TC, stake and payout: let's get some facts straight

Exactly!

Also to make an even more extreme example on the current payout method, let’s say an extremely good model stake is 50% of the total tournament stake. Since this is an extremely good model the portfolio optimizer keeps its contribution ~50% at every round, thus TC will tend to be 0. This incredibly great model will be paid ~0 NMR every round although its actual contribution is 50%! Sometimes it will even get negative TC, due to the little stake adjustments required to keep all the stake weighted portfolio optimal.

See here.

Being paid on TC doesn’t really make sense. TC is not a measure of the true contribution of a model, is a measure of how noisy the tournament is: more noise → more TC is required to keep the stake weighted portfolio optimal (with noise I mean market volatility, model changes, user stake changes)

1 Like