If I understand correctly, you start with any strategy for player 1, then you have player 2 maximally exploit that strategy, then you have player 1 maximally counter-exploit that strategy,...and after doing this many times, the resulting strategies become very close to the real GTO solution (and you can find an upper bound to the error). I think this is what's called "fictitious play."

Hi,

There are two major algorithms that are standard for computing GTO solutions and are mathematically proven to converge to GTO play for a very wide class of games (far beyond just poker), Fictious play and Counter Factual Regret Minimziation. You do not need to have the machines play each other infinitely, and GTORB doesn't do any simulation of a bot playing itself or anything like that. There are learning algorithms that involve self play that converge to GTO if they start near enough to it and play forever, but to my knowledge that approach is not very efficient / accurate.

Most modern solution software uses an optimized variant of either fictious play or CFRM, eg the UoAlberta program that solved limit used what they call CFR+ which is a CFRM variant.

Any algorithm that can do either fictious play or CFRM can also easily measure its own distance from an exact GTO solution (GTORB calls this the Nash distance). In game theory there is a scientifically defined measure of how close any non-GTO strategy is to exact GTO and you can measure that distance without knowing what the exact GTO strategy is. I have a free youtube video on this here: https://www.youtube.com/watch?v=j7xSHC90_Og

Anyways if you want to learn more I would start by reading the wikipedia page on fictious play and walking through the algorithm with pencil and paper for solving rock paper scissors to get a rough idea of how/why it converges.

Hope this helps,

-swc

Edit: just saw tobakudan's response and its pretty accurate for fictious play (you average the maximally exploitative strats together as you go). CFRM is a similar idea, but look at the "regret" of the EV your current estimate of GTO against a fixed opponent strategy and reduce that "regret".

so if i understand correctly, if i buy the $299/6 month package I will be able to do unlimited simulations to find the best action on the flop?

So i can do (Hero BTN) (CC'er SB) and then assign 3 flop cards?

And does this also include Turn (4 cards) and River (5 cards) ?

Thx

Hi,

Not quite, you get unlimited turn and river calcs, full access to the flop library of 500+ precomputed solutions (with more added almost every day) that are all instantly available for browsing and the ability to submit new solutions for addition to the library. User submitted solutions may or may not be calculated and added to the library depending on their quality (for the most part the only stuff we reject is either obvious duplicates of existing content or obviously broken or poorly designed ranges / betting trees), but you aren't guaranteed to get to run as many flop calcs of your own design as you want (or any for that matter).

You can look at what is in the solution library without a license and browse some of the free solutions to see if you are interested.

You can also contact me at gtorangebuilder@gmail.com or via PM if you have some specific stuff you want to study and I can give you more info on what the best option for you would be.

thanks!

-swc

Bought the flop library. Is it just me or are GTO ranges just weird? A lot of stuff doesn't seem to make sense or just seem counter intuitive.

I think in a fair number of spots (particularly when it comes to c-betting) GTO ranges can be pretty different from what "standard" play does and it can take some time to adjust your thinking to be able to construct ranges that are similar to what GTORB might show on your own.

That said, once you get a decent sense of what GTO ranges tend to look like and why, it can actually be possible to build up a pretty good intuition for what ranges would look like in certain spots. I think a good exercise is to find some of the library solutions that use the same ranges on 5 different boards and try and order them in terms of hero ev or hero c-betting frequency (or c/r frequency or whatever you are studying) before you open them and study them. Similarly, it can be constructive to take a single solution and guess how turn play might play on 5 different turn cards, then view those turn components of the solution to see where your intuition was right/wrong and why.

This approach can help get you thinking more deeply about why GTO ranges are the way they are in a variety of spots which is crucial to actually being able to apply GTO directly to your game as in practice there are thousands of possible flops you might encounter at the table (and millions of possible turn situations) so you are going to have to be able to practically apply what you've learned from studying GTO ranges in an intuitive way to use it at the tables in real time.

Great study tip!

Hi all,

I'm excited to announce that today we released a new strategy pack on in position c-betting, you can find it here: http://gtorangebuilder.com/#gto-dojo.

I also put up a blog post about the strategy pack with a preview video here: http://blog.gtorangebuilder.com/

Cheers!

-swc

Sorry if this has been answered before but why would someone not opt to use Simple Postflop's free version if it does in fact include unlimited turn solutions?

The interface you've designed is tops for sure but heck if its free its free.

Hi,

If your primary focus is to run your own turn and river calcs I would definitely use SPFs free version.

The GTORB Pro license also gets you flop library access to around 800+ fully precomputed flop solutions as well as the ability to edit and submit new library calculations.

Thanks,

-swc

Or get a one time licence to piosolver. I like the GTORB setup as it is very user friendly, however for the price of two video packs you can get a fully functional product. Perhaps you need to think about a new funding model.

My strategy packs sell quite well and I put a ton of work into each one and that balance means that I am well motivated to create more of them and I believe at the same time that most of my customers feel that they get good value from them. To me this is a good outcome and is how the free market is supposed to work.

The strategy packs are designed to give the viewer an amount of information that is equivalent to many many hours of individual, independent study and to give people guidance and focus in how they can use equilibrium calculations to improve their game, all while costing much less than an hour of private coaching from most coaches who coach midstakes+ players (including me). I consider myself an excellent coach and I would still say that many people who want to improve their c-betting would get more value from my IP C-betting strategy pack than they would from an equivalent $ amount of private coaching from me.

As an example, in constructing my last strategy pack I ran well over 100 CO vs BB and BTN vs BB scenarios with a wide range of sub-optimal strategies for each opponent to exploit, I then studied the examples, picked out the 30 most clear and illustrative ones, and made over two hours of video explaining the key take aways and how to apply the results to your game. Constructing all those game trees / ranges / exploitable strategies / etc and thinking carefully about all the results and organizing and presenting that material is a huge amount of work and is not something that is suddenly instantly free to someone if they buy an equilibrium solver.

The strategy packs are designed to complement other software, they are not designed to be a replacement for SPF / Pio / CREV / GTORB. In fact I highly recommend that my viewers go out and run their own equilibrium calculations in conjunction with viewing my strategy packs with whatever software they choose, many of the GTORB flop library submissions from users are modifications of calculations from strategy packs. A big part of my goal with my strategy packs is to give people with access to their own flop calculations an idea for new high EV impact scenarios that they should explore and a good theoretical framework for how to study and analyze a type of situation so that they can go out and use those software tools on their own to further explore the topic that I present.

Clearly something like the book the mathematics of poker has value even if I already know the mathematics required to derive their results. The book presents interesting specific uses of the mathematics that I might not come up with on my own as well as an example approach of how to analytically use mathematics to study poker. Similarly I believe my strategy packs have value regardless of what existing software the viewer owns because it presents specific solutions of interest with in depth analysis and provides a template for how to use software tools for independent study.

Obviously some people may feel differently and I don't expect everyone to buy my strategy packs, if you don't think you'd get value out of one then of course do not purchase it. But suggesting that my content would not have value if you owned CREV/SPF/Piosolver is not accurate I believe. Regardless, its a free market, I can make my content and price it as I wish and customers can decide if it is worth purchasing or not as they wish.

I respect your opinion and of course you are free to not buy my strategy packs but I also ask you to respect that other people may have their own opinions on the value of my work and that you let the free market take its course.

Fwiw I think your cbet defence pack is excellent value for money and one of the best theory videos period, and at no point would I suggest that you didn't put the serious amounts of work in.

I think the problem for me is then I want to tweak ranges bet sizing etc I'm locked in to set ranges and the yearly subscription still doesn't provide a standalone interface., but I'd rather use your interface. The other issue is I like to go places where there aren't any Internet connections (eg café) to study and then I have no way of using gtorangebuilder which is a problem.

I think your comparison of an hour of coaching vs the asking price of a strategy pack is a good one, and I personally would still buy strategy packs the same way I'd get coaching from time to time, but not having a tool for my personal use is a problem, and as I said I'd rather pay for yours, but not in its current funding model.

Or perhaps you don't need to do this and that's fine. GL all the same :-)

Got it, this makes total sense to me and I understand exactly where your coming from. For what its worth I am trying to make some steps in that direction but I don't think at the moment that making a download and install solver is feasible.

At the moment the closest thing to a solution for people who want to run their own flop calcs offline while still being able to use the GTORB UI is that the latest version of Simple Postflop's desktop software will let you compute solutions offline and whenever you have an internet connection you can upload solutions into GTORB and view them in the GTORB GUI (you can also share the link with other users). This feature is in early beta and is not 100% reliable yet but we hope to have it working solidly and to add more features to the integration (based on customer feedback) in the coming weeks / months, including the ability to interface with their cloud version.

This is an example of a solution I just ran with SPF, for some reason the nash distance is not uploading properly (it was 0.004) but it should serve as an example: http://gtorangebuilder.com/#share_sc...9d2b/root_v=31

At the moment this feature only requires an SPF license, no GTORB license is required although that may change at some point in the future.

What's the main difference between GTORB and crEV?

GTORB is 100% focused on equilibrium calculations (GTO and minimally exploitative) whereas CREV is an overall strategy simulator. Because of this GTORBs UI / toolkit is optimized for equilibrium focused tools (automatic tree building with GUI customization, easy to browse the different strategies that emerge on every run out, clear presentation of nash distance, etc), while CREV can do a lot more in terms of letting you simulate playing manually entered strategies against each other.

At this point their is a beta of an equilibrium tool in CREV and its possible that when they move it out of beta it will be as powerful / user friendly as the existing equilibrium solvers.

GTORB is also web based rather than installed and everything is stored in the cloud and shareable so that is a bit of a difference.

Hi all,

I just wanted to announce that we have ended the private beta of the GTORB Flop solver and it is now open to the public. A quick video on how to use the new functionality is here: https://www.youtube.com/watch?v=SZHPj7OiuVE

You can now easily edit strategy pack and library solutions and tweak ranges / bet sizing / etc and resolve them on demand. You can also of course create your own flop scenario s from scratch. All computation is done serverside so there are no CPU/Memory requirements for your PC and you can submit multiple calculations at once and have them solved in parallel. All on demand flop calcs are private (unless you share them via the share link).

Cheers!

-swc

I am little confused about the positioning of your product / your position.

I see you promoting a rival piece of software, Simple Postflop, which appears to have far superior functionality than GTORB does at this present time.

Are you guys now working together? Does GTORB use now Simple Postflop in anyway? Why would one buy GTORB rather than Simple Postflop?

This isn't trying to knock you / your product. You were first to market, but I am genuinely confused now about the point of GTORB, given the competition in the market offers a lot wider functionality and you are actively endorsing / promoting one of the two leading pieces of software.

Hi,

For running your own calculations on your own PC the SPF standalone version is definitely the best tool and is what I use personally so I recommend it to anyone who wants to run a large volume of calculations and their own PC. If you want to do something like solve 1755 scenarios on every possible flop and do aggregate analysis, GTORB is simply not the right tool for the job and that's not what its designed for.

However, I think GTORB is already at least as good or better than any cloud based solution offering at the moment and it will be improving in the coming month or two to be a clear leader I believe. GTORB will let you queue up large numbers of solutions that will be solved quickly in parallel in our queue, so you can for example submit 10 large flop game trees (that might each take 15 minutes to run on an 8 core PC) all and once and have all 10 done and ready to view in 30 minutes, which is not something any solution that runs on your own PC can offer. GTORB also lets you run solutions without slowing down your own PC or worrying about your PCs memory limits. Existing standalone solvers are not very useful on a low end PC.

The only other product that will run solutions in the cloud is SPF and their cloud version has similar pricing / performance / features to GTORB while many of our customers prefer GTORBs interface, sharability and accessibility. Furthermore, the GTORB solver easily integrates with the GTORB solution library and strategy packs so that you can load solutions from those sources into the GTORB solver with a single click, edit the ranges / bet sizing / strategy locking / etc and resolve and then share your results with the community / a study group / etc with a simple link.

The ability to edit and resolve strategy pack solutions was my biggest reason for releasing the GTORB flop solver as I believe it will allow our community of strategy pack and library customers to interact in a much deeper and more active way with our training content which is my primary focus.

I hope this helps,

-swc

Thanks for the detailed response.

Given you recent promotion of SPF and lack of activity on 2+2 thread, I was wondering if you were going to be stopping development of GTORB or merging with SPF, but from your answer it doesn't seem to be the case (or at least if you are it will be behind the scenes).

No problem, like you said GTORB is independent and will remain so for the forseeable future

Hi, can one of your softwares do that? http://forumserver.twoplustwo.com/45.../#post48900529


thank you

Hi,

No GTORB does not do any preflop calculation / equity calculation.

Sorry!

-swc

Hi, how strongly ranges differ when playing out of position on the flop? 18bb stack

They differ quite significantly, position is a major factor and it will change flop ranges and EVs a fair amount which of course results in substantially changed preflop play as well.

You can experiment with running the same flop scenario and just flipping the "hero is in position" setting and see the strategy and EV shift that happens even when all other variables are held constant. This is done in several of the strategy packs or you can do it yourself with the GTORB flop solver.