Quote:
Originally Posted by inneedformoney
well in that case we should do our postflop game from the database we have
The sample size for any postflop spot will be tiny. We've actually looked into this a few years ago for a different project. As opposed to this, the sample sizes for preflop spots are huge, especially for heads-up players.
Quote:
Originally Posted by inneedformoney
... different unusable hand percentage like 13% 42% 88% ...
Also would be great if hands was rounded to 0 50 or 100% in outcome, to use ranges more efficiently...
The exact frequencies are not something to obsess about. The most important thing to take away from a mix is that there is a mix in the first place. When a mix is used, the EV for both actions is identical and the player is indifferent between them. Rounding to 0% or 100% loses this information. Other than that, when trying to interpret large data sets, the individual data points are not something that should be altered. The trick is to try to find patterns in the data by using statistics, graphs, etc. However, the exact frequency of an individual combo will not be too relevant, given that a lot will be lost anyhow when translating the outcomes of GTO solving to someting that can be recalled and used in live play. Rounding accomplishes nothing in interpreting large data sets, and in fact only loses information.
Quote:
Originally Posted by inneedformoney
... and that what is bothering me... Have can we work on our postflop game if we dont have good starting point on preflop...
You should use your historic data for this, given that it actually applies to your own play and the play of your average opponent. As I mentioned earlier, a GTO solution won't really help you here. It will make you play hands preflop that you don't know how to play postflop, and vice versa, it will make you fold hands that are actually +EV in your own games. And, once again, preflop GTO only works for heads-up play; not 6-handed, not 5-handed, not 3-handed, etc. Just heads-up. And it only works if you have a computer with roughly 100GB of memory.
Quote:
Originally Posted by inneedformoney
And because you say that this can be easy made - why not to do this?
It would take roughly 3 months of development time, which we could instead use to develop original features of our own. And all we would wind up is a feature that already exists in about 3 or 4 other products already, where we could have used the opportunity to do something original. Other than that, once again, almost none of our users will actually be able to run this, or have any practical use for this.