1. SPT brings you the latest in amateur sports tracking device. Check your speed, distance and heart rate through games.

    Use the coupon code “BIGFOOTY” for a $15.00 discount - while stock lasts

  2. Congratulations to referee and wpotter for tipping 8/9 and 11 points off the tie-breaker margin. FREE Platinum BFSC for you!

    Get your tips in!

Moneyball

Discussion in 'Football Statistics' started by Knightmare, Jan 2, 2012.

Put it out there
  1. Knightmare

    Knightmare BFSC Platinum

    Collingwood
    Other teams:
    Chicago Bulls
    Joined:
    Sep 10
    Posts:
    6,376
    Just a couple of questions for those stats nuts out there.

    Doing some statistical research on which areas of statistical dominance best influence winning and looking at areas which perhaps haven't been appreciated as much or considered as much as they possibly should in recruiting decisions by clubs - using similar ideas to the book/film moneyball, but applying those theories to AFL as best I can and using my own interpretation first starting with team concepts, and if I get successful trends then perhaps I might look at specific undervalued players and potential draft prospects.

    Does anyone have any data on total long kicks and total short kicks?
    Also would have an interest in collecting data on kicking efficiency by team on long kicks and kicking efficiency by team of short kicks.

    Some other stats by team I would also have interest in acquiring include: meters gained (through running with the footy), total tackles inside 50, total hitouts to advantage + % to advantage, marks off opposition kicks, total spoils.

    The other big thing, while harder to calculate, total production of direct opponent. So disposals, goals, marks, basically all stats against that player directly. With the trend of players not going head to head and not sticking with the same matchups due to rotations this might be one of the more difficult ones to correctly record, but incredibly interesting if someone did indeed record this.


    Have been using this as my primary resource (hopefully accurate):
    http://www.afl.com.au/stats/tabid/73/default.aspx#page=team


    Any resources or numbers anyone has available to them in any of these categories would be very much appreciated.

    I'm very willing to discuss my findings with those interested if I come accross any strong data and more than happy to discuss theories behind the numbers and why they seem to have particular effects on winning.

    (Log in to remove this ad.)

  2. Supermercado

    Supermercado Club Legend

    Melbourne
    Other teams:
    AFC Wimbledon, Falcons, Braves
    Joined:
    Dec 02
    Posts:
    8,040
    Location:
    mfcdemonblog.blogspot.com
    Not to hijack your thread but I reckon this would be a good place to post stats that you'd like to see or have come up with.

    I've had this one floating around for a few years to measure who the most effective attacking player on a team is but in the grand spirit of baseball stats like VORP which take NASA computers to work out I've not got the resources/dedication to actually apply it.

    Attacking Efficiency is a number that ranks how useful each player on the team has been to his side's overall scoring. It's not meant to be applied to defenders/ruckmen so they will understandably probably score very little.

    Add one point each for -
    Goals (6) + Goal Assists (4) + Behinds (1) + Marks inside 50 (2) + effective inside 50's (2) + Effective disposals forward of centre (1) + tackles inside forward 50 (1)

    minus

    Ineffective disposals (1), Clangers/OOF (-2) and frees/50m penalties (-2)

    The minus points come off anywhere around the ground because if you're a forward getting the ball in the midfield and stuffing it up you're negatively influencing your team's attacking potential.

    I thought of changing the points for a goal based on distance/how it was won (should goals from free kicks count less in this measure?) but trying to work it all out gave me a stroke.

    Naturally if somebody kicks ten you probably don't need this stat to tell you they were the best attacking player in their team but having spent five years watching a side who have been utter filth up forward I'd like to use a stat like this to compare a Jurrah vs Green or a Petterd vs Dunn to see who is more likely to make an impact.
  3. FIGJAM

    FIGJAM FigBooty Legend

    Collingwood
    Other teams:
    VFL Magpies
    Joined:
    Feb 01
    Posts:
    18,812
    Location:
    Valhalla
    Certainly difficult to gauge match ups from a third party data collection source, but shouldn't be so bad from within the coaches box.

    Some simple stats like a time adjusted ratio showing scores for or against when on or off the field (or not playing).

    Leigh Brown of late 2009 and 2010 for us is probably a good example, as he could literally do sweet **** all, yet have a positive influence on the scoreboard (I am saying this by gut feel, as there is no stat for it!). He was a shocking relief ruckman and a sporadic forward (at best). Yet was vital to the team.

    Defence would be one area that I'd be keen to see it in, as I reckon there'd be a few unsung heroes getting some attention for generally keeping scores down just by being there. I reckon there might be a couple of "guns" who score cred for attacking, as opposed to the defensive function. Conversely, there are some rough headed butchers who keep the scores low, even if they don't superficially do that much.

    As they say in Moneyball, Q. "Do we care if it's a walk or a hit?", A. "No we do not.". Same in footy, it's about scores in and scores out, and playing ugly doesn't always signify poor performance.
  4. Knightmare

    Knightmare BFSC Platinum

    Collingwood
    Other teams:
    Chicago Bulls
    Joined:
    Sep 10
    Posts:
    6,376
    Any help is good help.

    Feel free to come up with more ideas - stats you'd like to see made available for public access or other collected stats that might not be available in other places that might have some influence on winning.

    Also any stats anyone believes has any correlation with winning/losing, feel free to post it here.

    Would be interested to know how you came up with this points system for the attack efficiency. What are your reasons for those exact scores you associate with particular stats?
    Did you get this formula from someone or some other resource or just from your observations?

    One factor with kicking, I'm finding more kicks as having a very high influence on winning. Obviously it has to be considered that the team that is stronger and wins more of it's own ball will get more opportunities to kick more often, but it would be very interesting if say long kicks in particular have a stronger effect on winning than a higher number of shorter kicks.
    I just don't have those team stats on long and short kicks to be able to put in that data at this stage though.

    With kicking for goal I guess you somehow have to apply accuracy, I don't have any data of how a 30m kick for example differs from a 40m kick in terms of accuracy.
    My feeling is that the more kicks and more opportunities closer to goal the better, and for this to occur more often you need quick ball movement to catch out opposing defenses to allow you forwards these opportunities within 20m from goal.
    Collingwood in the 1st half of 2011 did this exceptionally well through very quick ball movement, but then in the 2nd half of the season seemed from the naked eye to completely drop off in this area.
    But acquiring any of this data would be very interesting.

    My theory with the stats against is that it is half the game and yet there are no stats available as far as I can tell.

    A +/- statistic (as in basketball) for each player would be very interesting where it calculates how many points scored for and against happen when that particular player is on field. It would probably be very volatile due to momentum swings in games, but over the course of the season potentially could even out, but with such limited interchanges and so many players on the field I do question how much insight this could give us into the influence of one individual because teams are just so large and individual influence isn't as great as in other sports. But if a clear trend came up, then perhaps in some individual cases it might quickly become relevant.

    I think around the league stats against might be one of the unappreciated and possibly unevaluated, or at least not valued as highly as it perhaps could be if AFL clubs do have this data on hand.

    Those who get tagged - perhaps a Pendlebury or Swan generally don't have that many touches against unless they have a serious head to head against another elite such as an Ablett/Judd/Selwood among others so while their offensive influence can be huge, I don't think defensively they would be huge liabilities because opposition teams dedicate so much effort defensively just to contain their offensive influence. So I think through the midfield this might be less important than contested ball winning which should and rightly is the big thing clubs focus on.

    But for defenders in particular goals against, marks against among other stats could have a huge influence in gauging their efficiency. But there just aren't the availability of stats to look into this.
    I get the impression Toovey would be one who would rate incredibly high and might around the league be undervalued because he is considered a "role player" rather than a "high level player who by position is in the top few". - But this is just my interpretation and I don't have that depth of stats to back this up.
    The other stat while I haven't tested it I think as a defender marks off opposition kicks and marks 1v1 as a defender I consider to be huge - which is a big reason why I had Ben Reid in my bests nearly every round this year.

    Even as forwards stats against could be important in deciding whether in particular situations they need to pay more attention to a defenders run for example among other factors.
  5. Supermercado

    Supermercado Club Legend

    Melbourne
    Other teams:
    AFC Wimbledon, Falcons, Braves
    Joined:
    Dec 02
    Posts:
    8,040
    Location:
    mfcdemonblog.blogspot.com
    To be fair I did just come up with the numbers off the top of my head to try and reward both goalkickers, people who set them up and forward pressure players. Goals might actually be a bit high, and you're quite right that accuracy/degree of difficulty should probably come into it too but unfortunately the position we're in compared to baseball is that the location of the ball and the amount of people who are surrounding it is so variable that it's almost impossible to compare one play to another.

    Even in baseball it's hard to do any proper measures on fielding because the fall of the ball is so variable. Even when they do it becomes ridiculously complicated (see for instance http://www.fangraphs.com/blogs/index.php/the-fangraphs-uzr-primer/)

    If I was to get into half points and make it even more complicated (but again, compared to VORP or UZR it's a walk in the park) I'd probably reduce the score for kicking a behind to 0.5 because realistically by missing a shot on goal from any position you've provided the other team a low cost - compared to a goal - chance to take the ball out of your attacking zone (unless you're playing a team that can't kick out to save themselves like us).

    Perhaps if you got even more in-depth (and end up with an article as long as the one above explaining UZR) you could weight the score given for goals and behinds based on the difficulty of the shot, the location it came from etc..

    Either way this might not be the best measure but there needs to be a way to compare forwards (and on another measure defenders) to determine who contributed more to their team. When you've got two players who are similarly matched is it better to have the one who contributes 2.1 or the one who adds nothing on the scoreboard themselves but gives their teammates the best chance to add a score.

    Accepted wisdom says the player with the bigger contribution on the scoreboard offered more but with goal assists and forward pressure stats having come into fashion over the last few years it would be good to mix them all into one stat.
  6. Knightmare

    Knightmare BFSC Platinum

    Collingwood
    Other teams:
    Chicago Bulls
    Joined:
    Sep 10
    Posts:
    6,376
    This is all stuff that would really need to be number crunched to get any level of accuracy.

    An overall offensive efficiency level would be very interesting, but to get something conclusive and accurate would be very difficult. Certainly not something I have the skill to do.

    The other criteria I'd probably include from what you came up with, you had goal assists but score assists would be another. So if goal assists are 3, then maybe score assist could be 0.5, the goal and behind measurements are probably fine.

    The things I probably value more highly are the marks inside 50 (4), tackles inside 50 (2).

    But all this is so open to interpretation and just guesswork without crunching the numbers in detail.

    Maybe other stats need to be included in this such as clearances, contested ball, disposal efficiency and contested marking. But it depends on whether you want this to be purely for fwds or to include mids also in some type of measure of offensive efficiency.
  7. SJ

    SJ Moderator

    Geelong
    Other teams:
    Joined:
    May 06
    Posts:
    28,042
    Location:
    Melbourne
    I think Champion Data already do a lot of this, it's just the general public does not have access to it.
  8. Knightmare

    Knightmare BFSC Platinum

    Collingwood
    Other teams:
    Chicago Bulls
    Joined:
    Sep 10
    Posts:
    6,376
    Sounds like Champion Data might be the ones to talk to get access to some of these less public stats.
  9. morell

    morell Microsoft Excel FTW!

    Port Adelaide
    Other teams:
    All Blacks, MPFC
    Joined:
    Apr 06
    Posts:
    7,712
    There is one man who could do this.

    RTB.
  10. Knightmare

    Knightmare BFSC Platinum

    Collingwood
    Other teams:
    Chicago Bulls
    Joined:
    Sep 10
    Posts:
    6,376
    RTB? Can't say I recognise the username.
  11. morell

    morell Microsoft Excel FTW!

    Port Adelaide
    Other teams:
    All Blacks, MPFC
    Joined:
    Apr 06
    Posts:
    7,712
    Ron the Bear

    (Log in to remove this ad.)

  12. Belnakor

    Belnakor Club Legend

    Fremantle
    Other teams:
    Joined:
    Apr 05
    Posts:
    11,565
    Location:
    Perth
    its actually a very similiar scenario to how Sabremetrics was invented - the league (MLB) was refusing to give proper stats to the public. Champion data is the same, the public just get tiny tid bits. Proper stats would revolutionise AFL betting for example, which is probably partially why they don't release them.

    Stats AFL needs
    - % of marking contests won, spoils etc a much better measure than the terrible measures we currently have
    - effective range of players (ie: how they get their possessions, where, time between possessions) GPS of players across the board would give you awesome stats, but simply knowing where on the ground would mean alot, and be alot less crude than R50 and I50
    - the difficulty kicking measure they gave us last year at one point
    - a better way to measure marks than simply contested/uncontested. Ie: marks on lead vs pack mark vs truly uncontested marks vs pressured marks

    The meters gained stats we get tid bits of would be good if you could get across all players, and it was a little more granular, ie: meters gained on first possession, 2nd possession of the chain in order to find playmakers (I think prostats has something similiar)

    I think meters gained should be one of the most important stats in the game, but it is rarely used.
    ie: score would effectively be a set of meters gained, meters lost would be things like skill errors. Theres also not a proper measure of how much a "clanger" or skill error costs a team, which you could also measure in meters lost by the team.

    Score would be something like Score = meters gained - skill errors - 50% kick to contest (you'd need a better measure of contests to work out the whole team %).

    One thing for sure is that there is absolutely zero proof that the "official" champion data is any better or worse than pro stats.
  13. Supermercado

    Supermercado Club Legend

    Melbourne
    Other teams:
    AFC Wimbledon, Falcons, Braves
    Joined:
    Dec 02
    Posts:
    8,040
    Location:
    mfcdemonblog.blogspot.com
    I want them to rethink inside 50's and only tell me when it was a quality I50.

    There's no point commentators waffling on about how a team had 40 inside 50's when 20 of them either rolled to the 49m line on the boundary and OOB or were thumped aimlessly to a nest of opposition players standing on their own 30m out from goal.

    TV/Radio need to get over their obsession with the flat I50 numbers and either start recording/buy from Champion Data the stat which shows how many balls inside 50 gave their team a legitimate chance at adding to their score.
  14. Knightmare

    Knightmare BFSC Platinum

    Collingwood
    Other teams:
    Chicago Bulls
    Joined:
    Sep 10
    Posts:
    6,376
    Some good calls and I have to agree the situation is allot like that in pro baseball with the lack of available stats.

    As a fan I crave anything I can get. The more detail and the more in depth the better.

    There are some incredibly smart people who aren't club insiders, and if the public had access to many of these resources many of us would come up with entirely new research and innovations of our own as to how the game should be played and played most efficiently.
  15. Footynomics

    Footynomics Draftee

    Essendon
    Other teams:
    Joined:
    Feb 12
    Posts:
    10
    Just posted in another thread before I read this and you've said almost the exact same thing as me. Information can only be a good thing and trying to keep your information a secret except for those that are willing to pay top dollar is a sure fire way to start an opposition community who will recreate exactly what you've done, except better and for free. Look at piracy of music and movies for an example.

    As for team and therefore player evaluation, we really need to know if footy is by and large a territory based game like the NFL and presumably the rugby's and soccer or is it an efficiency game like basketball and are the numbers consistent, thereby implying skill or inconsistent, thereby implying luck and are not repeatable.

    Adjusted +/- as it is used by some NBA teams may have some merit especially in this era of high rotation in the AFL although it can be wildly inconsistent in the NBA and an efficiency based approach in that league is often more useful in assigning player value.
  16. 93567

    93567 Amateur

    Fremantle
    Other teams:
    Joined:
    Jul 10
    Posts:
    197
    Location:
    Melbourne
    My instinct would be that a reliable +/- for most AFL players would require many seasons of data, and hence be irrelevant for an individual player by the time it's statistically significant. I do not imagine that many players, once all other major confounding variables are accounted for, have a sufficient influence on the score to give a reliable +/- without a mammoth of data. The NBA is easy, they have 82 games/season and 80+ scoring shots per game. Footy is not sufficiently granular in terms of scoring, has too many players and too few games. Not enough data, not reliable enough to work off a small amount.
  17. Rourke

    Rourke Watching the Numbers

    St Kilda
    Other teams:
    Joined:
    Mar 06
    Posts:
    284
    Location:
    Melbourne
    Here's the other thing: when Champion Data did the work on a +/- several years ago, the recognised champions of the game were virtually all in the negative (although Jimmy Bartel was top five two years in a row). One theory was that they were being rested late in games as the score blew out, but that was never satisfactory. Yep, too little data
  18. Fudge

    Fudge Team Captain

    Richmond
    Other teams:
    Joined:
    May 07
    Posts:
    22
    Location:
    Melb
    It's very difficult for Champion to open all the numbers up for everything that's of interest because it costs so much to get the numbers in the first place. These costs have to be recovered somehow and that's where the commercial side of things come in to play. There's plenty of stuff happening though as the demands on new interpretations from the information gathered keep pushing for more insights into the game. Yep, I'm from Champion so I get to see what happens behind the scene. Mainly, it's the demand from clubs that drive the agenda, but the media and everyone else also feature highly in what we do, though it may not appear that way, sorry. I'm currently reading Moneyball and it's interesting to see the evolution of how the A's paid attention to what they thought important. And that's exactly what each of the clubs do too, though they each have a different idea on what matters.
  19. joe444

    joe444 Team Captain

    Geelong
    Other teams:
    Joined:
    Jul 06
    Posts:
    1,494
    Location:
    Franga
    Might be repeating myself, but pro-stats do provide plenty of interesting analysis.

    Anyone interested in the Moneyball side of things really should read an article called The No Stats All-Star about "a player ... widely regarded inside the N.B.A. as, at best, a replaceable cog in a machine driven by superstars. And yet every team he has ever played on has acquired some magical ability to win".

    http://www.nytimes.com/2009/02/15/magazine/15Battier-t.html?pagewanted=all
  20. Slattery_20

    Slattery_20 Premiership Player

    Essendon
    Other teams:
    Joined:
    Aug 06
    Posts:
    23,388
    Location:
    My house
    Too true. I remember watching one Essendon game and we handballed/ran it into then back out of the 50 about 4 times in the one passage of play, before kicking... padding the stats nicely!!!

    I know there is an 'inside 30m' that clubs have access to. If you can't make at least a marking/crumb opportunity from that close, then your forward line isn't working.

    I've always thought there should be a stat about how often you gain/lose possession for your side.

    Leave out the bit about kicking to contests etc.

    If it's a genuine 50/50 contest with no input from a player before the contest - ie a ruck, and you win it for your side, +0.5 to you, and -0.5 to the opponent.

    If the opposition have the ball, then you have it, it's +1 to you.
    If you have the ball, and kick it and the next person to have clean possession is the opposition, then it's -1 to you.

    And so on.
    All the players' scores in one game should add up to 0.
  21. Footynomics

    Footynomics Draftee

    Essendon
    Other teams:
    Joined:
    Feb 12
    Posts:
    10
    Like the idea. 50/50 contests are tough to assign negatives to though, how do you know who the direct opponent is? A regression analysis might come up with the values of a win at a stoppage and the +/- of gaining/giving away a turnover. Kicking or handballing out of bounds or to a pack that causes a stoppage is also a negative. You had control of the ball and now it's back to 50/50.

    I feel like this type of efficiency analysis is going to be the next frontier in player evaluation.
  22. SJ

    SJ Moderator

    Geelong
    Other teams:
    Joined:
    May 06
    Posts:
    28,042
    Location:
    Melbourne
    Champion Data does all this already.
  23. Footynomics

    Footynomics Draftee

    Essendon
    Other teams:
    Joined:
    Feb 12
    Posts:
    10
    Wonderful, but it'd be even better if the general public had access to it.

    As others on here have said, Champion Data statistics are useless for fans because we don't have access to all the information only snippets of it. As Champion Data have said, they need to charge a premium as the cost of obtaining the data is so expensive.

    Saying I know something you don't know is less than helpful in this instance.
  24. SJ

    SJ Moderator

    Geelong
    Other teams:
    Joined:
    May 06
    Posts:
    28,042
    Location:
    Melbourne
    But what is the point about talking about statistics that:

    a) are already collected yet sold at a price, or
    b) you could collect yourself but would have to pay staff to do so?
  25. Belnakor

    Belnakor Club Legend

    Fremantle
    Other teams:
    Joined:
    Apr 05
    Posts:
    11,565
    Location:
    Perth
    the point is that anyone can pay staff to collect information - how do we know how accurate it is?

    Champion data's competitive advantage should be that they are better, faster and more accurate. We, as the general public have no idea if that is true because they are hiding the information from us.

    You make a motherhood statement claiming they do collect this information - would nice to know what they actually *do* collect.

    Example 1: Taps to advantage - stat alot of people would like to have but champion data spoonfeeds to us
    Example 2: I30 - a stat most people are probably not aware of, but would actually add significantly to the experience if we could get it.

    the major issue is stats are being collected which would be informative and interesting for the public but the AFL chooses to allow a monopoly that hides this information. and that sucks.