AFL Stats Resource: Wheelo Ratings

[WheeloRatings 906]

For those who aren't aware, I manage a website Wheelo Ratings that has a range of AFL stats available. I have added quite a bit of functionality since last season, including player and team stats back to 2012, individual player profiles, match stats, and interactive charts.

Here are some of the key pages and features. I will probably add new functionality as the season progresses but if anyone has any suggested features, let me know.

Player Stats (2012-)
https://www.wheeloratings.com/afl_stats.html

Key features:

• View all statistical categories or save your favourites.
• Includes 60+ stats, including AFL player ratings, coaches votes, centre bounce attendances, one-on-one contests, ruck contests and all your usual stats.
• Filter by team, position, etc.
• Interactive scatter plot, with the option of highlighting players from a particular team.
• Download data as csv file.

Team Stats (2012-)
https://www.wheeloratings.com/afl_stats_team.html

Key features:

• Team averages, opposition averages, differential.
• Interactive scatter plot.
• Download data as csv file.

Match Stats (2012-)
https://www.wheeloratings.com/afl_match_stats.html

Key features:

• View specific match or all matches for a particular round.
• Interactive scatter plot.
• Download data as csv file.

Grand Final 2021: https://www.wheeloratings.com/afl_match_stats.html?ID=20212701

Player Profiles (players who have played AFL since 2012)
https://www.wheeloratings.com/afl_player_profile.html

Key features:

• View stats for each season, including overall career.
• Trend chart by season.
• Compare a player with up to four other players.
• Search for players or navigate from links on player stats and other relevant pages.

Clayton Oliver: https://www.wheeloratings.com/afl_player_profile.html?ID=2db9dd
Christian Petracca: https://www.wheeloratings.com/afl_player_profile.html?ID=231162
Max Gawn: https://www.wheeloratings.com/afl_player_profile.html?ID=22f360
Jack Viney: https://www.wheeloratings.com/afl_player_profile.html?ID=22f8be
Steven May: https://www.wheeloratings.com/afl_player_profile.html?ID=22ce7d
Bayley Fritsch: https://www.wheeloratings.com/afl_player_profile.html?ID=14074de
Angus Brayshaw: https://www.wheeloratings.com/afl_player_profile.html?ID=2311a7

Team Annual Summary (1897-)
https://www.wheeloratings.com/afl_annual_summary.html

Key features:

• View win/loss records and team ratings for all seasons in VFL/AFL history.
• Interactive scatter plot.

[Demonland 64,807]

This is great. Thanks @WheeloRatings  

We’ll have to get you on the podcast one week. @binman will be in heaven. 

[whatwhat say what 18,918]

nice one

[layzie 28,721]

It's good to hear from you @WheeloRatings

I promise I won't do this much this year but the novelty hasn't quite worn off yet.

[mo64 5,209]

This is awesome @WheeloRatings. I'm wondering if it could be pinned up the top because it's a fantastic reference.

[Sydney_Demon 1,175]

I love your analysis @WheeloRatings which is a fantastic resourse you have made available. One thing though that I'm not fully understanding is how your simulations are producing Melbourne as Premiership favourites:

https://www.wheeloratings.com/afl_simulations.html

The prediction at the end of the H&A is Collingwood 1st, Port 2nd, Brisbane 3rd, Melbourne 4th. That to me would lead to likely Qualifying wins for Collingwood & Port, highly likely Semi-Final wins for Brisbane & Melbourne, but then I would would think Port should be favourites against Melbourne (conversely, if Brisbane beats Port in the Qualifying, Brisbane would be favourites against Melbourne). So logically Melbourne would be most likely to exit at Prelim stage. 

You have Melbourne a 39.4% chance of playing in the Grand Final with Collingwood only a 35.4% chance. I would have thought that having Collingwood only a 53% chance of winning if they make the Grand Final seems remarkably low, given they would be favourites against Melbourne & strong favourites against either Port or Brisbane. Melbourne is given a 59% chance if they make the Grand Final, which seems high given very likely their opponent would be Collingwood.

I know I'm only looking at likely opponents and your simulations would include a number of different scenarios with Melbourne coming from other positions than 4th but the numbers seem a little incongruous, given that since the current AFL Finals 8 system came in in 2000 no team has won from 4th!

[WheeloRatings 906]

5 hours ago, Sydney_Demon said:

I love your analysis @WheeloRatings which is a fantastic resourse you have made available. One thing though that I'm not fully understanding is how your simulations are producing Melbourne as Premiership favourites:

https://www.wheeloratings.com/afl_simulations.html

The prediction at the end of the H&A is Collingwood 1st, Port 2nd, Brisbane 3rd, Melbourne 4th. That to me would lead to likely Qualifying wins for Collingwood & Port, highly likely Semi-Final wins for Brisbane & Melbourne, but then I would would think Port should be favourites against Melbourne (conversely, if Brisbane beats Port in the Qualifying, Brisbane would be favourites against Melbourne). So logically Melbourne would be most likely to exit at Prelim stage. 

You have Melbourne a 39.4% chance of playing in the Grand Final with Collingwood only a 35.4% chance. I would have thought that having Collingwood only a 53% chance of winning if they make the Grand Final seems remarkably low, given they would be favourites against Melbourne & strong favourites against either Port or Brisbane. Melbourne is given a 59% chance if they make the Grand Final, which seems high given very likely their opponent would be Collingwood.

I know I'm only looking at likely opponents and your simulations would include a number of different scenarios with Melbourne coming from other positions than 4th but the numbers seem a little incongruous, given that since the current AFL Finals 8 system came in in 2000 no team has won from 4th!

It's largely due to Melbourne being ranked #1 on team ratings by my model. Believe it or not, I actually have Melbourne favourites against Collingwood on King's Birthday, so a Collingwood v Melbourne final would be close to 50/50, or even Melbourne favourites. That's all based on the fact my model currently rates Melbourne 6.2 points higher than Collingwood, which I don't personally agree with, just to be clear. Collingwood's rating actually decreased from rounds 7 to 10 as their opponents were particular inaccurate (Adelaide 7.16, Sydney 6.12, GWS 7.13, and Carlton 7.15) and then didn't beat North by as much as expected.

Even though I have Melbourne ranked fourth on the aggregate ladder, they're still more likely to finish second or third than fourth if you look at the ladder distribution. Melbourne's strong percentage definitely helps. I should note that my model doesn't look at the historical record of teams finishing fourth, it simply looks at the individual finals match up for the given simulation. FYI, here are the premiership probabilities for all teams finishing in a particular ladder position this year, with fourth being much higher than historical results with the current final 8 system:

1: 31%
2: 24%
3: 17%
4: 14%
5: 5%
6: 4%
7: 3%
8: 2%

[Sydney_Demon 1,175]

53 minutes ago, WheeloRatings said:

It's largely due to Melbourne being ranked #1 on team ratings by my model. Believe it or not, I actually have Melbourne favourites against Collingwood on King's Birthday, so a Collingwood v Melbourne final would be close to 50/50, or even Melbourne favourites. That's all based on the fact my model currently rates Melbourne 6.2 points higher than Collingwood, which I don't personally agree with, just to be clear. Collingwood's rating actually decreased from rounds 7 to 10 as their opponents were particular inaccurate (Adelaide 7.16, Sydney 6.12, GWS 7.13, and Carlton 7.15) and then didn't beat North by as much as expected.

Even though I have Melbourne ranked fourth on the aggregate ladder, they're still more likely to finish second or third than fourth if you look at the ladder distribution. Melbourne's strong percentage definitely helps. I should note that my model doesn't look at the historical record of teams finishing fourth, it simply looks at the individual finals match up for the given simulation. FYI, here are the premiership probabilities for all teams finishing in a particular ladder position this year, with fourth being much higher than historical results with the current final 8 system:

1: 31%
2: 24%
3: 17%
4: 14%
5: 5%
6: 4%
7: 3%
8: 2%

Thanks @WheeloRatings for your clarifications on this and on The Run Home thread. Obviously ignoring any home ground advantage or other factors including team fatigue or the opposite (basically assuming 50% win odds for every team for every game) would produce premiership odds of 18.75% for Positions 1 to 4 and 6.75% for Teams 5-8, so your rankings including all factors make sense. I would assume the biggest variation factor would be higher ranked teams generally having higher team ratings.

I have this very unscientific take on the fact that Teams 1, 2 & 3 have pretty much an equal split of premierships since 2000 (with Western Bulldogs being the outlier coming from 7th in 2016). Basically, the 4th team has a very tough series of games (typically against 1st, 5th, 2nd, 1st teams) , whereas 2nd/3rd teams typically play 2nd/3rd, 4th, 1st, a much easier run. The top team has an easier 1st game but then usually has to play a battle-hardened 2nd/3rd team in the Prelim and then the other 2nd/3rd team in the Grand Final (who has had ean easier Prelim against a lower-ranked team). If the 1st team does unexpectedly lose against the 4th they then have the same tough run the 4th team more typically has.  I'm not sure how these possible paths play out in simulations but am assuming the fact that Melbourne is modelled to more likely finish 2nd or 3rd than 4th means they benefit compared to a team that has a higher prediction of finishing 4th. I'm also assuming Brisbane & Ports' premiership odds are increased by them having to play the Grand Final In Melbourne against a Melbourne-based team, especially as Collingwood & Melbourne are MCG co-tenants. 

[WheeloRatings 906]

13 hours ago, Sydney_Demon said:

Thanks @WheeloRatings for your clarifications on this and on The Run Home thread. Obviously ignoring any home ground advantage or other factors including team fatigue or the opposite (basically assuming 50% win odds for every team for every game) would produce premiership odds of 18.75% for Positions 1 to 4 and 6.75% for Teams 5-8, so your rankings including all factors make sense. I would assume the biggest variation factor would be higher ranked teams generally having higher team ratings.

The home advantage is probably the biggest factor overall actually, but mainly for matches played between teams from different states.

13 hours ago, Sydney_Demon said:

I have this very unscientific take on the fact that Teams 1, 2 & 3 have pretty much an equal split of premierships since 2000 (with Western Bulldogs being the outlier coming from 7th in 2016). Basically, the 4th team has a very tough series of games (typically against 1st, 5th, 2nd, 1st teams) , whereas 2nd/3rd teams typically play 2nd/3rd, 4th, 1st, a much easier run. The top team has an easier 1st game but then usually has to play a battle-hardened 2nd/3rd team in the Prelim and then the other 2nd/3rd team in the Grand Final (who has had ean easier Prelim against a lower-ranked team). If the 1st team does unexpectedly lose against the 4th they then have the same tough run the 4th team more typically has.  I'm not sure how these possible paths play out in simulations but am assuming the fact that Melbourne is modelled to more likely finish 2nd or 3rd than 4th means they benefit compared to a team that has a higher prediction of finishing 4th. I'm also assuming Brisbane & Ports' premiership odds are increased by them having to play the Grand Final In Melbourne against a Melbourne-based team, especially as Collingwood & Melbourne are MCG co-tenants. 

That sounds like a reasonable theory. I haven't analysed the different paths through the finals in the simulation results, but one thing it won't take into consideration is 1st playing a battle-hardened 2nd/3rd in the prelim. The simulations just see it as 1st being the home team against a likely lower-rated opponent so would win more often than not.

Brisbane and Port's premiership chances are definitely negatively affected by having to play the GF at the MCG, but Brisbane is affected more.