So, full disclosure: I'm fairly new to solvers in the grand scheme of things and I'm not a game theory expert. Perfectly open to others with more expertise correcting me.
Quote:
Originally Posted by 2+2=5
Do I understand it correctly that poker solver runs through all possible scenarios with all possible opponents and gives the best ev solution. it does not account for a specific player, it calculates best solution against all possible opponents in all possible scenarios?
It's funny: what you're describing is sort of accurate but also sort of the opposite of what a solver does.
A solver produces the strategy that is best in a vacuum. But, in practice, it is applicable to vanishingly few players. So it's sort of giving you the strategy for everyone and no one. Because until you can distinguish a player from "everyone", you should ideally just play unexploitably. But basically everyone has exploitable tendencies. (And you can often responsibly assume a lot about players' tendencies after just a few orbits of play with them. So, in most games, you should be shifting to an exploitative/exploitable strategy very quickly.)
To clarify further: a solver will give you an unexploitable strategy for both players in a pot. (You can also do multiway solutions with some solvers, namely Monker, but my limited understanding of game theory is that beyond two players a lot of theoretical underpinnings begin to fray.)
Essentially, you pick a game and you pick a street (and board cards, if appropriate) and then assign ranges to each player (or don't, if you're starting from nothing preflop).
The solver then creates and iteratively adjusts both players' strategies until neither player can unilaterally increase their EV against the other. (In practice, usually you need to define a point at which the solver stops trying to find small edges, otherwise it can fine-tune for a long time.)
As a result, what you are left with is two unexploitable strategies: what both players are "supposed" to do in the given situation, with their ranges and the board as it is.
As you can maybe imagine (and this is an important part that I feel like a lot of hashtag GTO bros are NOT forthcoming about): you actually need to do a lot of guesswork at the ranges you're starting from. There are still a lot of assumptions built into solver-based study, including an assumption of no assumptions (which, in fact, assumes that both players are playing unexploitably).
Beyond stipulating ranges to start a solver sim with, node-locking is another way that we can make assumptions and start to find exploitative strategies.
I think an example may illustrate it best:
Yesterday, I was looking over a NLH hand I played using a NLH solver.
It was CO (me, RFI) vs BB (call), and I used what little data I had on BB to stipulate their preflop range, and I used what I know to be my typical CO RFI range. I then plugged in the flop that we played, and I ran the sim.
I then went street-by-street based on how the hand played out.
Check/check on the flop sent me down a particular path of the sim. The combo I actually had was checking back the flop some percentage of the time, so that's good.
Villain bet the turn. The solver sim has them doing that sometimes on the turn card we saw, so that's fine too. I raised the turn with my hand, which the solver sim also includes in its unexploitable strategy. Villain calls, which they're doing sometimes in the unexploitable strategy the solver produced. So far, so good.
The river comes off. In real life, villain had bet.
In the solver sim, villain is betting literally 0%.
Uh oh.
So, what I need to do here is node-lock. Minor spoiler: I won with a raise on the river and no showdown, so I don't know what villain bet with. So I need to look at their river range and say "okay, I am locking this node so that they are betting this hand, this hand, and this hand". I'm literally guessing.
Once you have node locks in place, you run the solver again. The solver now looks for unexploitability with the caveat that it MUST include the strategies that you've node-locked for. In my case, villain MUST be betting x/y/z on this particular river.
Once it's done, you go back to the river and see what the strategies look like with this edit in place.
That's node-locking. It's basically computing the most optimal strategies possible given non-optimal constraints.
But these constraints need to be manually edited in. And the further your opponents are from unexploitable play themselves, the more node-locking you need to do.
And I think you can see from my description above that, even for what was a fairly simple river situation (I didn't go into detail, but it was really pretty simple), it can all become very time-consuming.
I hope that made sense. I'm very caffeinated as of this writing, which can do strange things to my writing. (If nothing else, it makes me write a lot.)
Quote:
Originally Posted by any four cards
lets not forget about propokertools
Seriously, PPT's free web equity calculator is probably the single most valuable tool for improving in PLO. Getting a feel for to what equity you can expect versus various ranges in various situations is vital.
If you're serious about PLO and don't know PPT syntax like the back of your hand, learning and using it is key, imo.
I think that, in PLO, the main things we need to grasp to be successful are (in this order): rake (don't play too many hands), equities (try not to get all in with the worst of it so much), balance (find some bluffs so you can get paid when you want to).
A solver will keep you from making -EV preflop plays WRT rake (as long as you entered the rake correctly) and it's obviously going to be fairly equity-guided in its strategies, but the whole point of a solver is to create a strategy that's so balanced that it cannot be exploited.
And when balance is of tertiary concern (and can be achieved to a workable level without a solver), I just think solvers are not all that helpful to most PLO players.
