Picking the Right Semi-Bluffs

As I often like to remind readers, I am not a professional poker player; I am just a writer who is well-versed in solvers. My writing at GTO Wizard is very much colored by the perspective of a student trying to plug leaks in his own game.

With that in mind, today I am writing about a big leak I fixed in 2024. My coach, Dara O’Kearney, got me on the right track initially by explaining the error of my way, and since then, I have used GTO Wizard to really hammer the lesson home.

The leak was related to semi-bluffing.

Most bluffs should be ‘natural,’ as in they have some property that increases their chances of winning, even if it is not a ‘made’ hand.

The leak I had belongs to the second category. My natural instinct was to make my biggest bluffs with my biggest draws. For example, I would only overbet-bluff or check/raise-bluff with hands such as the nut flush draw. I would rarely get too out of line with more speculative draws like gutshots or backdoor draws.

My logic was that if I were going to take on the risk of making a big bluff, I would want to have the best chance of getting there if I got action.
This, it turns out, is often backward logic. It’s probably best to jump into a hand to explain why.

Picking Semi-Bluffs on the Flop

This is an NL500 cash game example where UTG has opened, and the BTN has cold-called. The flop is J♥T♥2♣.

This is UTG’s first-in flop action:

For a hand to be considered a bluff, it needs to be weak enough so that taking down the pot uncontested is one of the best possible outcomes.

These combo draws have too much equity, and we want to make sure we realize more of it, so we check to induce bets and/or see the next street.

An interesting note about why the (club) backdoor draws bluff the most. This is because they have a prospect of profitably firing second barrels on many turn cards. If they don’t turn their straight when called, they might turn a club, which means they can bluff the turn with a straight and flush draw. And in case they get reraised on the flop, it’s easy to let go of it. Practically speaking, the times they do make the club flush by the river, they have a very well-disguised hand that most real-life opponents would discount entirely.

Another way to determine why these hands rarely bluff is to look at what folds. This is the response to the 33% pot bet:

Some small pocket pairs fold, and some Ace-x hands fold, and the rest is trash essentially. In equity terms, our combo draw range was massively ahead of all of these hands. Especially A♥K♥ and A♥Q♥, which were dominating the Ace-x and would have been very happy to turn top pair against them. The hands that fold are all hands we would have preferred to keep in the pot with our mega draw.

This was the first-in flop betting range for UTG as the OOP player. It’s a common strategy to check/raise big draws, so let’s quickly rule that out. If UTG checks, this is the BTN’s betting range:

We won’t explore it in detail because it’s very repetitive, but the BTN has a lot of the same combo draws as UTG, and they keep to the same strategy of betting least often with straight+flush draws, and betting most with straight+backdoor flush draws.

When the BTN does bet, it is mostly an overbet, and this is the UTG response:

It’s a very narrow check/raise range. All our combo draw hands mostly call (and are calling a larger bet than we would have made if we were first to bet it).

However, the few times we do check/raise bluff with a draw, we do so at a higher frequency with the combo draw that has hearts. It’s such a small part of the range that it is almost insignificant, but as the BTN has strengthened their range by raising, we need a much stronger bluffing candidate when we reraise with a draw.

The overall betting frequency has increased as the BTN has weakened their range by not betting. Let’s see what happens now with our combo draw hands:

There is still a significant amount of raising, although a bit less than there was on the flop. The same sort of hands fold but at a higher frequency—trash, small pocket pairs, and weak Ace-x.

With one card to come, this is a much better outcome for our combo draws, even more so since there are even a few hands that continue that our higher card bluffs beat. We can “bluff” with A♥K♥ and A♥Q♥ and still get called by worse straight draws like A5s, A4s, KQs, K9s, and worse flush draws.

And once again, let’s compare the checking frequencies to determine which of our combo draws bluff when they miss:

Once again, our missed flush draws check the most (i.e., they bluff the least); twice as often as the other suits, in fact.

This is a very useful heuristic to note:

Missed straight draws make better bluffs on the river than missed flush draws, most of the time.

The reason is simple: when you’re bluffing, you hope that your opponent folds. Their folding range usually contains a lot of missed flush draws, but by having a flush draw yourself, you reduce the likelihood they’ll fold. A missed straight draw, on the other hand, interferes less with the folding range, making it relatively more likely they’ll fold.

Let’s look at the response to the 132% pot bet:

KQs bluffs all the time, regardless of suit; it makes for a very good bluff. It makes Ace-x and small pocket pairs fold, which is a great result. It also blocks AKo, QQ, KJs, QJs, KTs, and QTs, which all continue in some capacity.

In contrast, 87s bluffs half the time except when it is 8♥7♥. A hand like 8♦7♦ makes a good bluff because it’s almost never ahead, so any hand it makes fold was otherwise winning. It crucially unblocks missed heart draws, all of which will fold. 8♥7♥ takes a lot of heart draws out of the BTN’s range, weighting them more towards a bluff-catcher or stronger.

A quick look at the “Blockers” tab shows us this:

The 8♥ and 7♥ are both among the cards that lower the BTN’s folding frequency by the greatest amount.

In contrast, this is the same tab for the K and Q:

In some instances, having either of these cards increases the chance the BTN will fold, and even for the heart variant, the decrease in fold likelihood is quite low.