Hey guys,

I was wondering what you all consider an acceptable amount of EV loss when simplifying strategies from a solver.

In spots where the solver c-bets at 85%+ frequency, most people will just round up to 100% and bet their entire range as a simplification. At what percentage of EV loss do you draw the line? I've seen numbers anywhere between 1% and 3%, and I'm curious what thresholds you guys use.

Thanks.

Don't have anything like a conclusive answer to offer you but I think this is a really interesting topic. Surely from a practical point of view it depends most on how people will respond to your simplification.

Like in the example you used where you're cbetting range rather than 85% of the time, if people are already raising too little anyway and won't adjust (or won't adjust enough) to your simplified strategy then you might not be sacrificing EV but actually increasing it.

great question

anyone?

I doubt the validity of the solvers that quote a percentage of exploitability, or quote a “nash distance” number. Such numbers are only estimates based on the bet size, which is either pre-built into an AI like pokersnowie, or is entirely input by the user in PIOSolver.

If you *understand* why the solver takes a given action with 85 percent *instead* of 100 percent, then you will know when to take the more profitable action.

Most (?) solvers have the option to calculate maximally exploitative lines for one player, which helps confirm the validity of the nash distance.



I don't think anyone can tell you that losing 5% of EV = 0.05bb/100 or some such figure. Some spots you study are also more valuable than others. Basically, if anyone is to give you a number, it's probably just an arbitrary figure because no one would know the implication of losing 5% of your EV, nor is it practical to know.

What I would do is look at the EV of my whole range. For example, you call a BTN raise from the BB in a game with no rake. the EV of your worst calling hand should be = or >2. If your simpler strats cause one of those hands to have an EV <2 ...then you have some work to do.

P.S. this is assuming that the BTN raised to 3bb pre. So BB is calling 2bb preflop. So if your hand makes back <2bb postflop on average because of a simplifying strat, you have to adjust somewhere.


Also, remember that your worst calling hand will be in many different situations. A hand like 65s will obviously make near 0bb on a flop of AK7, but will fare better on 743. Does it make back your pre flop call on average? You'll have to find out for multiple flops.

Any solver that is calculating a maximin solution against a particular player is only as reliable as the data you input for that player. The more quickly the data converges, the less useful the data becomes.

Regardless, GTO is not interested in any exploit, no matter how reliable the data. GTO starts and ends within one hand of poker. When there are only two players dealt, or the hand has become heads-up, then GTO can not lose money in the long run. There is no way to exploit a GTO solution heads-up, and any solver that purports to teach otherwise is perpetuating an endless circle of self-leveling.

Why?

.

I'm not sure if OP is actually asking about a particular setting to use for 'Nash distance', so I might be missing the point, but if we're just alking about simplifying the strategy in general...
I wouldn't be particularly interested in the EV of my entire strat. I'd look for combos that have pure strategies (as opposed to mixed strats) and try to make sure I make the correct actions with those. e.g. If the solver says "C-bet 85% of your range" and some of the hands use mixing, I'd look for the few combos that should always be checked. After all, if something like QQ is a "must-check" on Kxx, then betting that hand is what would drag down your overall EV. You can look up how bad it is to bet combo XX, and compare it to the EV of checking it.
In any spot where the optimal betting frequency is <100% and some combos should be pure checks, there will always be some combos that are the worst c-betting candidates. Your job is to find them.

If it turns out that every hand in the range in a particular spot can/should be bet at some frequency (it's just that mixing brings it down to 85% overall) then it's not going to be a problem if you go with 100%, since - in the vacuum of that one hand - you'd still be playing GTO. The only EV mistakes you can make are when you bet with hands that should be 100% checks, or you check with combos that should always be bets.

It's still going to be a problem, since your betting range becomes exploitable when you bet 100% of it.
Once regs start exploiting you, many hands become higher EV as a check back.