Round 23 Prediction – shuffling the pieces


While our fight for the top 8 is basically over, there remains some important games this weekend. Our top 4 is certainly not set, with only 1 win separating 1st from 7th. As such, our match importance rating has thrown up some big games this week – largely because teams chances of various finishing positions fall close to 100% or 0% based on a win or a loss, obviously dependant on other results.
Continue reading “Round 23 Prediction – shuffling the pieces”

Round 23 Ratings and Simulations – Set in stone

As the Dee’s finals hopes melted away like the snow their fans had been enjoying all winter, our final 8 became set in stone. Barring North losing by a record margin and St Kilda winning by one, we won’t be seeing any changes in our finals teams.

Continue reading “Round 23 Ratings and Simulations – Set in stone”

Round 22 Predictions – Ph-inishe-D

After having a rough start to the week (I busted my ankle playing local Aussie Rules in Brisbane), I’ve had a monumentous end, resubmitting my PhD thesis to the UQ Graduate School after responding to the examiners comments. Fingers crossed that gets ticked off soon! The downside of this is I’ve run out of time to get a decent preview up unfortunately.

My predictions are below however I won’t include a writeup this week. I’m off to celebrate 🙂

Continue reading “Round 22 Predictions – Ph-inishe-D”

Round 22 Ratings and Simulations – Sydney and Adelaide standout

As we enter our final 2 rounds of the season, we continue to see two standout teams. North has also failed to again lock up a finals position – leaving the door ajar with their loss to the Hawks

Continue reading “Round 22 Ratings and Simulations – Sydney and Adelaide standout”

Round 21 Results

I must admit to not catching a lot of footy over the weekend due to having Olympics fever, which is probably good considering my ELO model had a tough round this week! With a few upsets, and 5 games decided by a goal or less, we struggled picking winners this week. The model managed to tip 5 out of 9 with a mean absolute prediction error (MAPE) of 27.7.
Continue reading “Round 21 Results”

Round 21 Predictions – Saints won’t go away

While I keep writing that our top 8 has been set for awhile, the chance of someone jumping into the finals hasn’t gone away. The Saints are still a 17% of making it and the order of our top 4 in particular is still very fluid meaning there is still plenty of important games!
Continue reading “Round 21 Predictions – Saints won’t go away”

Round 21 Ratings and Simulations – Hawks back to the pack

With a big upset loss on the weekend, the Hawks is the biggest loser from the weekend results. They have dropped back into the pack of clustered teams below our clear leaders and now face a possibility of losing top spot before the season is out.

New ratings

In fact – despite having a relatively good week for head to head tipping, our ELO model has shown some considerable adjustment to the rankings this week. This relates to some big upsets and misses in the margin, which we ultimately use to reassess our team ratings. We’ve seen 12 teams change places on our table (although only 5 of our top 8). The biggest changes come out of the Hawthorn/Melbourne and St Kilda/Carlton games.

In our top 8 race, Sydney and Adelaide have consolidated their grip as the clearly the two best rated teams. They have both opened up their lead on the remainder of the top 9 ranked teams apart from West Coast. In fact, Sydney would now start favourite against any team other than Adelaide regardless of the home ground advantage (HGA) boost (nominally set at 35 rating points). Adelaide is approaching that status, sitting 30 rating points above the next best side in Geelong.

plot of chunk unnamed-chunk-1

In the race for the worst team for the season, the Bombers have jumped above the Lions. The Lions have in fact edged close dropping below 1300 mark, which is almost 8 goals worse than an average team. In fact, only 12 teams in history have dropped below that ELO rating within a season. I promise I’ll write a post before the season is out about these two teams and where they sit in historical seasons but depending on how we measure ‘worst’ seasons by a team, they are on track to both finish within the worst 20.


Onto our simulations, and I’ve finally got around to adding in some finals sims!

Firstly to our finals race, and we know have 5 teams that have clinched a finals spot. Two more teams (Bulldogs and WCE) will likely add to that list with a win this week, with only North Melbourne any realistic chance of dropping out. St Kilda improved their chances to jumping into the 8 to 17%, to be the most likely. They need to make up 2 games and percentage from the remaining 3 games to do that, but based on the relative strengths of the teams involved, we’d expect it to occur almost 1 in 5 times, so the excitement continues!
The top 4 race remains likely to come from 5 teams. The Hawks unexpected loss on the weekend sees their chances drop considerably but still remain at around 3 in every 4 sims. We actually give Geelong and Sydney more chances of making it, despite sitting 1 game behind the Hawks.

We see a similar distribution in top 2 chances, however interestingly  the Hawks are still our favourite to clinch the Minor Premiership, although their chances dropped from 67% down to 29% this week.

Finals Sims!

Our first set of finals simulations and we are again favouring the same 5 teams as our most likely premiers. Our predictions for the Prelim are very similar to our top 4 chances, with only GWS having less than a 64% chance of making it that far. In terms of the grand final, Sydney are our favourites to make it, reaching the final weekend in almost half of all simulations. They also win that match in about 1/4 of simulations, making them our premiership favourites.

It will be interesting to see how these finals simulations progress – as Hawks showed last week, an unexpected loss could change the picture dramatically!

plot of chunk unnamed-chunk-2

Round 20 Results

For the second weekend in a row, the ELO model has done well with its predictions albeit without being very accurate. Round 20 saw the ELO model tip a total of 7 out of 9 with a mean absolute prediction error (MAPE) of 34.7.

The two main misses for tipping were considered upset wins. The first one was the Friday night clash between Richmond and Collingwood. Our model gave the Pies the slight edge at 52% however Richmond were able to overcome the Deledio effect, winning against expectation without the star. The other big miss was Melbourne’s upset win over Hawthorn, which will have some big effects on our expectations for the remainder of the season. I’ll release our simulations tomorrow and discuss these effects. A few other games were quite a lot closer (see GWS v Gold Coast) or much bigger blowouts (see St Kilda v Carlton) than we expected, driving up our MAPE.

These results lift our season total to 124 correct tips out of 171 games (72.5%), with a MAPE of 29.4

Results Plots

Our results plots show that we’ve been tracking along nicely with our tipping percentage over the last period, although our accuracy in the margin has been slipping above our goal.

plot of chunk unnamed-chunk-2plot of chunk unnamed-chunk-2

At the end of the season I’ll do a bit more work on this plot as a potential source of improvement for our model. For now, I’ll keep posting it for reference.

plot of chunk unnamed-chunk-2

Round 20 Prediction – shuffling the pieces

Leading into the final month of the season, match simulations suggesting that the top 8 is basically set. In fact, the top 8 hasn’t changed since before Round 7. As we wrote about at the time, that isn’t all that surprising. Given that – we’d perhaps expect that the remaining games aren’t all that important, as measured by our match importance metric. However, since very little separates teams in our top 8, matches involving those teams are particular important this week!

Round Predictions


Unlike last week – our ELO model sees some harder to predict matches this week. There are few matchups between teams close on our ELO rankings. While our two closest games have very little ‘importance’ in terms of the season, one could argue that each team has a fair bit to play for. Richmond v Collingwood on Friday night shapes as a big match for a Tigers team under siege, while Carlton v St Kilda is probably the key match between two up and coming teams on the tail end of a rebuild. As per our norm though – I’ll discuss some of the more important matches below.

Match Importance

plot of chunk unnamed-chunk-5
Sydney v Port Adelaide, Sat 1:45pm, SGC.

Port Adelaide continues to keep themselves in with an (albeit) small chance of making the top 8. This match looks like being their last chance to remain in with a realistic chance. A win sees their probability jump to ~20%, while a loss sees it drop down to <2%. Given Sydney is one of 4 of our best rated teams vying for 3 spots in the top 4, losing this match puts a big dent in their hopes of top 4. We see a 40 percentage point swing between a win and a loss. Given the relative strength of Sydney and their home ground advantage (HGA), our ELO model is giving Sydney a 63% chance of winning this one. Sydney by 21 points. 

Western Bulldogs v North Melbourne, Sat 7:25pm, Etihad.

Both the Bulldogs and Kangaroos have been struggling of recent weeks for varying reasons. In this time, they’ve dropped to 9th and 8th respectively in our ELO ratings. The Bulldogs early season work does see them more than likely to hold onto 8th spot but if they want to make a run for top 4, this match is important, showing a 35 percentage point swing based on a win or a loss. For North, they’re biggest risk is dropping out of the 8 and a loss here sees those chances jump to 60% probability of missing! The Bulldogs, despite a slight rating deficit, gets the edge here with HGA, with ELO model placing them at a 53% chance in a hard one to predict. Bulldogs by 5 points. 


Weighted Importance

The remaining matches – while seemingly important – highlight the need for me to weight these importance ratings based on the probability of the win or loss actually occurring. As an example this weekend – a win or a loss sees their top 4 hopes change by almost 50 percentage points. The actually probability of the Lions upsetting the Crows however from our ELO model is only 12%. I’m not sure I’ll have time before the season to fix this as I want to get my finals simulations completed but it is something to note!