And so we wind up with the team Rated 1st by both ChiPS and MARS playing in the Grand Final against the team Rated 2nd by them both. After a week of seeing various statistics quoted about the success of teams of yore whose only similarity to the Dockers was that they also finished as Minor Premiers and/or were playing at home in previous years' Preliminary Finals, my faith in my understanding of the types of patterns than matter and those that don't in analytics is restored.
For the most part, historical statistics are of limited, if any, predictive value - though they are interesting, don't get me wrong. What matters almost exclusively is the revealed form of the teams that are actually playing in the contest in the current year, not that of other teams from other years that have some tenuous similarity. At some point I plan to do a whole blog post on the topic of what I'll call "predictive relevance", but for now let's focus on the Preliminary Finals just played.
Not that I'm all that keen to return the focus here, because it was another sorry tale for Investors this week, the last-minute lazy miss by the Eagles preventing even a consolation collect in that game. So it was 3 losing wagers from 3 for the round, lopping 2.2c off the Head-to-Head Fund, 1.3c off the Line Fund, and therefore 1.6c off the Overall Portfolio, which now stands at just 63.2c.
TIPS AND PREDICTIONS
Far better was the news for the Head-to-Head Tipsters, a majority of which tipped both results, leaving C_Marg still four tips clear of the pack and now with a 72% accuracy level for the season. Two other Tipsters, Bookie_3 and Combo_7, are also tipping at above 70% accuracy.
Not only did C_Marg manage to select the right teams this week but it also performed amazingly well in predicting the final margins, its Mean Absolute Error (MAE) a remarkable 6 points per game. That has comfortably restored its season-long MAE to below 30 and has left it with a 139-point lead over the 2nd-placed ENS_Greedy. That is, surely, an unassailable lead.
More broadly, again this week, the ordering of the Margin Predictors remained unchanged.
C_Marg, Bookie_3 and Combo_NN2 still have season-long profitable line-betting records, and are joined by H2H_Adjusted_7 in having that status.
Amongst the Head-to-Head Probability Predictors, C_Prob returned the highest probability score for the week, although all eight Predictors recorded strongly positive scores. Barring a daft and errant probability assessment by C_Prob in the final game of the season it looks set to finish 1st, which will mean that a member of the ChiPS Forecaster family will take top spot in all of the three competitions.
The Line Fund algorithm had a very poor week, its probability score for the round the 2nd-lowest of the season, better only than the result for Round 2.