2015 - Team Ratings After Round 10

/

Just five teams changed their ChiPS Ranking this week while 13 changed their MARS Ranking, though nine of those MARS changes were only by a single position, so the contrast isn't as stark as it first appears.

From the viewpoint of the competition the most significant of ChiPS' changes was the elevation of Sydney into 3rd place at the expense of Fremantle, whose loss to the Tigers cost them 3.4 Rating Points (RPs) and leaves them now trailing the Swans by a little over half an RP.

More dramatic, however, was the Dons' slide of four places back into 11th after surrendering 4.2 RPs to Geelong, the transfer of those RPs also precipitating the other major move on ChiPS, which saw Geelong rise three places into 8th. The only other team to move on ChiPS this week was the Roos, who climbed one spot into 7th.

The drama in MARS Rankings was the three-spot climb by Carlton into 14th, the two-spot climb by Port Adelaide into 5th, the two-spot decline by Adelaide into 7th, and the two-spot decline by St Kilda into 17th.

Those four MARS multi-spot Ranking changes, along with the eight one-spot movements have left just four teams Ranked more than two-spots differently by ChiPS and MARS:

  • Richmond (Ranked 5th on ChiPS and 9th on MARS)
  • The Kangaroos (7th and 10th)
  • Port Adelaide (9th and 5th)
  • Adelaide (10th and 7th)

The rank correlation between ChiPS' and MARS' team Rankings now stands then at +0.90 and the correlation between their Ratings at +0.97. Clearly, ChiPS and MARS still agree a lot more often than they disagree.

MASSEY, COLLEY AND ODM

Time then to introduce a few second opinions via MoS' other Team Rating Systems. The most divergent of those new opinions belong, as usual, to Colley, though its Rankings generally moved closer to ChiPS' this week, as evidenced by the increase in the rank correlation between Colley and ChiPS Rankings to +0.80. Colley's rank correlation is still higher with every other System, however, and highest with ODM at +0.93.

Colley only finds itself significantly alone now in its Ranking of Hawthorn, which it Ranks 7th and the other Systems Rank either 1st or 2nd. ChiPS, meantime, is similarly alone in its Ranking of the Roos, which it has in 7th and no other System has higher than 10th.

Together, ChiPS and MARS are also divergent from the other Systems in their assessments of GWS, Ranked 12th by them and no lower than 8th by the other Systems, and of the Western Bulldogs, Ranked 13th by them and no lower than 8th by the other Systems.

ODM OFFENSIVE AND DEFENSIVE RANKINGS

Fremantle, Sydney, West Coast and Hawthorn, in that order, remain the Top 4 Ranked Defensive teams according to ODM this week, with Essendon dropping out of the Top 5 and making way for Richmond.

On Offence, the Top 4 remain Hawthorn, West Coast, Fremantle and the Western Bulldogs, also in that order, with the Roos supplanting GWS in 5th.

The largest differences between team ladder positions and  Defensive and/or Offensive Rankings are the following:

  • Sydney (2nd on the ladder, 2nd on defence, 8th on offence)
  • Collingwood (4th/8th/10th)
  • GWS (5th/9th/7th)
  • Hawthorn (6th/4th/1st)
  • Richmond (8th/5th/12th)
  • Port Adelaide (9th/6th/11th)
  • Geelong (10th/12th/6th)
  • Western Bulldogs (11th/10th/4th)
  • Essendon (13th/7th/16th)
  • Kangaroos (12th/13th/5th)

PREDICTIVE ACCURACY

This week, as well as reviewing the simple predictive accuracy of the MoS Team Rating Systems for picking winners, I thought I'd also assess the simple margin predicting abilities of the ChiPS and MARS Systems for estimating final game margins.

The scatter plots below chart the actual margins in all 90 games this season against the difference in team Ratings that prevailed directly before each game, and reveal that:

  • Pre-game differences in ChiPS' and in MARS' Ratings explain about 30% of the variability in game margins this season, closer to 33% for ChiPS, and a tick under 30% for MARS.
  • The difference in team pre-game ChiPS Ratings is (as it was designed to be) a good indicator of the final margin after we adjust for home ground advantage (HGA, which about 5 points if we use ChiPS Ratings). We can say this because the coefficient on the Ratings Difference variable in the equation shown on the chart at left is 0.96, which is close to the ideal value of 1 that would apply if the exact difference in the team Ratings was the best estimator of the final margin after adjusting for HGA. This 0.96 figure suggests that the raw Ratings Difference tends to slightly overstate the final margin.
  • The difference in team pre-game MARS Ratings needs to be reduced by a larger amount in order to derive the best possible margin prediction after adjusting for HGA (which is estimated at about 5.7 points if we use MARS Ratings). Roughly speaking, we need to reduce the raw difference in Ratings by about 9% in order to provide the best margin estimate. That's a lot lower than the reduction of about 25% that I've estimated as being required in previous seasons.

Despite ChiPS' marginal superiority in margin prediction across the season so far, MARS outperformed ChiPS and every other System this week in predicting winners. It picked one more winner than ChiPS, Colley, ODM and both ODM component Ratings, and picked two more winners than Massey.