Oscillating Around Equilibrium

This is not radically different than just changing gears, but the idea is still a bit more specific: including more or less bluffs than optimal from a GTO perspective for a time, and then reversing that when you start to feel that opponents in a hand have begun to adapt to your non-GTO play. Against stronger opposition, this in turn eventually reverses itself again, and then again, with the oscillations gradually diminishing in size.

I’ll typically implement this by significantly under-bluffing in a new environment for a few hundred or a few thousand hands, and then reversing that and bluffing at unusually high rates once I observe that people are over folding.

If it is an environment where people generally already over fold, begin instead by over bluffing, and then flip that when you start to see a significant uptick in call rates. Ideally these oscillations around equilibrium end up converging back at equilibrium, though in practice this is probably a bit hard to achieve.

Again, this is just one form of changing gears… let me know what you think.

In my opinion they are not overfolding but overcalling but its only my opinion but that what you wrote sounds like good reverse strategy

1 Like

Sure, I think that is more typical on a play money site; but that will change depending on the tables you play at and the particular players. For live cash games, defending too narrow a range is more common than defending too wide; but again, you’ll find yourself at tables where that is probably not the case.

1 Like

Suppose GTO parameter for raising flop cbet stat is 12.5% and with your underbluff it becomes 6.3%. Agaist reasonably near-perfect opponent, it’s -EV due to missed oppotunity cost. Also you also need to “change gears” before opponent has adjusted to them, that’s also implies -EV risk. Suppose, over 100s of hands you have raising flop cbet stat ~ 6.3%, but now you changed gear to raising cbet 25%. If I observe your raise cbet stat as ~ 6.3% but now I see that you raised last 4 flops of 16, thus making hypothesis that your raise cbet stat is 6.3% is wrong at P-value 0.998. Even if you will switching to raising cbet to 12-15%, you still will not last much against rational agent who does proper bayesian updating and 1 iteration will be enough).

I wouldn’t worry about “reasonably near perfect opponents”. If you are playing a human, they are not that.

But still, it is true that anytime you deviate from equilibrium, it is true that you are creating the possibility of being counter exploited. I’d group opponents into three main groups here:

  1. Opponents with wildly non-GTO play that also never seem to modify their play: don’t bother to oscillate here, but just exploit their play (I tend to avoid going for a maximally exploitative strategy even here, as it just seems like they’d just have to adjust at that point, LOL)
  2. Opponents with wildly non-GTO play that seem to adapt to your play: these are the targets for this kind of adaptive strategy
  3. Opponents that seem moderately GTO: strive for balance yourself