I'm a little confused on these concepts. So for example, on the river if villain call's his bluff catchers 5% more of the time than equilibrium, If we adjusted to a maximally exploitative strategy we would never bluff because our bluffs are no longer indifferent. But if we instead used a minimally exploitative strategy we would only bluff 5% less to just match his imbalance and avoid getting counter exploited. So when you node lock a specific imbalance into a solver, it's using a minimally exploitative strategy, right? because it's balancing our strategy around villain's imbalance, not exploiting it necessarily? So essentially where still playing GTO, just with a higher overall EV? if that makes sense

The solver is always trying to maximize value against whatever mistakes you force one player to make. MinES and MaxES are both exploitable.

Modern Poker Theory defines MaxES/MinES as follows:

Minimally exploitative - We assume villain will play suboptimally at one specific node, and perfectly everywhere else. The MinES is somewhat vulnerable to counter-exploitation and harder to detect.

Maximally exploitative - We assume villain will play suboptimally everywhere. The MaxES is extremely vulnerable to counter-exploitation and much easier to detect. We need to know villain's entire strategy and nodelock every point in the gametree with their imbalances. For this reason, it's not usually practical to solve this.

There's also a distinction between reactive and proactive exploitation:


Yes, because you only locked one node.

The solver will just try to maximize value against whatever mistakes the player is making. It assumes the locked player can't change that point in their strategy (so it doesn't consider counter-exploitation).

MinES is still exploitable, and it's a lot more complicated than simply bluffing 5% less often. You rebuild the entire strategy around that leak

The solver IS exploiting villains imbalance for as much value as it can get. The only difference is it can't get too out of line because it knows villain will play perfectly outside of that point in the game tree.

Makes sense, thanks man. Guess I gotta re read Modern Poker Theory, it's been a while.

I love your username btw lol

I read that part of MPT and maybe this is me just beeing dum, but idea of Minimally exploitative strategy seems unnecessary.
In book author calculates MinE strategy for river polar vs bluff catchers game by changing payouts for one player(if i remember correctly?) who is calling too much. This makes no sense from theory protective at least, if you node-lock this into slover he just gives up all the bluffs.
There is nothing minimal about strategy that you get if you node-lock this is again Maximally exploitative given the assumption.

The idea is probably that you can't 100% know what your opponent's ranges look like (unlike solvers). If you decided to maximally exploit a player that you've seen overbluff once before, he could in turn then exploit your "overexploitation" and make profit vs you. He could intentionally overbluff in the 1st hand, just so that you'd overadjust in the 2nd hand and lose even more EV than he lost in H1 by overbluffing. That's probably the idea.

Another reason for minimally exploitative strategies might be to "stealthy" exploit the opponent, so that he doesn't fix his leaks. For example, if you keep 3betting 100% of the time vs someone that overfolds to 3bets, he's soon going to realize what you're doing, so it might be more profitable to only slightly adjust. There is some logic in making short term -EV plays if they increase the EV of your future hands.

Quick question if we find leaks in the population as a whole can we profit by blindly assuming all unknowns have that leak and playing a maximumally exploitive strategy for example say the pop overfolds from BB versus a SB RFI do we profit by betting SB 100% of the time against 100% of players . Or us it possible that we lose enough money to good players that we lose overall?

I agree with every point you made. Problem i have see is that notion of maximally exploitative strategy is very well define but minimum ES is not.

You can say minES is strategy that maximizes EV over infinite( or finite?) number of hands vs reactive opponent. My point is this is not strategy that you get from solver when you node lock and it's not strategy that author of MPD found in his book.


If you do it vs every player then its fine. If you do it only vs unknowns then there is possibility that average unknown doesn't have that leak and you end up losing ev.

Great question!

It really depends how counter-exploitable your adjustment is. Some exploits are more vulnerable than others. If the exploit is too fragile then even a small minority of players countering you can destroy any profit gained from the adjustment.

For example, if an exploit normally gains 1bb against most of the population, but can lose 5bb against someone countering us, then you need to be right about that exploit 80% of the time to break even with it. (In reality, it's not gonna be +1/-5 binary but you get the idea).






In the above example, the MES against a nit is to push any two and print 29.4bb. But if that nit starts countering us by calling down wide, now we're losing 62.4bb. That's an example of an extremely vulnerable exploit.

Conversely, against the player that calls too wide we're supposed to tighten up a bit and play more value-heavy. This adjustment gains an extra 6.3bb. If that station starts to nit up, the most they can counter-exploit us for is 2.5bb. This is an example of a less vulnerable exploit.

The MPT example isn't very good. A better example of MinES would be nodelocking a flop, knowing villain cbets too wide. The solver will exploit this for all the value it can get. However, it can't get too out of line because it knows villain will react optimally going forward.

We needed a way to distinguish attacking one node from their entire strategy. This seems silly right now because we don't have the tools to effectively study multistreet nodelocks. Imagine a solver that allowed you to slightly overfold all the hands that were mixed folds in one player's range, on every node. Then recalculate a MaxES against that. Now we're exploiting a tendency across the entire game tree. Something like that ought to be categorized separately from exploiting a single node.

Sure but that is still maximally exploitative strategy given the assumption we have.
Interesting point about pool expolits. If we update our pool analysis regularly we should see if pool is adjusting to our strategy and make changes.

As far as I know minimally exploitative play was first used as a term in poker by Alex Sutherland, who is the dev of GTO Range Builder (one of the first solvers to be available to the public). Here's a video where he talks about it

https://www.youtube.com/watch?v=xfMmaWO42o8

The definitions given in MPT are correct, but applying the term minimally exploitative to river play doesn't make much sense as there is no future action left. On the river if you exploit the adjustments are very dramatic and usually all or nothing. However that is contingent on your read being correct. In reality we are working with degrees of confidence. At any rate, minimally exploitative play involves nodelocking a leak and assuming all future action is played at equilibrium (whatever that new equilibrium is).

Another term that I think is more useful for river play is the concept of exploiting at the margins. Meaning you start at the optimal baseline and marginally add or remove combos from your range based on how confident you are in your read. If I'm really confident the guy is going to fold I'll bluff everything except the worst blockers. If I'm not that confident I'll only bluff the best blockers. Notice that regardless of my read, I still always have a bluffing range and a give up range.

Maximally exploitative play is just a theoretical concept that can never be achieved in practice in a game as complex as poker. To take your example of people overfolding and you betting 100%. What makes you think that is the maximal exploit? If people are overfolding do you really want to bet all your invulnerable made hands? What the maximum exploit should look like at that node is not that easy to say, because you would need to know how your opponent is going to play in all future nodes as well.

Than you for the video.
So nod locking strategies are called minES. They don't anything to do with how easy they are to detect or contra exploit( but they usually are more difficult to detect and expoit)
MaxES you need to know opponent whole strategy and exploit.
It's just matter of definition. I would call nod lock strategy maxES because it is maxES vs strategy solver would play vs nod lock.

It all comes down to if is villain going to counter exploit ot not.

If not, which is usually going to be the case, just don't make -EV plays, like bluffing in this instance, because those plays lose money period.
On the other hand, if he sees you check a certain bluff combo on the river and he knows that combo to be a slam dunk bluff, and he subsequently adjusts by never calling, you should still bluff with a few slam dunk combos.
This second playertype is maybe 5% of all players and most of them play very high stakes, so in your case, just don't make -EV plays.

I might be late to the party but just stumbled upon the same issue with the MPT example. I've used GTO+ to set that toy game on the river and after locking bluff-catcher strategy to call 55% of the time, solver just removed all the bluffs. No miES, just maxES.

Does anyone have an actual sample of a toy game that can be used to explore differences between minES and maxES strategies using solver?

A nodelock is just a node in villain's strategy that can't adjust. "Hero" will take only the best EV line against that node, and obviously if bluffs are -EV then Hero will check.
MinES just means that the rest of the nodes in the gametree are played with full freedom to adjust.
If you're on the very last node of the tree than all exploits are maxES for obvious reasons.