The credibility of champion data stats/ranking points

["Black Kavanagh"]

For a while now champion data stats and player rating points have seemed iffy, but these three examples for me prove that it's all one big joke of a system at times.



Marcus Bontempelli (Rd 12 vs Sydney, last night): 18 disposals, 38% DE, 6 clangers, 11 tackles, 0.1 - 92 RANKING POINTS

Hugh Greenwood (Rd 11 vs Geelong): 18 disposals, 55% DE, 11 tackles - 72 RANKING POINTS

JP Kennedy (Rd 12 vs WB, last night - BOG): 37 disposals, 51% DE, 3 tackles, 1 goal - 101 RANKING POINTS

Looking at these it can be seen that Greenwood's game last round was comparable to if not better than Bontempelli's game last night - however somehow recorded 20 less ranking points.

Also from last night, Josh Kennedy who was a fairly unanimous BOG only attained 9 more ranking points than Bontempelli - suggesting he was only 10% better than Bontempelli which is pretty ludacris imo. Apparently CD don't count contested possessions alone as they are made up of other things, with i50's and clearances not counting AT ALL (WTF?)

Can we just scrap this "complicated algorithm" dribble and just be done with these fancy "advanced" rankings? I don't have an explanation for the above examples, can never substitute watching the game as it is to determine the most influential players.

 

[Jobe Watson]

Here's how Bontempelli got his points last night:

The Ranking Points system is far from perfect (it will never be able to give Key Defenders a proper score based on their influence, for example), but it's the best method we have for combining the dozens upon dozens of stats recorded each game into one simple number that can be used to measure a player's output in a game.

 

["Black Kavanagh"]

This seems to be a better indicator - however I'm not sure how the two can spit out such different results

 

[Dadda Dave]

I really hope the Crows give the saints a good 100 point pasting tonight !!..Eddie to kick 6..

 

[Fadge]

This seems to be a better indicator - however I'm not sure how the two can spit out such different results

That system is still extremely flawed.

Bontempelli was not that bad.
Hunter was nowhere near third best on ground.
Kennedy ranked 11th best and Franklin 23rd?

Forget about the algorithms - just have a knowledgeable unbiased observer watch the game and distribute the pool of points amongst each player on the ground.

I still think that 200 points being about the maximum anyone can reach is underselling some of the best individual performances in any given game.

Think Carey in his best games - 20 marks (mostly contested) and 8 goals, Ablett's 9 in the 89 GF, Lockett and Dunstall's bags of goals. Those efforts should be approaching 400 points...

 

[Bazza97]

That system is still extremely flawed.

Bontempelli was not that bad.
Hunter was nowhere near third best on ground.
Kennedy ranked 11th best and Franklin 23rd?

Forget about the algorithms - just have a knowledgeable unbiased observer watch the game and distribute the pool of points amongst each player on the ground.

I still think that 200 points being about the maximum anyone can reach is underselling some of the best individual performances in any given game.

Think Carey in his best games - 20 marks (mostly contested) and 8 goals, Ablett's 9 in the 89 GF, Lockett and Dunstall's bags of goals. Those efforts should be approaching 400 points...

Yep, I don't put any stock into CD stats at all. It's a joke pretty much. I just go by what I see and it's better off doing that. Like you said, having a representative who just goes to watch the games and then rank the players without any stats is much more better then just going by a stats based system to rank players.

 

[Demonic Ascent]

Here's how Bontempelli got his points last night:

The Ranking Points system is far from perfect (it will never be able to give Key Defenders a proper score based on their influence, for example), but it's the best method we have for combining the dozens upon dozens of stats recorded each game into one simple number that can be used to measure a player's output in a game.

So the Bonts defensive efforts were what got him such a good ranking. Makes sense, I am one who think CD have done a great job with their advanced stats, just wish they were more accessible. I read the Prospectus at the start of every year, gives a great insight.

 

[TheMcManusNose]

I was amused when they rated Brad Hill something like 3rd worst on ground on for the Essendon game where he had 37 touches, 30 kicks with only 3 clangers, 16 marks and a goal.

I think he missed another 3 shots on goal (for the other 0.1) and only had 3 contested possessions and no tackles which is where the 'bad' rating has presumably come from.

But watching that game live it was ****ing amazingly obvious that he was BOG for us. No other player was spreading as hard and providing a link up option (which is the single worst thing about Freo this year, can't ****ing move the ball). Without him it would have been a grindfest we would have been lucky to win.

And that's the core of the problem with statistical measures like the ranking points. They're weighted in certain ways to try and measure impact in multiple areas which are often not part of players roles. Especially in a game where there is so many different roles and some events which are very difficult to measure. So there's occasional times where they will miss things or give too much emphasis to a player whose impact has turned out to be somewhat unimportant. Or visa versa.

 

[mr bagcroft]

If listen to David King, and his flog mate "Luffy" talk stats, , then you will quickly summise that Champion Data have no credibility at all. They just invent stats so they dont get bored and so they can try and sound a bit fancy.

 

[Belnakor]

Forget about the algorithms - just have a knowledgeable unbiased observer watch the game and distribute the pool of points amongst each player on the ground.

Have you thought about having potentially 3-4 people on the ground at all times who are really close to the action? They could then meet after the game and allocate points for the best player that day.

 

[Fadge]

Have you thought about having potentially 3-4 people on the ground at all times who are really close to the action? They could then meet after the game and allocate points for the best player that day.

Great idea!

That way all the midfielders would receive 90% of the fantasy points, leaving the remaining 10% to be allocated across the other 20 or so backs and forwards on the ground!

 

[blitzer]

In any fantasy football competition its important to remember that Bontempelli's total score is multiplied by the fantasy multiplier factor of 1.5.
That's why I got him for supercoach. 18 possessions? No problem - a few points off best on ground.

Some players are just a mystery how they score so well. Nankervis is another one. Gets beaten badly in hitouts and doesn't get that many touches but kills it in points.

FWIW i have both Bontempelli and Nankervis in my team.

 

[calyam]

It is always possible to find an individual result within these statistical systems that appears incorrect. But these measures shouldn't be used to simply assess an individual game - where there is a lot of noise and volatility - but rather to assess a players performance across a season or a few seasons.

Both measures are calculated very differently. The AFL Player Rankings are far more sophisticated than the SuperCoach points and so we could reasonably conclude that they will do a better job of ranking the games best players. The SuperCoach points are more sophisticated than DreamTeam points so they do a better job of ranking players.

Naturally there are complexities to the AFL that are difficult to quantify. The AFL Player Ratings are among the most sophisticated ranking systems in world sport but there will always be actions that either under or over-rated or not measured at all. That doesn't mean they should be ridiculed but it does mean that caution should be used.

I feel a little embarrassed for those who so vehemently deny the usefulness of statistical analysis. I don't know whether it is simply ignorance or fear of what you don't understand or even some misguided belief that you are capable of consuming and interpreting every action and signal that occurs during a football game. But the game is genuinely richer when we are capable of understanding it and statistical analysis helps to achieve that. It shouldn't be the only thing a person relies on - shoutout to David King - but it should be part of any well informed supporters toolkit.

 

[Fadge]

I don't think anyone in this thread has 'denied the usefulness of statistical analysis' (myself included).

What we are saying is there are far too many outliers in all games for the Champion Data scores to be an accurate assessment of the players' performance in a given game.

I watched the game on tv last night, didn't take any notes or look at any possession counts.

However, I am very confident that if I were to rate the players' performances relative to one another I could do a significantly better job than the ratings provided by the graph shown in post #3 of the thread.

 

[timtamWB]

Forget about the algorithms - just have a knowledgeable unbiased observer watch the game and distribute the pool of points amongst each player on the ground.

The problem with this is that there is no such thing as a knowledgeable unbiased observer. And even if there is, as soon as he does and he gives, let's say Sloane a 135 and B Crouch a 125 when a few people think Crouch played better, the vitriol that comes his way would be ridiculous.

 

[Scodog10]

Here's how Bontempelli got his points last night:

The Ranking Points system is far from perfect (it will never be able to give Key Defenders a proper score based on their influence, for example), but it's the best method we have for combining the dozens upon dozens of stats recorded each game into one simple number that can be used to measure a player's output in a game.

Reading through that and my primary takeaway is that some of the acts that usually go unrecognised aren't correctly weighted, but I can justify the scores each received. Basically it's a case of Kennedy needing to work harder defensively to get huge scores!

 

[TiggerMan22]

They're awful.

Here's how Bontempelli got his points last night:

The Ranking Points system is far from perfect (it will never be able to give Key Defenders a proper score based on their influence, for example), but it's the best method we have for combining the dozens upon dozens of stats recorded each game into one simple number that can be used to measure a player's output in a game.

Where are those stats from?

 

[Bardo State]

Why would I need 'statistics', 'numbers', and 'evidence' when someone on Bigfooty can tell me from 3/4 of a TV viewing with one eye on their phone exactly how each of the 44 players performed taking into account every action a player took or didn't take when involved in each play?

 

[The Swert]

When did Bontempelli make 11 tackles???

Does hugging a teammate after a goal count or something?

They should only count tackles that lead to a ball up or free kick.

No way his ineffective tackles were worth that much.

 

[Fadge]

Why would I need 'statistics', 'numbers', and 'evidence' when someone on Bigfooty can tell me from 3/4 of a TV viewing with one eye on their phone exactly how each of the 44 players performed taking into account every action a player took or didn't take when involved in each play?

Was Josh Kennedy the 11th best and most influential player on the ground last night? Franklin the 23rd?

They would have both been in my top 5, and I'm sure the top 5 of most other neutral observers. But not the Player Ratings above...

 

[citizen-erased]

When did Bontempelli make 11 tackles???

Does hugging a teammate after a goal count or something?

They should only count tackles that lead to a ball up or free kick.

No way his ineffective tackles were worth that much.

That stat surprised me from the night
I don't recall him being that effective at the contest.

 

[Fadge]

The problem with this is that there is no such thing as a knowledgeable unbiased observer. And even if there is, as soon as he does and he gives, let's say Sloane a 135 and B Crouch a 125 when a few people think Crouch played better, the vitriol that comes his way would be ridiculous.

I wouldn't have thought so.

What it would tease out are instances when Franklin's game is downgraded significantly because he kicks 3 goals 5 behinds.

Similarly, Bontempelli wouldn't have been the third worst player on the ground, only marginally ahead of Tim English, primarily because he had a number of dubious free kicks paid against him.

 

[Jobe Watson]

Where are those stats from?

Champion Data posted them on their Twitter account.

 

[Richo83]

Here's how Bontempelli got his points last night:

The Ranking Points system is far from perfect (it will never be able to give Key Defenders a proper score based on their influence, for example), but it's the best method we have for combining the dozens upon dozens of stats recorded each game into one simple number that can be used to measure a player's output in a game.

How come Kennedy got more for his short effective kicks than Bontempelli did?

 

[Richo83]

When did Bontempelli make 11 tackles???

Does hugging a teammate after a goal count or something?

They should only count tackles that lead to a ball up or free kick.

No way his ineffective tackles were worth that much.

It's not uncommon for players to do a bad handpass which spills free or goes to another player under the pump as a result of a tackle, you'd ignore them under your system.