Given that the fixture for Round 12 is as fluid as the entire season’s schedule was last year, we’re not only late this week, having waited for the final decision on sites, but we’ll be really quick about our picks this week because we still don’t trust what the weekend will look like and also want to talk about something else of predictive importance.
So, as of Tuesday night, here’s what the numbers from Following Football say about the Round 12 games.
Brisbane will defeat Melbourne straight up. With the game finally settled in at Giants Stadium, the line has been logical but wrong: Melbourne by 4.5 points. However, ELO-FF gives Brisbane a two-goal edge, and when you dig in you’ll see why. After a 1-3 start, the Lions have won seven straight games and are flying, while the Demons, although coming off an impressive controlling of the Bulldogs, have not been able to handle opponents the way that Brisbane has.
We could be wrong – we were picking against Melbourne last week – but we haven’t gone wrong picking for Brisbane in recent weeks. Our usual (purely hypothetical) $2 wager would pay out $4.20 if we’re right.
Sydney will defeat St Kilda – ELO-FF ratings say by 23 at the SCG. While neither the Crows nor Swans have been able to hold up their stellar starts, Sydney showed last week that they’re favourites to stay in the eight despite the competition immediately behind them. And St Kilda’s inconsistency make them a really good bet to fall short – both from finals and in Round 12 itself.
With Brodie Grundy down, ironically I think it’s a rally-the-troops moment for Collingwood. Our numbers say to take them against the 14.5-point spread at Adelaide, and that’s the reason I’ll agree with the numbers. Not that they will necessarily win, but they will keep it close, and I will bet that it stays close until the end.
Richmond will defeat Essendon, but our ratings show a much closer game than the 12.5-point spread the pundits are offering as of Tuesday evening. It was a point higher last night – punters know what’s what. Beware: we’re taking the Tigers only to win, not cover.
Carlton are favoured over the Eagles by 1.5 points; it’s basically a coin toss. If it’s at Optus, it would be an easy call: West Coast all the way. Anywhere else and the road rating for the Eagles plus Patrick Cripps’s improved play under pressure suggests Carlton deserves that slight edge and our hypothetical $2 wager, which was paying $3.66 as of Tuesday evening.
The Bulldogs are favoured by 17 over Fremantle. Our numbers say it should be 28 – it went up a point since opening at 16. Longtime readers know what I’m about to say: Footscray plus the points equals easy money over the long haul. Yes, I know it lost last week.
Like everyone else, we won Star Trek: Voyager last week (seven of nine); we lost one of our seven conservative bets – Geelong didn’t cover thanks to the fourth-quarter awakening of the Magpies – but the other six paid enough to make two per cent on the round. We are $9.50 up on $198 hypothetically wagered this season, which is less than five per cent but better than what the establishments want you to make.
Overall the ELO-FF numbers have been on the winning side of the outcomes 60-38 with one tie this season (61 per cent) and have picked the winners 72 of 99 times (73 per cent).
Here’s something we haven’t talked about all season: burst ratings.
Each team has a rating under the ELO-Following Football system that adjusts slightly every week depending on how far from expectations the game result is. Team rating changes are capped at ten points per game, and once it hits eight points a circuit-breaker slows down the change, and it does the same in garbage time of blowout games.
The important thing to remember about ELO ratings of any kind is that whatever one team adds to its rating the other team subtracts from its rating, so the overall total will always remain the same. In the AFL the average team rating is 50, so the total for the 18 teams in the league is always 900.
However, we can also start experimenting with something we call burst ratings this season. I’ll be honest: we stole the basic idea from a couple of other computer system folks who were doing this same thing – Matt Cowgill at The Arc was the first one I saw using this.
The point is to give us an idea about what a team looked like in a one-game sample rather than over the long term. It doesn’t seem to be useful by itself in terms of being predictive, but it might shed some light on how teams are faring over a specified length of time. Let’s use Friday night’s potential blockbuster game between Brisbane and Melbourne to show you how it’s used.
You read burst ratings the same way you read any of the ELO-Following Football ratings: 50 is the average, and the higher the better. But because they’re one-shot ratings, there isn’t any correlation between one week and the next except that they’re being produced by the same team. There are already teams that have hit triple-digit burst ratings one week and negatives the next.
Here, for example, are the burst ratings for the first six games of the season for the 10-1 top-of-the-ladder Melbourne Demons – all wins, all well above average:
Now look at their most recent five match-ups – objectively overall no more difficult than the first six:
To make the distinction more dramatic, take out the victory over the Bulldogs – the average for the four weeks leading up to it was just 52.6, or 32 points lower than the previous six games. And since they’d undoubtedly have been aiming for that game, it makes sense that the ones on either side of were lower.
By comparison, here are the first four games for the then 1-3 Brisbane Lions:
And here are the seven subsequent wins for the now 8-3 Lions:
We haven’t done enough data collection or analysis on burst ratings to know how functional or predictive they are, but we will, and if they’re worth the time to share with you in our analyses, we’ll definitely include them in our forecasts next season. As far as this goes, it’s at least anecdotal evidence to support our belief that Brisbane will do very well against the Demons on Friday night.
I did some messing around and discovered that a continuation of those two trends comes closest to continuing exactly if the Lions win by 20 points. The ‘normal’ ratings came up with a winning margin of a fraction under 12 for Brisbane. The burst ratings feel like anecdotal support of that nerve-wracking prediction – picking against the league leader is generally a fool’s game!