Two solvers totally different solutions in simple spot
I have been on a dinner with two friends, one working with runitonce and the other one with plomastermind and they discussed preflop play for 6max PLO.
Who is wrong, or are both right?
Situation:
BU AAxx facing a 3bet from SB
runitonce:
Call 43%, 4bet 57%
Plomastermind:
Call 6%, 4bet 94%
6 Replies
Are the rake settings the same? I’m not familiar with the default, but ranges are typically influenced by rake a whole lot
Sent from my iPhone using Tapatalk
Should be the same. At RIO it‘s 400-600 and mastermind 500
And the 4bet range is 5% and 8%
Those are not pictures of a solver. They are pictures of a -library of solves-. It's different.
You'd need all constraints and how much it solves down to to be identical for you to really compare. It's also really likely that however people are solving PLO they are doing a lot of abstractions/bucketing, so that would also have to be the same.
All preflop and postflop allowed sizings would need to be identical
Rake settings would need to be identical
Abstractions/bucketing would need to be identical
How far it solved/exploitability would also need to be identical
A solver is really just a computational modeling software... It's computing the nash equilibrium given a set of constraints (model), if the constraints are different between two solves, then you're not solving the same model.
Thank you for the explanation. I already thought that the assumption made by the two versions have to be different.
I just thought that the models should be close together for a simple preflop spot like 4bet range BU vs. SB.
I didn‘t think that the variation of the abstractions made by the creators are so big, that they find totally different results.
What do you recommend - find a library that I like more/suits my playing style more, if I want to find a solution for a certain spot?
Thank you for the explanation. I already thought that the assumption made by the two versions have to be different.
I just thought that the models should be close together for a simple preflop spot like 4bet range BU vs. SB.
I didn‘t think that the variation of the abstractions made by the creators
You'd have to look at their sites or contact the developers to see what they used to solve these libraries and what the constraints were.
Even if the parameters are identical, there usually many strategies that satisfy the accuracy thresholds, so different solvers can give you different strategies.
In practice, we don’t solve to perfect accuracy but to a threshold, often defined as "the strategy must lose no more than 0.3% of the pot against a perfectly exploitative opponent." This threshold, known as Nash Distance, introduces some wiggle room, allowing freedom in how strategies develop.
In my own experiments, I’ve observed that solvers with identical parameters—same ranges, bet sizes, rake, and accuracy thresholds—can produce vastly different strategies. For example,
, I solved the same spot using three different solvers, and each found a unique strategy:
These differences arise because multiple strategies can satisfy the same Nash Distance threshold. It doesn’t mean one solver is wrong—just that there are several ways to achieve comparable exploitability.
In your case, while we can’t confirm if the parameters are identical, PLO’s additional complexity often necessitates even more relaxed accuracy thresholds, widening the range of acceptable solutions.
