How does your Prediction Program compute the "Reliabilities" it operates with?
We have to load "starting data" into the Program to calculate the initial "Reliabilities", and for that we use the actual results and score-lines for the "Staying" teams, while for the Relegated and Promoted teams we use the data from the artificially created "Starting Matrix" (where artificial match results are derived using the true Win Quotients of the Staying teams and the artificial WQs of the Relegated and Promoted teams, plus the overall average Scoring Abilities and Vulnerabilities of the Relegated and Promoted teams, derived by "cloning" them against the records of the Staying team considered closest in terms of "performance capability"). Yes, that explanation hurt, we know!
After that it's easy. All our Program does is replace old match data and artificial match data with the results from actual played matches, thereby enabling the updated Reliabilities to be calculated. Needless to say, as the season progresses the "Reliabilities" data becomes, well........more reliable!