This year, MAFL's/MoS' 10th, was a record year in a number of ways, with the most traffic to the site, the most comments (though still far fewer than I'd like), the best tipping and prediction relative to the bookmakers, but also the largest wagering loss.
For 2016, I want to build on the improvements and remove some of the stuff that I feel might have had its day. This blog will serve as a running to do list of things I'm considering and I'd encourage you to leave comments if you've any thoughts or suggestions for what you'd like to see or like to see less of.
So, here's the current list:
- Rebuild the Head-to-Head Fund algorithm; possibly allow it to wager on Home and on Away teams
- Rebuild the Line Fund algorithm; possibly allow it to wager on Home and on Away teams
- Consider an Over/Under Fund and algorithm
- Retire the Heuristic Tipsters (apart, maybe, from Consult The Ladder, Bookie Knows Best, and Home Sweet Home)
- Retire the H2H, ProPred and WinPred Tipsters, possibly replacing the H2H Tipsters with something derived from the rebuilt Head-to-Head Fund algorithm
- Retire the H2H, ProPred and WinPred Predictors, possibly replacing the H2H Predictors with something derived from the rebuilt Head-to-Head Fund algorithm.
- Retire Combo_NN1
- Retire the H2H, ProPred and WinPred Head-to-Head Probability Predictors, possibly replacing the H2H Predictors with something derived from the rebuilt Head-to-Head Fund algorithm.
- Consider an ensemble-based Head-to-Head Probability Predictor
- Rebuild the Line Fund algorithm as noted above
- Stop calculating Massey, Colley and ODM Team Ratings/Rankings
- Track ChiPS, MARS and MoSSBOD each week
- Refine the Home Ground Advantage (HGA) estimates used by ChiPS and MoSSBOD to better reflect the most-recent performances at the Venue. Resolve (probably) that Venue, not designated Home Ground status, is what drives HGA, so calculate each team's HGA at the Venue for every game.
I also have some unformed thoughts about incorporating player data - maybe somethings as simple as aggregate player experience - but I'm not sure exactly how yet (or how I'l source the data).
I'll continue to update this list as new ideas arise or existing ones get done, get altered, or get ignored.