How much predictability there is in aggregate features in leagues? By 'aggregate features', we're talking about promotion, relegation, etc., as opposed to individual games which are not very predictable.
This model uses Markov Chain Monte Carlo together with a Poisson model of goal-scoring and a prior distribution from the previous season to predict future matches. The rest of the season is 'rolled out' many thousands of times, and the proportion of times a team gets promoted/relegated/in-the-playoffs/... is displayed in a the tables linked to here. The predictions early in the season are not going to be that accurate, though the later ones seem to be quite reasonable. (They don't take into account real-world knowledge such as what has happened in the transfer market, who is injured, etc..)
It should automatically update itself every evening to take account of newly played matches.
(Any bets are at your own risk...)