Fish Clint Eastwood

I'd personally block or 75% on the turn to get called by all hearts, SD, and pairs, and maybe random overcards.

Not sure what kind of fish this is or what you're considering doing on the river since there's no info given.

The precise spectrum of FISH CLINT EASTWOOD is unknown at this juncture

I can't tell you how I played the hand because I want you to tell me what you'd do

EDIT: actually it doesn't matter, I called

Seems pretty standard fold on such texture changing turn. Both ranges become uncapped and river sizing underbluffed. Raise doesn't make much sense as we try to make him fold trips+

I think the top 5 hand combos in villains range are:

A5o: 7.5%
65s: 5.7%
A5s: 2.2%
A9o: 2%
54s: 1.9%

Villain probably has trips 19% of the time and a flush 26% of the time, and a full house 10% of the time! So we are crushed 55% of the time.

However, they are probably bluffing with nothing 23% of the time and will have a worse two pair than you about 9% of the time.

So overall, you can probably expect to win with a call roughly 31% of the time. But given the river bet size, you need to win 36% of the time.

So this seems like a slightly losing call IMO.

I was also curious if maybe you could bluff raise the river? So I tried different raise sizes for you, but they literally underfold to every size so as an exploit would be a good spot to raise for value.

In fact, if you jammed for literally 94bb on the river, they would call you 35% of the time! If you have a full house then you would win at least 90% of the time too, so a jam on the river would be worth almost 51bb in EV.

I would fold river. Just think that sizing is rarely a bluff at these stakes mate

Agreed. Jamming would be a huge losing play,

Fish defend more than MDF. So The only time you rebluff fish is when they have an air heavy range.

This sizing is underbluffed so I would fold as well,

I don't disagree with folding really. He definitely has all the fives.

Some reasons I called in game:

- most fish bluff most FDs on the flop, or at least the kind with river OBs for value in their arsenal (so as usual, incongruency)

- BU v BB. Wiiiiide range etc

- as a fish he is, in all likelihood, horrendously unbalanced in this line and i have a strong blocker

- I was feeling lucky

I'll post results later unless anyone else wants a crack

Out of curiosity, why do you think villain has zero flushes?

I have a predictive model that estimates villain will actually have have a flush 26% of the time, and will have trips or full house 29% of the time when they bet this size on the river given the action so far.

ok. I'd fold against a fish when they ~50 turn and overbet a bricky river unless I know they're some kind of aggro fish that has a high PFR. I don't think an average fish is going to show up with much mergy stuff or worse 9x here as PFR and IP.

You built your own predictive model? Sounds awesome.

Can you tell us more about it? Your analysis is very good from what I’ve seen.

Yeah I did! Thanks for the kind words!

And absolutely, I have been working on it for a while now. It is a set of large neural network models that were trained over hundreds of millions of hands to predict what players will do and their ranges.

I'm trying to figure out the best way to make use of it and turn it into a useful tool for others, so if you have any suggestions then I'm all for it!

So far it's been really useful for understanding what spots are people overfolding or underfolding, what is their range when they bet or what hands I should be value betting based on how often I'll win after a call, etc.

I've also tested it on many new hands its never seen before and it is very accurate.

You can also give it a set of past hands you have on a player (or your player pool) and it will customize its predictions to that specific player based on how they played their prior hands.

Sounds exciting. Is that different than MDA? Or are you saying it goes further in how it processes range assumptions?

I suppose having any kind of interface for MDA decision making would be pretty awesome

I would say it's much more accurate than MDA based on my tests and is different. It's essentially a large model that was trained to be able to predict player actions and ranges in any player pool or situation.

For example, let's say you're player 50NL and button has an 85bb stack and opens to 3.5bb and you call from BB. Flop is king-high with a flush draw, will the button overfold to a donk bet here?

If you were using MDA, you would need to try and find many hands with similar circumstances with a "shorter stack" villain with a 3.5x open from button, and then a king high board with flush draw and donk bet? Even if you have millions of hands, that situation might have only occurred a few times and will not give enough accuracy on their range or their fold frequency.

Also, MDA is limited because it can't take in the history of a player and make predictions about them. For example, let's say we see a hand where villain calls down bets to the river and shows down with a hand that should have folded flop. Well my predictive model could see that hand and adjust its predictions for what that player will do, etc.

Sounds wild squire, I'm all for it.

HAND RESULT:

**************

PLOT TWIST

FISH CLINT EASTWOOD wasn’t a fish… It was actually the real Clint Eastwood all along making a not too awful bluff given common thread responses

I instinct fold normally but then I remembered something Unguarded said once about fish overbets on the river in weak lines being generally FOS so i looked him up. I think what FreakDaddy said is true.

Does anyone have percentages on certain pools based on how many players are
a) shitty face up regs/ thinking recs who simply bet turn and river because they now have value
b) players who try to play reasonably theoretically
c) players who realise this sort of node is overfolded and just do this with range

That sounds like the future of poker because MDA is flawed in that it needs past hand histories to predict future events. And you need many samples to have any type of accuracy. Some spots are so rare that they almost never happen so we just end up making pure guesses about them.

As far as how to get this information to the public I'd suggest going the route of GTO Wizard and make it available to everyone by having them pay a monthly subscription.

For example, I looked at this hand and knew X-B-OB was underbluffed, but the sample size in the database was still only 100 hands so the margin of error will still be a significant amount.

Awesome work and I'm going to private message you to get more information on this project.

X% of the time he's betting the flop w/ a flush draw, and most of the time if he's not betting the flop, he's not over betting the river w/ a flush considering the action and paired board. He can't possibly be expected to get paid w/ that line.