Closed JackFawthorpe closed 4 months ago
This is done. Whilst doing this, a bug was identified when the ranking difference is very large. What would happen was the huge discrepency in rating would mean the win likelihood was 1.0 when 1 is an asymptote of the function. This is the nature of computers and probability.
Although this was fixed and the clamping is done there needs to be further work done to limit the breaking of the system
There has to be a bug somewhere, everytime I come back after a long period of time the winner has an insane rank
Clamp volatility and std deviation as a preventative measure till we figure out why it happens