All games with a prediction 200 games. 132.5 wins (66.3%) Level stakes profit = +0.04%
To all intents and purposes, the predictor breaks even betting on the predicted winner regardless of odds available. It picks the winner in almost exactly two-thirds of games. A random selector would pick 50% of winners and lose approximately 3.8%, so this is quite good overall although nothing to get hugely excited about.
Recommended bet games 33 selections. 18 winners (53.8%) Level stakes profit = +£8.35 (+25.3%)
This is a return I'm pretty happy with - definitely more than I'd imagined, and also evidence that using Kelly Criteria to determine stakes is indeed more efficient/effective! The lower proportion of winners is expected, given the average odds of the selections, and here the percentage return is the crucial factor.
Given the weekly update (and cranking out of all predictions/selections/stakes) only takes about 40 minutes, it seems worth the time invested, and I'll be having another go from Week 5 onwards next season. Fairly sure I might be hard pushed to match the nap strike rate from this year (an unlikely 9 from 13), but already looking forward to the challenge!
So all calculations/bets based on bet365 prices available as of Thursday each week. A general comparison I did early on (in the first weeks of the season) indicated that there was very little difference between those odds and the betfair exchange price minus commission, with their markets generally at 103-104%, but with fewer 'micro' changes along the way.
And thanks both - I will look at whether there are any little tweaks or additional factors that might be worth considering or monitoring, but also conscious that it would be easy to overly engineer a backfit model based on the (small, single season) data that could be misleading. Also, I don't want it to become subjective or start taking much longer to crunch through each week.
So all calculations/bets based on bet365 prices available as of Thursday each week. A general comparison I did early on (in the first weeks of the season) indicated that there was very little difference between those odds and the betfair exchange pri