[SumZero]

Simulation.

The best way to show him you are right is to simulate it. Create a population of 1,000,000 players with various bankrolls. Make only 10% of them winning players, but have the typical random outcomes when people play. Give people bankrolls that once they lose they stop. Simulate the population for a while. Subsample the population of players (looking only at a small % of the hands played) to simulate the types of looks your tracking player gets. Show that if you only look at the subsamples you think a lot of players are profitable that if you look at the full data and/or the actual a priori settings you realize aren't.

That may be more work than you are willing to put into it.