Thanks, glad you're enjoying them!

Yes I did allow the solver to play a donking strategy on the river in the triple barrel line in all scenarios. What I've found is that the solver does relatively little donking on the river after IP has double barreled. On a runout where the nuts on the turn is a flush and the river remains a flush, there is obviously no reason to lead because IP has a polar range.

However even on a run out where a flush is the turned nuts and a full house/quads is the nuts on the river, OOP still doesn't donk very often at all. This is because the solver does a better job selecting it's turn double barrels than humans do. The solver includes more sets (and weak two pair as bluffs) so large range asymmetries are rarely created. In practice, donking these spots vs humans might be very effective if they construct their turn ranges poorly. In solver land this doesn't happen.

Regarding the question of which flushes to fold: There wasn't a clear pattern which is why I lumped them together. Sometimes you just have to play mixed strategies in theory. In practice, I would just always fold the lowest flushes which is exploitable (villain could theoretically use specific blockers to exploit this strategy) but this will never happen in human play.

Nice vid!

Out of curiosity, how can villain use specific blockers to exploit the strategy of folding the lowest flushes?

I believe that the clairvoyant bluffing player will be able to select slightly better blockers against a strategy that say folds 7 high flushes or worse 100%. Now they really don't want to bluff with a 7d-2d in their hand. That should allow the clairvoyant player to select a little more intelligently from the pool of possible blocker hands.

It also might be possible that knowing I fold 7 high flush or worse 100% allows you to alter your value range. For instance Q8dd might be a worse value bet than Q7dd now.

You're correct, QJ indicates QJ[T-2][T-2], but excludes AQJ which was already covered under the AQ category.

On the charts themselves I use "ds" to indicate a double suited hand and "ss" to indicate a single suited hand. "r" indicates a rainbow hand. dsi indicates a hand that is double suited improperly which just means (AK)(52) rather than a properly double suited hand (A5)(K2) or (A2)(K5). I just found that parenthesis were too difficult to look differentiate when looking at a long list of hands.

Hey Sean,

Yeah you're interpreting the charts correctly. The only other thing I might have been trying to show was the weakest 8 high hand and the weakest 9 high hand with a pair that is opened, but that doesn't make sense in this context because if we're opening 8655ds, then we are also opening 9766ds.

I imagine that 8655ds is the weakest hand the solver is opening here, but as long as you get close to opening these hands, your frequencies will be much better than the competition.

Thanks for pointing that out.

Really glad you're enjoying them!

Yes, this is exactly what I've observed. While it's always better to make middle set than bottom set, with SPR=4 or similar stack sizes, the opposing player as a lot of incentive to stack off with hands like top two, top and bottom with redraws and other combo hands. Against ranges like this, even 22 does really well on J92ss.

At 100bb, if we x/r 22 on the above board and get called, we're going to have an unpleasant experience on most turn cards.

One caveat is that the these preflop ranges are generated by monker with an abstraction. The abstraction doesn't seem to be greatly impacted by the postflop bet sizing I've given it, but there is always a small chance that either I or the abstraction is causing the preflop ranges to be skewed towards these hands. That being said, the logic of the first paragraph seems pretty airtight to me.