One problem when nodelocking in solvers is that the solver will try to re-adjust in unlocked nodes to minimize the EV loss caused by the imbalance introduced in the locked node.

Quite often this is undesirable because it leads to unrealistic behavior on earlier/later nodes.

It's usually easy enough to lock the earlier nodes to original GTO, because it's usually only one or a few nodes that need to be locked.
I also see that Pio has an option to "lock all parent nodes".

But future nodes are much more difficult to nodelock because there are exponentially more future nodes as the game tree branches out.

I have heard of some guys that have custom scripts/plugins to do nodelocking for future nodes, but I don't think it's publicly available?

****************************

What about this as an easy solution: just have an option to lock all combos in all future nodes to the original GTO frequencies?

I think this will automatically lead to more realistic nodelocks in a lot of cases.

What we see in reality is that when humans arrive at the river with too much air, they will often overfold/overbluff river.
When they arrive at the river with not enough air, they will often underfold/underbluff river.

****************************

Simplified example: We play the nuts-or-air toy game where IP can bet 1x pot or check.

IP start with 10 value combos on the turn and 25 air combos.
IP bets each air combo 50% of the time on the turn = 12.5 air combos.
IP then bets 5 air combos on the river and gives up 7.5 combos.
So IP is betting each air combo 5/12.5 = 40% of the time on the river.

So let's use our tool to now easily nodelock all rivers so that air combos bet 40% of the time, just like in the original GTO solution.
Now we also nodelock turn so that villain bets each air combo 60% of the time on the turn instead of 50%, giving us 15 air combos instead of 12.5 air combos on the river.

Now he will automatically bet 6 air combos on the river (15 * 40% = 6 combos).
So bluff% automatically goes up to 6/16 = 37.5% without having to do any additional manual nodelocking.

****************************

The problem: this would only work for combos that are at least mixed continues in this line on earlier nodes in the original gto solution, right?
Since we couldn't lock any original GTO combo frequencies on future nodes for combos that are not even present in the future nodes in the
original GTO solution.

Possible solution: Couldn't we use something like QRE to mix in all combos in all lines at very tiny frequencies? Just enough to allow us to use this technique of locking all combos in all future nodes to the original GTO combo frequencies but without significantly altering equities/EVs/strategies?

Seems reasonable to me. Have you talked to Kuba about it or mention it in the PIO discord?

Dunno. If you locked all future nodes, to GTO or any way, mixed or pure, then you wouldn't get the exploit?

i.e. doesn't the solver exploit a node precisely by balancing and reorganising later EV? Like it's literally pulling EV from those later combos to attack current imbalances. Locking future possibilities just cuts off the fuel.


I thought about this dream solver before but my sense was that the reason we don't have 'future node locked to fish subtype' solutions yet is because 99% of the time most regs and nearly all fish wouldn't even know what they'd do with certain combos/runouts until it happens 'on the day' with all the table dynamics involved with that, so it's effectively way too abstract to make any of the necessary assumptions required (and that's before compute size etc)

Yes, you would get the exploit. That was the point of my example. I'm obviously talking about locking my opponent while keeping myself unlocked.

Why not just check the hand EVs after nodelocking? That should give you the results you’re looking for.

What ZK said.

Just nodelock it, then check EV **before** you re-solve it. It will calculate the EV against the current equilibrium strategy.

Ah, thanks guys. That's clever!