Hey, the Lucid Poker solver 4-bet jams in many situations, but GTO Wizard and peakGTO solvers dont take this action, so I'm confused should I learn it or not? Maybe its better to incorporate it so I can respond if someone does it.
But the more important thing is how come two solvers have different outcome?(I know you a human determines the actions and bet sizes a solver can take, but I assume GTO Wizard engineers or others know about this action).
The perfect man to answer this question quickly posted a reply, racking up many times more upvotes than the original post.
Hi, Mike Brady from Lucid and Upswing here. I can ramble about this.
First, I'll talk about this 4-bet shove with . Why the solver does it. Why you maybe should or shouldn't do it in practice.
Then, I'll talk about your bigger question (why two different solver outputs can come out looking differently).
Remember throughout all of this that the solver build strategies at equilibrium -- i.e. it assumes both players are omniscient and play perfectly.
The Ace-King 4-Bet Shove Mechanic
I think this strategy will resonate with a lot of people. I'll explain why with an example that most of us will have experienced.
Cash game. A player raises in late position. You 3-bet from the blinds with offsuit. He calls.
The flop comes three cards with no or . Maybe it's a good flop for your range ( ) so you c-bet, maybe it's a bad flop ( ) so you don't.
Either way, it can be a bit of an uncomfortable spot. It's so easy for your opponent to (correctly) put you on two unpaired high cards, and more specifically (and live players love putting their opponents on ).
There's not much you can do about it here. You obv made the best preflop play by 3-betting. And it's not a huge deal overall, because your 3-bet range isn't super tight. You'll usually have plenty of other hands that actually do hit whatever flop came.
But this does become a legit problem in 4-bet pots, where ranges are much narrower.
Suppose I raise from UTG (6-handed, aka LJ) and the next player 3-bets. It gets back to me and I 4-bet.
No matter what solver you look at, you're going to 4-bet a pretty narrow range of approximately 50 combinations here.
Ace-King (offsuit and suited) are 16 combinations of that. If your opponent calls and the flop doesn't have an or a , a third of your range has missed the flop -- and both of you know it. When the flop does contain an or a , your opponent can easily put you on and get away from fairly strong hands by the river.
This is a problem (both in practice and in theory). The solver has a couple of ways of negating it.
One option is to just call with versus the 3-bet. You will see solvers take this option in very tight range spots (such as UTG1 vs LJ at a 8-handed table). Here's a screenshot of Lucid's live cash sim (150bb, no rake) showing offsuit and suited both call at a solid frequency:
Occasionally, you'll even see the solver mix in some folds (only in VERY tight range spots).
The other option is to avoid the possibility of postflop play altogether by shoving all-in.
You make this play with a balanced range -- usually including , plus some combos of medium-big pairs. This way, your opponent can't just always call with a pair knowing you have (because they'll be crushed by your bigger pair sometimes).
Here's a spot where Lucid does that -- 6max cash, LJ vs HJ 3-bet (I've highlighted the shove range in yellow). Rake structure is 5%, 1bb cap by the way. GTO Wizard's most similar solve (500NL rake, same positions) make this play at a very similar frequency.
ReadYou'll find this in loads of other spots too, including when 3-betting all-in in tournaments when you're at 40-50bb stacks.
It's a cool play that I think performs very well against good players who are 3-betting with wide ranges and playing tough postflop.
But this play SUCKS against nitty opponents who don't 3-bet nearly as much as they should. You're just pumping too much money into the pot if your opponent's range is + and himself. Punish those players by just calling (or even folding) with vs their 3-bet. Folding is a weapon against nits. Don't give them action.
Why Solvers Have Different Outputs
It comes down to three things:
1. Inputs.
Open size. 3-bet size. 4-bet size. The rake. How many cold calls are allowed. You have to specify all of this before running a solve, and small changes (especially early in the game tree) can have big impacts.
2. How long you run it.
The longer you run a solve, the less exploitable it gets. At Lucid, we run our solves down to absolute minimal CI (convergence index -- nerd term you don't need to worry about). In laymen's terms, the longer you cook a sim, the better the output.
3. The software you use.
At Lucid, we use CFR solvers. Commercially available ones like Hold'em Resources Calculator and PioSOLVER on privately rented servers, plus a proprietary GPU-based solver we'll soon be releasing for commercial use.
These are the types of solvers that have built top players' strategies for the last 10 years. But in the last couple of years, AI solvers have come onto the scene. Maybe GTO Wizard uses those in their cloud sims. Tough for me to say for sure.
Overall, a solve is most useful if you understand the context and the why behind it. So, when you spot a solve that's different in two places, I'd suggest digging in. Check the raise sizes.
See if there's a rake difference. Sometimes a simple input explains it -- and that will help you broaden your understanding of the game ("oh, if they 3-bet bigger, I 4-bet shove more often").
I checked the spots you mentioned in another comment in both GTOW and Lucid (LJ vs HJ and CO vs BTN) and both have the 4-bet shove with as part of the strategy.
So, I can't speak to exactly what you were seeing since I couldn't replicate it. But hopefully this novel was helpful for you anyway. GL
The subreddit's members thanked Brady for such an extensive response, which easily could have been its own blog article.
— Man, I appreciate your well elaborated answer, and thankful for the whole team who is always making us better. Also, keep up your amazing videos bro (and yeah, the AK 4-bet jam is a good idea for a YouTube video).
— Lovely response. On a side note, are different open sizes coming to Lucid poker soon? All the solves sims are for 2.5bb open which isn't super common in 6max or hunl these days.
Yeah, for 6-max we've got a few different sizes in the internal version (testing phase).
Will be available in your app soon.
— Not OP, but I really appreciate the writeup and the content you make.
Thank you! Might turn this writeup into a YouTube video for Lucid.
A man of his word, Mike Brady really did make a video a week later.
"By the end of this video," he says, "I think this crazy looking strategy is going to resonate with a lot of people."
If you're looking to upgrade your training, you've got plenty of avenues to take.
1. Pay for the full Lucid Poker subscription or browse powerful discounted solvers like DeepSolver.
2. Let our specialists recommend a solver that truly suits your format and stakes.
The GipsyTeam Shop always carries software at the internet's lowest prices. We've been helping players succeed since 2009, and the right tools are part of their arsenal.
Equip yourself, fix leaks, and level up your game.