FAO: USAC

Yesterday's post about that league table predictor Mikey found, just out of interest what makes you think it's a normal distribution?

User Posted Link

I may be missing something here, but isn't the creator of this thing simulating remaining games so that three possible outcomes for each game is L (0 pts) D (1 pt) W (3 pts)?

I'm assuming under for the weighted method they've applied different probabilities to these different outcomes (based on who the home team is and what their record has been to date) somehow, which would give each of those three point scenarios (0,1,3) a different probability of occurring each simulation to 33.3% each.

So I'm not sure how a normal distribution with 1.333 as the mean (expected value of pts per match) would be derived from the above.

I get that by using Monte Carlo simulation techniques it's likely they used a normal distribution to (pseudo) simulate from the "weighted" / skewed distributions so as to build up an aggregate distribution over a large number of simulations and taking into account correlation with opponents (so that if one team W the other had to L or both D etc)

But what made you say they've simply applied a normal distribution to their assumptions? Intrigued more than anything.

Mikey?

Posted By: essexcanaryOTBC, Mar 23, 13:26:26