that type of calculation to generate bounds for arbitrary bet sizes. Anyone being scientific about their work should always have in-model epsilon equilibrium distances calculated, but out of model ones in full generality are very hard.

Regarding 2) Yes I'd say that is correct. I don't really think of GTORB as a tool to solve poker but rather as a more practical tool to help players win more money. Its entirely possible that the "standard" pre flop ranges (and bet sizes) people use even at high stakes are very wrong, solving poker would involve figure that all out exactly. Regardless of what optimal play pre flop is, most people play one of a few reasonably predictable and static ways. Answering the question "if we assume they play that way preflop, how do I play post-flop optimally" is a very directly useful and +EV question to answer. In fact, if our opponents do not adjust their pre-flop play, answering that question is going to be higher EV than knowing "assuming everyone plays GTO pre flop what does optimal post-flop play look like".

Finally regarding Snowie, my understanding is that they removed all claims to snowie being GTO from their marketing, I think their CEO posted on 2p2 a while back correcting some of their early marketing that claimed the program played GTO. I think that they are solving a very different problem them I am, they are trying to make a strong poker AI, not solve for GTO play. They want their AI to be non-exploitative but that is very different from it actually being GTO. Of course just simple epsilon equilibrium measurements (which you can make with CREV in simple cases) would be enough to spot non-GTO play that it might make in simple spots (e.g. on the river).

In terms of the value of AIs, obviously an AI that could just play true GTO in all spots would be the holy grail, but it doesn't mean that a non-GTO AI couldn't be better than any player in the world theoretically speaking and thus very valuable to play against/learn from. This is the case in chess and people regularly train against AIs despite chess being far from solved and unlikely to be solved this century. Whether Snowie is a good enough AI to constructively practice again and learn from is in the eye of the beholder.

Every year there is a computer poker championship and my understanding is that the winners of that are still not generally at the level of top humans. I think one of the former champions stated their bot was comparable in skill to a winning micro stakes grinder.

Some AI approaches solve abstracted (simplified) versions of poker for GTO strategies, often ignoring blockers, simplifying bet-sizes, making strategy assumptions etc and play that way in the full game, including the winner of this years event. A GTO focused bot both won the most chips total against a pool that included fish and had the best record against tough opponents this year which I think should convince a lot of people who think GTO would break even against weak players that they are mistake. There is a neat article on it here: http://www.cardplayer.com/poker-news...apon-for-poker. However, I think most academics have never played poker at a high level and greatly overestimate how well their bots would do against top players so I would take the hype with a grain of salt. AI research in incomplete information games is a young field. The AI tournament this year was the first to feature ANY 3-handed play of any kind and they played limit Kuhn poker (the AKQ game), which is infinitely simpler than NLHE.

Will there be a price increase above the turn price point when you release the ability to solve flops as well?

Thanks.

I'm still figuring out the flop pricing, flop calculations require more expensive servers and currently take hours to run, so I may end up charging per calc (or per server hour of computation since that is what drives my costs).

I don't think the turn or river license pricing will be affected at all by the introduction of the flop calculations although I may give all existing turn license holders a free flop calc or two to experiment with.

-swc

Ah very interesting, do you have a preliminary guess at how much it will cost per flop calc? I understand if you don't want to speculate on the pricepoint early on but I'm curious

I think I'd rather not speculate until a really figure out all the dimensions of my costs and a better sense of demand at various price points. The solutions are gigabytes for large scenarios so i need to figure out the cost of storing them in the cloud forever in addition to calculating them, then I also have some ideas that might 2x my computation speed which would half the calculation cost, etc so its just too early to say.

I understand, thanks for your replies!