Thanks
Quote:
What's the best way to model a BTNvCO preflop spot with the preflop solver? Or for that matter any preflop spot where the action would not be closed, say MPvUTG, or COvMP, etc.
There is currently no 100% reliable to do that. One try is to estimate how much money is lost to remaining player using stats from your db and then removing that amount from the starting pot (blinds). While this is in no way perfect it's one reasonble way to estimate the ranges.
Quote:
I'm getting results for the computed BTN strategy that just don't pass a sanity check. Things like T2s being a pure preflop call, etc. I imagine that's because I included 2.4bb of dead money (the blinds and antes) in the pot that's not actually dead.
Yeah, especially if you want calls as well it's not going to be easy to do. If you only want 3bet/fold then the method describe above is likely to be reasonable but with calls it's just not going to happen until there is multiway solver (but that will have disadvantages as well as it will no longer be possile to solve without abstractions for anything other than 3way or maybe 4way).
Quote:
Finally, is it safe to say that the # of flops we fit into our preflop tree is significantly more important than the strategies we specify in the preflop call nodes?
Yes, as long as postflop play isn't missing important EV options (at least on sizings everywhere, it's safe to remove donk bets OOP and simplify the play a lot by using cap of 3 or maybe 4) it's way more important to add more flops than to make postflop play more detailed in the preflop tree.
Quote:
Is there a limitation to how many scripts you can run at once or is it just about memory?
Just memory but just to be sure: they will run slower on average if you run a lot of them as they are going to compete for CPU cores.