20:55 so here i think we cant do it at 100% frequency thats my guess, we have i think too many such combos like Jx Qx 7x I feel like without looking it up, but thats my guess.

25:20 thats just a value jam vs the shortstack, definitely not a bluff. He has the Qh not Ah if that is the spot you mean?

oh sorry very different spot yeah. I agree this hand having Ah he should check back a lot/all the time? I think all the time, he has better bluffs indeed and still a little bit of showdown value.

Blocking 98 is good ofc, all other A9 combos can bluff all the time maybe and then Q9s J9s i wonder since they block the QJ which is bad but also block 98. Will look this up later. Would assume all non heart 9x bluffs at least most of the time maybe always, but not sure. AQ AJ we probably check more and rather bluff the 9x. Then also wonder about small pairs we probably bluff some on turn sometimes and then same river like sometimes bluff. But just my guess off top right now about the bluffs. Will look later

so actually looks like Ah still wants to bet a lot except with the 9c as a kicker. The reason is he has more Ah combos that call for value than ones that fold. AThh and ATo with Ah. Opponent did not get to call with too many Axhh combos on the flop, he had to fold a bunch already then checkraised some. But overall the bluffs come indeed from 9x, some 8x, 44 for the pocketpairs that got to the river, and then which i did not think of, A2s-A4s the combos that got like this to the river.
His specific combo in solver is supposed to bluff always. These damn blockers and unblockers, always confusing :D

let me post just the bluffs as it is hard to see in the pic dont know if this is etter

also noteworthy here: With his sizing he can range c-bet the flop and it is fine and simple. However solver wants to play rather 41% bet strategy on a structure like this and then mix in checks. We just want to put more money into the pot with a lot of hands on a favorable flop like this. I used here 50% of field left ICM and then its postflop ICM too, to simulate an MTT scenario.
But as a human to just simply bet smaller and therefore be able to bet range and make no mistakes plus know exactly our combos on turn, could still be better.
And I am not even sure why solver wants to bet larger, I just compared the EV for both strategies, we actually have slightly more EV with the 25% sizing and range cbet...Not sure then why it wants to bet 41%? In any case, even more reason to think 25% range cbet is better for a human.
Also, we can of course mix these two sizings but right now I am all for simplification, I do not see a good reason to complicate the game-tree for us like that on flop