The application of ratings systems based of predicted % likelihoods

[Clotty]

I was wondering if someone could help me out. I have two main questions, involving determining a probability from ratings. I'll start off with the easier question. Let's assume I ussed an ELO type ratings system and had one team with 1400 playing at home and another team playing awat with 1600 points (with the starting points total being 1000). How, from this data, could I work out the implied % chance of a home win, away win and draw? Second question, is this possible to do from one value? I'll use an extremely simplified version of my over/under/ predictions ratings. Let's say I add the number of goals scored by a team with the number they conceed and divide it by the the number of games played. I do this for the away team they are playing, too. Assume the home team have, on average, 2.37 goals in a game, while the away team have 3.46. I then average these two figures to produce a very, very rough estimate that, on average, games between these two sides will produce 2.92 goals (to 2 d.p). Is there a way for me to use this figure of 2.92 to predice the probability of over/under 2.5 goals? Thanks for any feedback.

[cavello]

Re: The application of ratings systems based of predicted % likelihoods Over/under is strongly related to the relative strengths of the teams. Note that matches with big odds on shots often produce plenty of goals. The Racing and Football Outlook use their ratings(similiar to ELO) to produce computer true odds for matches. These have IMO been produced by using historiical data concerning these ratings to generate a regression line and formula.