I remember this being a surprisingly effective tactic that Libratus was using during the heads-up challenge with Polk + friends a couple of years back. IIRC it was mostly on the turn, but I think it would potentially be more effective in certain flop situations. However, I don’t recall any explanation of why it was effective at the time, beyond “How the hell do you play against that?”

I’ll try to reason through the incentives to massively overbet. The first thing that occurs to me is that this would be better OOP rather than in position. There’s not much difference in how often we see the river if our draw retains equity on most turns but we’ll have more ability to pot control in position, and better opportunities to extract value when our draw completes. We’ll also be more likely to find profitable bluffs. Drawing to non-nut flushes should also have less reverse implied odds when we act after our opponent. The EV of playing later streets in position probably outweighs the fold-equity + guaranteed equity realization of making a massive flop bet.

I think a huge part of what could make this tactic preferable to other options is precisely those reverse implied odds. In the example of 65s on T42ss, I would be very hard-pressed to call massive overbets with hands as strong as NFD+2 overs. Could be folding all of them, maybe calling A3s/A5s but likely folding even AKs. Although I’d expect to be a favorite over some parts of a massive bet range, I’d be screwed if someone was doing this with half their sets as well as their 12+ outers, unless they were also doing this with some 1-pair hands (a la Mike “The God” Postle). Folding out draws that will often stack us when the flush completes is obviously a huge win.

I want to get into some EV calcs, range construction, and also talk about the merits of smaller shoves when holding draws but I don’t have the time or brain function to do so atm. To be continued…