Another popular late-July activity is speculating about the finals and I think it's time we did a little of that.
So, using a simple model that allows me to predict the victory margin for a game based solely on the participating teams' MARS Ratings and whether or not the game is an Interstate Clash, I've projected the remaining six rounds of the season to determine which teams are most likely to extend their season into September and from what ladder position they're likely to commence their finals campaign.
The model I used for this was the following:
Predicted Victory Margin = 0.711959*(Own MARS - Opp MARS) + 13.9253*Interstate_Clash - 0.805606
In using this model I've assumed that teams' MARS Ratings remain constant at their current level for the remainder of the season.
This model explains about 26% of the variability in actual victory margins across the last 10-and-a-bit seasons, and seems to be about as good as I can manage without the predictive fillip that comes from knowledge of the bookie's prices for each game (and, right now, I doubt even he knows what they'll be for Rounds 18 to 22).
As a quick aside, it's interesting to note that this model, which was grown using Eureqa and is therefore in some sense optimal, does not use the Home_Team variable. The Interstate_Clash variable, it transpires, has about a +0.7 correlation with the Home_Team variable and so shares about half its variability with the Home_Team variable, but the Interstate_Clash variable has historically done a better job of explaining victory margins, and so makes it into the model while the Home_Team variable does not.
Here's how this simple model projects the six home-and-away rounds ahead of us:
This model explains about 26% of the variability in actual victory margins across the last 10-and-a-bit seasons, and seems to be about as good as I can manage without the predictive fillip that comes from knowledge of the bookie's prices for each game (and, right now, I doubt even he knows what they'll be for Rounds 18 to 22).
As a quick aside, it's interesting to note that this model, which was grown using Eureqa and is therefore in some sense optimal, does not use the Home_Team variable. The Interstate_Clash variable, it transpires, has about a +0.7 correlation with the Home_Team variable and so shares about half its variability with the Home_Team variable, but the Interstate_Clash variable has historically done a better job of explaining victory margins, and so makes it into the model while the Home_Team variable does not.
Here's how this simple model projects the six home-and-away rounds ahead of us:
Let's run through each team's projected future.
Collingwood win all but one of their remaining contests, losing to the Cats in Round 19 by just 2 points in what turns out to be the contest for the minor premiership. This leaves them ruing their draw with the Dees and heading into the finals in 3rd place. Aside from the loss to the Cats, the Pies other results are projected to be Comfortable or Big wins (ie by 13 points or more).
Geelong win all their remaining clashes, three of them - those against the Lions, Blues and Eagles - comfortably, and the other three by margins ranging from just 2 to 10 points. As a result, they finish 1st. If MARS is right, however, it's easy to see the Cats finishing lower down the ladder.
St Kilda also win all their remaining matches, and only those victories against the Hawks and the Crows seem to be in any way uncertain. They finish 2nd but only need the Cats to slip up once or twice to snatch the minor premiership.
Fremantle finish 4 and 2 to drop a top 4 spot to the Dogs and finish 5th. This assumes a Round 19 win against the Roos and a Round 21 loss to the Hawks. Altering either of these outcomes would probably change Freo's final ladder position.
The Western Bulldogs go 5 and 1 over the remainder of the season, which allows them to leapfrog Freo into 4th. Rounds 19 to 21 seem most likely to determine their final ladder position as they're projected to win two of these contests by 10 points or fewer and to lose the other by just 4 points.
Sydney finish 3 and 3, though none of the results are projected to have margins exceeding 13 points. With the projected finish they wind up 6th, but the narrowness of projected results suggests that a much higher or a much lower finish is possible.
Hawthorn also finish 3 and 3. Two of their losses are by margins of 12 points or fewer, and two of their wins are by only 13 or 14 points, so they too could easily finish a couple of spots higher or lower on the ladder. As it stands they are projected to finish 7th.
Carlton are another team projected to finish 3 and 3. Their Round 17 clash, which is projected to be a narrow win for them, is the game that is mostly likely to produce a different result. Barring that they are projected to finish 8th.
The Kangaroos are the first of the teams projected to complete the remainder of the season with a sub-50% performance. They're projected to go 2 and 4, though 4 of these results - 2 wins and 2 losses - are predicted to be decided by 6 points or fewer. They are projected to finish 10th but could snatch a spot in the eight if they do better than this and Carlton drop another game.
Adelaide are projected to finish 3 and 3, which secures them 9th spot. Four of their projected results - 2 wins and 2 losses - are predicted to be decided by 12 points or fewer, so it's also conceivable that they could make the final 8.
Melbourne are forecast to win only 1 of their remaining 6 clashes, which sees them drop to 12th on the final ladder. Four of those losses are narrow - 14 points or fewer - so a higher finish is possible, but a spot in the final 8 seems unlikely.
Essendon are also projected to wind up their season with just 1 additional win, which sees tham finish 14th on percentages (which again this year I've proxied by the difference in Points For and Points Against). Only this week's projected loss to the Roos by 5 points can be seen as reasonably reversible, so there's virtually no chance that we'll see them running around in September.
Brisbane Lions scrape 1 more win for the season and lose the last 4 games by 11 points or fewer, which sees them finish 13th. Finals football is, at the earliest, a 2011 project for them.
Port Adelaide win Rounds 20 and 21, which is enough to lift them into 11th spot by season's end. Four of their remaining games, including both of their wins, are projected to be decided by 13 points or fewer, so they could finish a couple of spots higher or lower than this, but a place in the finals looks impossible.
Richmond are forecast to lose the next 5 matches before finishing with a consolation victory in Round 22. Nonetheless, they finish 16th.
West Coast win two games, both narrowly and at home, to lift them to a final position of 15th. Two other games appear winnable, so there is some hope for a slightly higher finish.
Here's the projected final table:
No comments:
Post a Comment