Thread by @nntaleb

If employees of UBS knew how to run simulations, they wouldn't be employees of UBS, but customers of the Private Banking group of UBS. How can't people think in 1 ½ order effects?

https://twitter.com/business/status/997061733566107649

Giving Germany a 25% chance of winning is a very arrogant statement, it is equivalent to giving Germany ~100% chance of making it to the semi-final.

Don't listen to bank research pple with no #SkinintheGame who talk about probability.

Very similar to elections. Under high variance, you go closer to 1/n probability per team, where n is the number of teams, with minor variations (except for the very bad teams such as Saudi Barbaria).
The more uncertainty, the more distributed the bets.
arxiv.org/pdf/1703.06351…

There is a set of probabilities that can give Germany 25% of winning without 100% probability of being in the semi-finals: that CONDITIONAL on making the semi-finals probability increases.

However, the opposite holds: the higher up, the MORE 1/n.

Gabish?

The other counterintuitive aspect is that EVEN if Country X won EVERY past game, injecting a tiny bit of randomness into the system can make its probability of winning drop below, say, 10%. A tiny bit of randomness percolates quickly across the branches of the tree.

Soccer is fundamentally a high variance game.

https://twitter.com/Mehdi_IR_DC/status/1013792319550705665