Iron
Norm Smith Medallist
- May 26, 2014
- 7,170
- 15,280
- AFL Club
- Adelaide
A hangover from what? PTSD maybe.
Champion Data simulated season
- Thread starter Luv_our_club
- Start date
- Tagged users None
CD publishing this s**t doesn’t help them. I’m all for stats. They’re great for determining HOW games are won and HOW to beat sides.
But putting an arbitrary value on stats is ridiculous. For example, a team that likes to handball and share it around will likely have far higher uncontested possessions, and probably higher disposal efficiency. A team that goes as direct as possible with long kicks will have Lower uncontested possessions and a lower disposal efficiency. CD will rate the handball team higher, but in reality that may not be the case.
The best use for those stats is clubs trying to work out how to beat a team in a few days. Rather than watching 4 or 5 games with only a vague idea of what they are looking for, they can skip through 6 or 7 looking for specific examples of things and gain a greater understanding of the intricate details of the oppositions game plan.
But this is basically just SuperCoach. It’s basically click bait.
But putting an arbitrary value on stats is ridiculous. For example, a team that likes to handball and share it around will likely have far higher uncontested possessions, and probably higher disposal efficiency. A team that goes as direct as possible with long kicks will have Lower uncontested possessions and a lower disposal efficiency. CD will rate the handball team higher, but in reality that may not be the case.
The best use for those stats is clubs trying to work out how to beat a team in a few days. Rather than watching 4 or 5 games with only a vague idea of what they are looking for, they can skip through 6 or 7 looking for specific examples of things and gain a greater understanding of the intricate details of the oppositions game plan.
But this is basically just SuperCoach. It’s basically click bait.
Footy statistics that don't account for improvement in players and players dropping off are over valued in football analysis.
- Aug 26, 2016
- 12,136
- 15,082
- AFL Club
- Collingwood
You ******* beauty!!
- Aug 26, 2016
- 12,136
- 15,082
- AFL Club
- Collingwood
The only error i can see is they have Richmond playing finals.
Bluelegs
Celebrating what we're good at
Inb4 David King starts quoting this as fact.
TJ Hendricks
All Australian
I'm not good linking on my mobile but it went
GWS
Sydney
Western
Adelaide
Port
West Coast
Collingwood
Geelong
Hawthorn
Richmond
Fremantle
Melbourne
St Kilda
Gold Coast
Essendon
North Melbourne
Carlton
Brisbane
- Feb 17, 2016
- 3,196
- 6,674
- AFL Club
- Fremantle
They got Brisbane right, not that many got that one wrong last year
I'm actually astonished at how average and selective Champion Data's analysis' are.
Even in games like soccer and basketball - sports that are far easier to predict based on list talent and performance - aren't as predictive with a previous season's stats as Champion Data are.
They are really clutching at straws with some of their rankings. Me thinks this is the AFL pushing them into making the statistics more accessible.
Fact is that the sport itself is extremely anarchic and unpredictable. Most preseason predictions are carbon copies of the season prior's results, with maybe some minor tweaks. No one takes risks, and even the data is bizarrely limited in its theoretical application of performance.
Based on stats from season prior, which is pointless analysis.
The better analysis would be the rate of growth relative to previous alterations in the data. I could have come up with this model in year 9 advanced mathematics.
Even in games like soccer and basketball - sports that are far easier to predict based on list talent and performance - aren't as predictive with a previous season's stats as Champion Data are.
They are really clutching at straws with some of their rankings. Me thinks this is the AFL pushing them into making the statistics more accessible.
Fact is that the sport itself is extremely anarchic and unpredictable. Most preseason predictions are carbon copies of the season prior's results, with maybe some minor tweaks. No one takes risks, and even the data is bizarrely limited in its theoretical application of performance.
Based on stats from season prior, which is pointless analysis.
The better analysis would be the rate of growth relative to previous alterations in the data. I could have come up with this model in year 9 advanced mathematics.
That list clogger from port
ydraw
Norm Smith Medallist
- Apr 27, 2015
- 5,515
- 6,325
- AFL Club
- Fremantle
Is it though?
A proper statistical model would take into account changes over time - e.g. Someone like Buddy Franklin or Mark LeCras' output is likely to decrease this year because their best footy is behind them. Likewise players in their early 20s would be expected to increase their output.
But these ratings simply total up the ratings from the last 2 years and assume each player and team will be the same as they have been the last 2 years.
Kane McGoodwin
Make me an Admin!
- May 21, 2001
- 81,513
- 82,127
- AFL Club
- Adelaide
- Other Teams
- Adelaide Crows
Problem with Champion Data is they don't apply ratings to coaches like they do the players...
Kane McGoodwin
Make me an Admin!
- May 21, 2001
- 81,513
- 82,127
- AFL Club
- Adelaide
- Other Teams
- Adelaide Crows
Statistics are so black & white.
CD love Collingwood because they rack up large contested possessions. Now, yes, teams which do win the contested possession count do generally win the game. But Collingwood, a team which has a poor backline, a poor forwardline and a midfield that doesn't have the foot skills to take advantage of all the contested possession won. It's misleading when CD rank such teams highly, especially given Collingwood love to handball a lot in contested areas, which inflates their numbers, but ignores the fact that what they actually do with those handballs matters. They also win the inside 50 count often, which is generally a sign a side is playing good, attacking football and is likely to win, until you actually see their forward entries. And this is borne out in the stats, with 2017 Collingwood being one of the least efficient teams in terms of generating scores from inside 50s.
BEEG
Hall of Famer
- Aug 2, 2007
- 31,046
- 37,284
- AFL Club
- Collingwood
- Other Teams
- Dallas Mavericks
Melbourne 8th hehehehe
captain blood 17
El Jefe
- Dec 9, 2006
- 50,329
- 111,993
- AFL Club
- Richmond
- Other Teams
- Aston Villa Traralgon Football Club
Confirms a lot of things i think about with regard to Champion Data.
Sent from my iPad using Tapatalk
Sent from my iPad using Tapatalk
2017 Collingwood finished with a better % than Dogs, Saints, Hawks.
We were only 6-7% points behind Dons, Dees and Eagles.
But by losing so many winnable games, we ended 13th and get a bottom 6 favourable fixture in 2018.
So predicting us to win a couple more games than those teams really isn’t that dramatic.
The problem with us is that the once more plenty of key players have had interrupted pre-seasons...with a tough first month we really need to hit the ground running.
Its not actually that bad.
Richmond played better than their list suggested
Collingwood underperformed
Geelong did better than expected
Other than that, what's their margin of error? 2 games?
It is what it is.
I do agree that there are some key factors for Growth.
You're never gonna ever build a successful model, there are way too many variables. The maximum for predictions is about 82% for a season.
It doesn't take into consideration wet weather, finishing strength, close finishes etc.
- Banned
- #45
champion data, who rate shane edwards (who i actually like) as better than trent cotchin
No they don’t.
They class Edwards a ‘forward mid’ and compared to other players they categorise as a ‘forward mid’ Titch is elite
Cotchin is classed as a midfielder.
Snake_Baker
The one true King of the North
- Apr 24, 2013
- 81,024
- 153,169
- AFL Club
- North Melbourne
- Other Teams
- Essendon Lawn Bowls Club
- Banned
- #47
They've simulated the 2018 season 100,000 times based on their team and player rankings to come out with the result, which seems like it'll have the Pies in finals more often than not.
The top 8 is just about right and so is the bottom 4.
The Doggies have claims for the 8, but the rest of them look about right to me in the 9-14 bracket also.
GC is a little puzzling. CD must be expecting their kids to click.
Last edited:
I thought the model is supposed to use historical data to explain the success of teams based on a composition of ratings of the players in that team. In that case, if the model is working properly, it should actually closely reflect last years ladder at the end of the season, with the only possible deviations in overall team ranking based on off-season player movements?
