reply

what I am saying is the distribution is like a normal distribution.

All these systems have many problems.

They don't take into account a lot of factors that effect future matches.

Almost none take into account cup matches, European Matches, Days rest between cup matches and LG games.

We KNOW for a fact these things are important.

Also the result correlations are important.

If teams need a draw to stay up then a draw is more likely than if they needed a win to stay up.

A number of teams will be "on the beach" plus a number of players wont risk anything in a meaningless match if they are Euros bound.
Some teams who are safe will be tempted to throw in some younger players.

All systems are flawed to some extent.

I know how the professional groups simulate games and it is via a Monte Carlo type simulation
(run the game a billion times to get the percentage of each result)
with a ton of fudge factors to account for a number of the things I have mentioned above.
That PLUS inside info. (ie line ups and injuries) to get you ahead of the market.

What I was trying to suggest in a snarky way was that you could get a close estimate by just taking PPG data and extrapolating, using that as the median point.
It should be obvious this is the most likely out come if no other info was available.
Over a large sample of games, say 20 games, the relative strengths of opposition evens out somewhat.
The fewer games remaining, the more valuable the information to adjust becomes.

When you are down to the very last game you are playing with virtually perfect information.

Posted By: usacanary, Mar 23, 15:01:31