How good is a solve?

The longer version

What each scene is saying, spelled out. Tap a note to jump to its scene.

1. 0:00How good is a solve?
2. 0:03No solve is ever perfect. Every pass brings the strategy closer to equilibrium, and none of them arrives. So the real question is: how close?
3. 0:08Imagine a perfect opponent who knows your whole strategy and plays the best possible reply. Whatever share of the pot that opponent can take from you, averaged over every hand, is your exploitability.
4. 0:13How low is low enough? As rules of thumb: around 5% of the pot is rough and still has big errors; around 1% is usable and the shapes are right; around 0.3% is tight, with the fine detail settled.
5. 0:17Each step costs more. The error falls fast at first and then flattens, so the last improvements take most of the time.
6. 0:20That is why you pick a target and stop there. Running longer buys very little.
7. 0:24Exploitability is the honest number. Check it before you trust a solve.