2015 - Round 4 Results: Hot ChiPS

It's an odd weekend's football when the three predicting progeny of the ChiPS System wind up as the best Head-to-Head, Margin and Probability predictors.

The C_Marg-derived Head-to-Head Tipster tipped seven of the nine winners, two more than the TAB, which was enough to elevate it into joint leadership on the MatterOfStats Leaderboard, a position it now shares with five other Tipsters.

Amongst the Margin Predictors, C_Marg produced the only sub-30 Mean Absolute Error (MAE) for the round, kicking it almost 40 points clear of Combo_7 at the top of its group. One of MoS' newest Margin Predictors, ENS_Linear, now sits in third, less than a point behind Combo_7.

And, to complete the ChiPS story of triumph, in the Head-to-Head Probability contest C_Prob returned the round's only positive probability score, leaving it comfortably leading its peers.

While ChiPS might have been collecting, Investors weren't, the Line Fund's losses from its 0 and 2 performance more than offsetting the Head-to-Head Fund's profits from its 2 from 3 performance. All up the Overall Portfolio shed just under a cent this week to finish the round down by just over 7.5c on the season.


ChiPS aside, a considered look at the week's tipping and predicting results would have it that it was a relatively difficult round.

The Head-to-Head all-Tipster average was just 4.3 from 9 games, dragged below the BKB score of 5 largely by a succession of 2s and 3s from most of the Heuristic Tipsters. C_Marg's 7 was, as noted earlier, the round's best score, no other Tipster scoring more than 5. Across the season, the leading score amongst the Head-to-Head Tipsters is currently 24 from 36 (67%).

Margin Predictor performances were quite varied this week, Predictor MAEs spanning about a 10-point range from C_Marg's low of 27.5 to Combo_NN1's 37.5 point high. The all-Predictor average was 34.5 points per game. C_Marg now has the season's best MAE at 31.0 points per game and the all-Predictor season average stands at 32.8 points per game.

The bar chart in the middle of the Margin Predictor section of the Tipster Leaderboard shows how infrequently Predictors have been able to predict margins very close to the actual result this season. C_Marg, for example, has tipped within 6 points of the final margin in only about 6% of games, and Combo_7 has done it only a little more often, producing this result in about 11% of games. What's differentiated Predictor MAEs this season has been the avoidance of huge errors moreso than the accumulation of a series of smaller ones.

Head-to-Head Probability Prediction was also relatively difficult this week. As noted earlier, C_Prob was the only Predictor returning a positive probability score, the best of the rest being the two H2H Predictors' mildly negative scores. ProPred recorded the round's worst result and now sits in 5th on the Leaderboard behind C_Prob and the three Bookie Predictors.

The Line Fund algorithm also struggled with its probability assessments this week, finishing with a slightly negative probability score for the week, the fourth time in a row it's recorded such a result. This week's result was better than those of Round 2 and of Round 3, though that's about as quintessentially a small comfort as there is.