MoS for the Men's 2025 AFL Season

/

For those purely interested in MoS forecasts, 2025 will look a lot like 2024, although there will have been non-trivial changes under the hood for the underlying algorithms that provide those forecasts. For those also - or instead - interested in the wagering, the differences will be a bit more substantial.

There will, once again, be weekly posts predicting results for the round ahead, reviewing the results of the round just gone, updating Team Ratings and the Team Dashboard, and projecting the remainder of the home and away season and the Finals. There might even be a few posts looking at specific issues (clients willing).

A summary of the main changes follows.

ALGORITHMS

The algorithms for MoSSBODS and MoSHBODS have been substantially rebuilt (as detailed in this blog) with the main visible result being that Team Ratings are now measured in standard deviations - around recent average team point-scoring for MoSHBODS and around recent average team scoring shot production for MoSSBODS. Less visibly but more importantly, they also perform slightly better on historical data, but not shoot-the-lights-out well, which is always a hint at overfitting.

MoSHPlay has also been reoptimised to include the 2024 season and to take into account the revamped MoSHBODS algorithm. As well, a bug has been removed from the Player Rating component, which last year saw the algorithm include the entirety of a player’s AFL Player Rating data in coming up with a one-game ahead forecast when it should, instead, have only been including history going back no more than some fixed number of days. The impact on the forecasts was probably only small, but would have varied by player, depending on career length.

The Player Rating component no longer decays one-game ahead forecasts for players who have missed one or more games, or returned after a long period out of the game, and limits all forecasts to be in the range 7 to 15. Only the last calendar year’s worth of actual AFL Player Ratings are included in its calculations.

For use in MoSHPlay, a team average of its one-game ahead forecasts is calculated (excluding the substitute), which is combined with the relevant MoSHBODS’ margin forecast for the game passed through a fitted MARS model with hinges (more on which maybe on another day).

FUNDS

The same three Funds will operate in 2025 with the same weightings of Line Fund 65% / Head-to-Head Fund 25% / Over Under Fund 10%.

All three will still fractional Kelly stake at 20% of the Kelly recommended bet, but applied to the original, not the current, bankroll. A review of last year’s results, applying the bets to the current instead of the original bankroll, produced results very similar to what was actually achieved.

That same review also suggested that imposing a fixed 5% minimum on the edge required to wager both put some wagers at risk ill-advisedly, but also left some money on the floor. In particular, the review reinforced a view that I’ve long held and only occasionally abandoned to my cost: betting on home teams is different from betting on away teams.

So, this season, the Funds will face the following constraints:

Head-to-Head Betting

  • On home teams: no minimum edge and 25% maximum edge

  • On away teams: 10% minimum edge and 50% maximum edge

Line Betting

  • On home teams: no minimum edge and 50% maximum edge

  • On away teams: 20% minimum edge and 50% maximum edge

Totals Betting

  • On under and overs: 10% minimum edge and 50% maximum edge

The rationale for imposing minimum edges is fairly self-explanatory: a model doesn’t have to be wrong by much to drag a should be “no bet” situation into a “let’s make a small bet” situation. The rationale for imposing maximum edges is that, at some point, it’s more likely that your model is incorrect and missing some key piece of information than it is that a bookmaker has made such an enormous pricing error.

The final thing to note is that MoSSBODS has been optimised for Totals predictions and so will solely be used for wagers in the Totals market, leaving MoSHBODS to handle the Line and Head-to-Head markets.

The overall effect of these changes will likely see Investors wagering on fewer games, but with higher average wagers on away teams (because of the higher floors) and lower average wagers on home teams (because of the lower floors).

Profitability however, remains resolutely non-guaranteed.