FSI - Full Strength Indicator

[jazzadogs]

A new initiative by Champion Data this year, the Full Strength Indicator is designed to eliminate cries of "BUT WE WERE MISSING OUR BEST PLAYERS!!" from the mouths of Hawthorn supporters.

Champion data has selected a best 22/25 from each side, and each week ranks the selected 22/25 against this 'best' team, providing a percentage representative of how close a team is to full strength. Better players provide a greater percentage (I imagine if Riewoldt misses, St Kilda would be at about 40% :p ).

Whilst I think this is a good idea, I wonder as to how they developed the 'best 22', because there would seem to be anomalies. The most relevant example, obviously, is the Bulldogs.

According to Champion Data, our best 25 (positions not relevant) reads:

B: Harbrow Lake Morris
HB: Hargrave Callan Gilbee
C: Cross Boyd Hill
HF: Murphy Hahn Higgins
F: Akermanis Hall Giansiracusa
R: Hudson Cooney Griffen
Int from: Minson Addison Everitt Johnson Picken Ward Eagleton

By not having Eagleton and Callan in, we have an FSI of 96%.

A decent effort, however Williams and Moles would seem certainties in the eyes of most Bulldogs supporters, at least for a best 25 (Williams at least)?

This link will take you to the FSI for the Dogs vs Pies game. Link (http://www.heraldsun.com.au/sport/afl/match?matchid=1111112046393)

St Kilda have Lovett, Jesse Smith, Gardiner and Pattison in their 'best 22' ahead of Armitage, Peake, Gwilt and Geary.

Sydney has Henry Playfair in ahead of Jesse White.

Will people be taking any notice of this throughout the season? I think a lot more research needed to go into it, seems like a fairly haphazard attempt, but I imagine that we're going to have it rammed down our throats.

[AndrewP6]

Will people be taking any notice of this throughout the season? I think a lot more research needed to go into it, seems like a fairly haphazard attempt, but I imagine that we're going to have it rammed down our throats.

Only if we lose and our FSI is low ;)

[chef]

Williams and Moles would seem certainties in the eyes of most Bulldogs supporters, at least for a best 25 (Williams at least)?

St Kilda have Lovett, Jesse Smith, Gardiner and Pattison in their 'best 22' ahead of Armitage, Peake, Gwilt and Geary.

Sydney has Henry Playfair in ahead of Jesse White.


Not sure I like it that much(or take to much notice of it) if this is the case.

This is going to give supporters another excuse when they lose.

[ledge]

It seems it has to be taken on last years team, say a kid comes in and plays really well does he move into the top 25 or does this 25 stay the same all year?

[LostDoggy]

After reading the best 22/25 you mentioned for St Kilda jazzadogs - I think I will be ignoring this for the most part!

[mjp]

It will be based on CD Stats for the previous year by position I guess. Good idea...will evolve over time though as guys nudge their way ahead of others in the pecking order, or even change positions?

How could Callan not be ahead of Moles? Moles hasn't played in a game yet.

[LostDoggy]

It's a decent effort -- I'm with mjp in that Moles can't really be in our 'top 22/25' since he hasn't even played a game and was just promoted off the rookie list.

It also probably quantifies actual performance statistically rather than take into account things like potential or 'excitement generation'. In your Sydney example, Jesse White hasn't really set the world alight and is exciting because of what he is becoming, not what he already is. Clubs who are 'developing' their lists will play some players before their time just to fast-track them -- Jordy could justifiably consider himself part of Richmond's top 22 players but they'll play kids ahead of him. Statistically speaking, they are definitely then below their optimum current FSI, which is technically correct.

Williams would seem to fit in the same boat as Jesse White (in that he is more potential than performance at present) but, as with White, their value goes beyond individual performance to contribution to a team's structure BEYOND their immediate contribution (ie. allowing other players to play on suitable opponents etc), and that's probably where the FSI rating seems to fall down for me (especially when you say 'positions not relevant') as often a team is at its best not when all its highest ranked players are out on the park but when its structure is sound and everyone has the right match-ups.

[soupman]

I think whilst it's a good idea in theory, it should be marked slightly differently.

Only players unavailable through injury, suspension or otherwise should be included. That is, McMahon's absence from Richmonds 22 shouldn't be reflected in the FSI because he is available, but not selected. Wouldn't an available player not selected in the 22 be regarded as not in that sides best 22?

Another example will be this week and us. If say Brad Johnson misses through injury then his absence should be taken into account in the FSI. We can rightly say after the match that we weren't at full strength because Johnson didn't play due to injury.

However, say Ward doesn't play because we select Moles and Everitt ahead of him, meaning he plays for Willy. At the moment his absence would affect the FSI, however we cannot say after the match that we weren't at full strength because ward didn't play, because if he was that good he'd have been picked.

[mjp]

Only players unavailable through injury, suspension or otherwise should be included. That is, McMahon's absence from Richmonds 22 shouldn't be reflected in the FSI because he is available, but not selected. Wouldn't an available player not selected in the 22 be regarded as not in that sides best 22?


I agree, but they can't do that.

1/.Because they dont know WHY he wasn't picked. Just because the clubs say he isn't injured, doesn't mean he isn't injured.

2/.Because all of their assessments are based on numbers...and until the numbers say otherwise, Tim Callan is in the best 25 and Brodie Moles isn't.

[jazzadogs]

I agree, but they can't do that.

1/.Because they dont know WHY he wasn't picked. Just because the clubs say he isn't injured, doesn't mean he isn't injured.

2/.Because all of their assessments are based on numbers...and until the numbers say otherwise, Tim Callan is in the best 25 and Brodie Moles isn't.
Yeh, I read something to day saying it is based on a 'form assessment'. So over the course of the year, this 'best 22/25' will change to reflect the form of different players. I'd say if someone is out with injury though (e.g. Andy Otten from Adelaide) he will stay in that best 22 no matter how good the player who has replaced him becomes.

Needs clarification and adjustment, but I think it will become a good tool over time, just like SuperCoach scores have evolved to provide a better indication of performance.

[Mantis]

I notice this week we are at 94% in terms of the FSI.

In my mind we are missing 4 members of our best 22, Johnson, Aker, Ward & Picken. I understand that Champion Data use some special formula which takes into account SC ratings, etc.. but when I went to school 4/22 = 18.2%. So using extremely simple maths our FSI should be 81.8%

[Doc26]

I notice this week we are at 94% in terms of the FSI.

In my mind we are missing 4 members of our best 22, Johnson, Aker, Ward & Picken. I understand that Champion Data use some special formula which takes into account SC ratings, etc.. but when I went to school 4/22 = 18.2%. So using extremely simple maths our FSI should be 81.8%

Of course that's assuming each of the 22 players have an identical weighting which isn't reasonable. With that said the combination of Johnson, Akermanis, Ward and Picken out of our 22 must surely be greater than the indicated 6% gap although Cooney and Griff on current form are probably nudging 25% alone ;)

[Mantis]

Of course that's assuming each of the 22 players have an identical weighting which isn't reasonable. With that said the combination of Johnson, Akermanis, Ward and Picken out of our 22 must surely be greater than the indicated 6% gap although Cooney and Griff on current form are probably nudging 25% alone ;)

The system is totally flawed.

Example: Not knowing and without looking at stats (cause I can't be stuffed) Liam Picken is probably rated somewhere between 17-22 in terms of importance as he probably only gathers 10 touches a game and say 50pts. But his role in the team is to negate the influence of one of the oppositions more important players.

Last week Picken plays on Bateman and keeps him quiet and he gathers only gets 50 pts, which is far different to his season average of 100pts. (all made up)

Therefore Liam Picken as well picking up 50pts has saved us 50pts, meaning his net worth is 100pts ranking him up in the top 5 or 10. Same would go for someone like Dale Morris who is a pleb in terms Fantasy Footy rankings, but we all know his worth to the team is far greater than this ranking system.

Summary: The FSI is a big pile of shit!!!

[Doc26]

The system is totally flawed.

Summary: The FSI is a big pile of shit!!!

Certainly at face value would agree with you which also extends to the SuperCoach scoring concept which has little if no regard to any negating role or even on field leadership value. It's skewed heavily to the offensive side of the game.

[Sedat]

Summary: The FSI is a big pile of shit!!!
The Full Strength Indicator sounds more like a rating system for beer brands. It also sounds like a load of tosh.

[Sockeye Salmon]

The system is totally flawed.

Example: Not knowing and without looking at stats (cause I can't be stuffed) Liam Picken is probably rated somewhere between 17-22 in terms of importance as he probably only gathers 10 touches a game and say 50pts. But his role in the team is to negate the influence of one of the oppositions more important players.

Last week Picken plays on Bateman and keeps him quiet and he gathers only gets 50 pts, which is far different to his season average of 100pts. (all made up)

Therefore Liam Picken as well picking up 50pts has saved us 50pts, meaning his net worth is 100pts ranking him up in the top 5 or 10. Same would go for someone like Dale Morris who is a pleb in terms Fantasy Footy rankings, but we all know his worth to the team is far greater than this ranking system.

Summary: The FSI is a big pile of shit!!!

Picken kept Bateman to 15 pts!

Therefore Picken = 50 pts + 85 pts = 135 pts. Make Picken your Supercoach captain this week!

Picken > Ablett.

[Scorlibo]

I notice this week we are at 94% in terms of the FSI.

In my mind we are missing 4 members of our best 22, Johnson, Aker, Ward & Picken. I understand that Champion Data use some special formula which takes into account SC ratings, etc.. but when I went to school 4/22 = 18.2%. So using extremely simple maths our FSI should be 81.8%

Except we don't play with 18 players, 4 come in to replace the 4 that went out, and their 4 average SC scores are added to the total. Meaning that the FSI is the (total SC scores of selected 22)/(total SC of best 22). And the best 22 is chosen as the 22 highest scoring players. Fairly primitive system I guess but the concept of a FSI is one which I think we all take a bit of a fancy to, and so hopefully it will be furthered.


The system is totally flawed.

Example: Not knowing and without looking at stats (cause I can't be stuffed) Liam Picken is probably rated somewhere between 17-22 in terms of importance as he probably only gathers 10 touches a game and say 50pts. But his role in the team is to negate the influence of one of the oppositions more important players.

Last week Picken plays on Bateman and keeps him quiet and he gathers only gets 50 pts, which is far different to his season average of 100pts. (all made up)

Therefore Liam Picken as well picking up 50pts has saved us 50pts, meaning his net worth is 100pts ranking him up in the top 5 or 10. Same would go for someone like Dale Morris who is a pleb in terms Fantasy Footy rankings, but we all know his worth to the team is far greater than this ranking system.

Summary: The FSI is a big pile of shit!!!

It would be a really great idea for champion data to develop their rankings to account for what each player's opponent is doing. They now track matchups so intently that knowing exactly who was on who for what period of time wouldn't be an issue.

Problem is that at the moment the rankings (or SC scores) weigh in most defensive stats (spoils, tackles, 1%ers), and to implement a system where a player's defensive quality is calculated by their opponent's score (ie. Picken gets an extra 50 points for reducing Bateman to 50 below his average output), you would need to make the initial score a purely offensive ranking, so that defensive efforts wouldn't be overly represented; so that a defender wouldn't get points for a spoil and then get more points for how that spoil affected their opponent's game, but rather they would just get points for how the spoil affected their opponent's game. Where would you draw the line with offensive vs defensive plays? If Brian Lake takes one of his trademark marks, halting an opposition attack, does that count offensively or defensively? The truth is it counts both ways, but no one can ever know in what ratio, and in what ways it will affect a number of forwards, not just his direct opponent.

Of course there are other problems also:

What if Bateman was just playing badly? How much of his poor performance can one attribute to Picken's influence?

Do you work with raw points or percentages? Say if Picken got 60 offensive points, would he get those 60 points plus 50 points for 50 less of Bateman's usual? Or would he get 2*60 points for reducing Bateman to half of his usual output?

There are so many ifs and buts to do with effectively approximating the output of a player in regards to their opponent, but it is clear that defenders and taggers get a rough go under the current system, and it would be refreshing to see them accurately represented statistically. For the timebeing though, the FSI is working off the most complex statistical approximation we have, and I guess that's all you can ask for.

[LostDoggy]

It sounds like a bit of a crock..

Brisbane's was 85% or thereabouts?

How does that make sense? Who's missing from their best 22? Do Mitch Clark & Ash McGrath make up that 15% or something..:rolleyes:

[Mantis]

There are so many ifs and buts to do with effectively approximating the output of a player in regards to their opponent, but it is clear that defenders and taggers get a rough go under the current system, and it would be refreshing to see them accurately represented statistically. For the timebeing though, the FSI is working off the most complex statistical approximation we have, and I guess that's all you can ask for.

I am not asking for it. Who is?

[Scorlibo]

I am not asking for it. Who is?

Given that the the FSI exists, most people, as shown in this thread, are interested in its accuracy, and they're using the best statistical ranking system the footy world has, hence me saying, 'that's all you can ask for'. If you're saying that you don't want/don't ask for a FSI at all then sure, I understand that. But I guess the thing to note is that really every journo and every blogger who has ever made a comment about a side being close to or nowhere near 'full-strength' is guilty of being a FSI, so this really isn't a new thing, it's just a more mathematical approach to a common football matter. If they hadn't brought this in, there would be more ill-informed journos from the Herald-sun trying to shove erratic opinions on the teams down your throat anyway.

[Mantis]

Given that the the FSI exists, most people, as shown in this thread, are interested in its accuracy, and they're using the best statistical ranking system the footy world has, hence me saying, 'that's all you can ask for'. If you're saying that you don't want/don't ask for a FSI at all then sure, I understand that. But I guess the thing to note is that really every journo and every blogger who has ever made a comment about a side being close to or nowhere near 'full-strength' is guilty of being a FSI, so this really isn't a new thing, it's just a more mathematical approach to a common football matter. If they hadn't brought this in, there would be more ill-informed journos from the Herald-sun trying to shove erratic opinions on the teams down your throat anyway.

Which, as I previously stated is flawed.

You cannot, I mean CANNOT determine a players worth to any individual team via a mathematical model which uses players stats as a determining factor, I just can't have it.

[chef]

Which, as I previously stated is flawed.

You cannot, I mean CANNOT determine a players worth to any individual team via a mathematical model which uses players stats as a determining factor, I just can't have it.

Who decides who is our best 22 anyway?

Do they ask Rocket or just make it up themselves.

[Remi Moses]

Should be renamed the B/S indicator,honestly some people have far to much spare time on their hands:rolleyes:

[Scorlibo]

Which, as I previously stated is flawed.

You cannot, I mean CANNOT determine a players worth to any individual team via a mathematical model which uses players stats as a determining factor, I just can't have it.

You cannot determine a player's worth to any individual team via ANY model.

I don't disagree, the FSI is flawed to an extent. But my last point stands, it cannot be any more flawed than mere opinion, and at least it is consistent. Therefore, it cannot be a bad thing, because previously opinion is all that footy fans have had to gauge team selection.

[Doc26]

I don't disagree, the FSI is flawed to an extent. But my last point stands, it cannot be any more flawed than mere opinion, and at least it is consistent. Therefore, it cannot be a bad thing, because previously opinion is all that footy fans have had to gauge team selection.

I would tend to disagree with your last point, well at least when it comes to defenders and negaters of the game, that based on the current modelling at least it IS more flawed than opinion, where opinion does tend to recognise their positive value to a team ala a Morris and Picken. It remains the 'black hole' in their endeavour to equitably weight ALL players on their real value.

You can claim that it is more consistent in its approach although this doesn't inhibit it from also being consistently flawed. As with much opinion there is also subjectivity involved e.g in determining effectiveness, in determining who receives credit for a clearance in a chain of play etc and in interpretation variances associated across the many statisticians involved.

Swinburne and CD are continually looking at ways to enhance their modelling but for mine it remains skewed too heavily towards offense.

For those interested this was an article from Swinburne's magazine last year on CD looking to incorporate an 'Equity Ranking' model that looks to place weight on the significance of a given statistic although this still doesn't address the case for the negating defender.

http://www.swin.edu.au/magazine/5/110/football-numbers-man-brings-players-to-account/

[Scorlibo]

I would tend to disagree with your last point, well at least when it comes to defenders and negaters of the game, that based on the current modelling at least it IS more flawed than opinion, where opinion does tend to recognise their positive value to a team ala a Morris and Picken. It remains the 'black hole' in their endeavour to equitably weight ALL players on their real value.

I'm going to play Devil's Advocate here.

What is their 'real value'? Common opinion seems to be taken as gospel when it comes to player value, so naturally statistical models are shot down when they differ from the norm. The common perception is that Morris, Picken etc. are good players because they typically nullify their opponent (so their 'real value' is high), yet most of the acts of nullification are covered in CD's system: Spoils, tackles, 1%ers, tap ons, contested possessions, contested marks, marks from opposition kicks, ball gets, etc. So is it in the value of these stats that the results are supposedly skewed, if it is not in the quantity?

As far as I'm aware, Champion Data does more research than any other person or company or newspaper for that matter into the value of particular acts, in particular moments in time, in particular positions on the ground, to winning a game.

For all this I lay a great deal of trust in the SC scoring system, yet I am also of the view that Dale Morris deserved his All-Australian selection because he's a good player. So I am conflicted, but I don't feel it is fair to definitively say that CD's ranking system is flawed when their opinion is apparently the most well supported and well researched opinion in footy to date. It would do them some good to come out of the dark, reveal their methods, exact some statistical values and remove the apparently from the previous sentence.


You can claim that it is more consistent in its approach although this doesn't inhibit it from also being consistently flawed. As with much opinion there is also subjectivity involved e.g in determining effectiveness, in determining who receives credit for a clearance in a chain of play etc and in interpretation variances associated across the many statisticians involved.


We're talking about values, Doc. And in order to form a value, one must enforce a scientific approach. Inconsistency will be the biggest flaw in any scientific approach, and it will flaw the whole experiment before you even get the chance to have flawed data. That's a really wanky way to argue my point, I know, but I feel it's appropriate:p. In terms of the subjectivity involved you are absolutely right, there is plenty still associated with collecting the stats, but I'm sure that Champion Data would run their collectors through rigorous training to ensure that interpretation becomes a very dependent variable. Perhaps there will be one miscredited clearance or effective kick per match, but in the scheme of the system, in most curcumstances it will have a miniscule impact, and that's the advantage of judging plays individually also.

In the prospectus a couple of years ago they talked about the Markov model, I'm not sure if anyone's familiar with it, but it worked off the probabilities of one play moving to another, and of the ball moving from one chunk of the ground to another. With these probabilities they could run random match simulations and predict results. The markov modelling predictions were put in the fantasy freako's email for a year, before sadly disappearing. Anyway, the point is that if CD could ever set up a Markov model for use with player rankings, it would prove invaluable, and take away much of the generality associated with the current system, as well as accurately crediting negators (whether that be pitting them closer to common opinion or to the current stats model).