Uncomfortable calldown

You are facing a 30% pot river bet, which means you only need to win showdown 23% of the time to be profitable assuming a 5% rake.

I think this is definitely a profitable call. My predictive model estimates that the EV of calling is 18bb, and that your hands equity is 40%.

Villain is likely to have a high card bluff 25% of the time. They have hands like KT, AK, AT, T9s, etc.

They will also have a random underpair 6% of the time with 99, 88, 77, etc.

They will also have worse two pairs than you 9% of the time with hands like Q9, Q8, A6, etc. We even chop with QT 1.6% of the time.

The top hands in villains range are:

KK: 16%
AA: 12.3%
AQ: 10.4%
AK: 7.4%
KQ: 6.3%
AJ: 5%
KJ: 2.7%
KT: 2%

Overall, I think it is a very profitable call, and folding would be a huge mistake on the river IMO.

Cheers Ben. I'm not sure I can get behind this one though:

Just seems so unlikely to me a fish is able to triple these hands, particularly given thin river sizing that can't be anticipating (m)any folds.

But hey, I don't know! It could be right. Hunch says you would need incredible amounts of MDA to guesstimate OOP's range accurately here given unique sizing scheme.

Do you ever think the model is less accurate in certain spots?

Thats fair, and thanks for sharing your thoughts!

The predictive model was trained on many many hands, and it is designed to try and accurately predict in unique spots that it has never seen before.

But also, it is totally fair to say that it is likely to be less accurate in rarer spots than it is in more common spots. So in an extremely rare scenario, I would have less confidence in its accuracy than I would in a very common scenario.

However, that is true of anything. The more rare and unique a spot is, the more difficult it would be for anybody to predict what a players range is, and it would certainly be difficult for MDA as well.

So even in rare situations, I would tend to lean towards trusting the model as default because it's probably going to be more accurate (on average) than anything else we might have.

As a quick test, I looked over my own personal hands and I filtered for river call spots where a player made 2+ streets of bets and made a river bet that was < 50% of pot.

I had 181 spots, and on average my model predicted that 19.1% of the time villain should have a high card hand. When I look at their actual revealed hand, I see 33 high card hands, which is 18.2% of the hands.

TL;DR: The model is probably more accurate in common spots compared to rare spots. But even in rare spots, it is likely to be the most accurate tool that exists for predicting player ranges

Ok. Appreciate your take sir

I'd call down as well given MDA. He can have AK enough of the time. The hard part about bluff catching is if somehow you knew you would lose 75% of the time in this spot, you would still have a profitable call.

It's a shrug call for me. Anyone bad enough to min squeeze OOP is a fish/whale and is likely to massively overbluff and merge with small sizes on a bricky runout, but it really depends on what their preflop VP/PFR looks like. Could also be a complete nit and always have AQ+ too.

Easy call