Each AFL team's midfield rankings

DC - I think inside vs outside player is a distinction that's worthy of consideration. I have supposed that you would do this by looking at what percentage of their possessions are contested vs uncontested - but even then if you say a player with 50% contested possession and over is an 'inside player' - do you call a guy who gets 48% an 'outside player' - seems to me that you would get into the same tricky territory for classification no matter where you draw the line. For me it seems just as effective to make the distinction for yourself after the fact i.e. if I rate two players both worth 25 and you go 'yeh but player A works on the inside and player B runs out on the wing' then you could just determine for yourself which one you'd prefer based on your teams needs. I don't know if I've given you a satisfactory answer there but get back to me if you want to extrapolate

Well if you think that is an interesting exercise you go right ahead and do it mate. I put it to you that Brownlow voting and peer assessment are not really truly objective measures of potential output, those are more like measures of what everyone else thinks subjectively. But look I really think we are having the same argument and not making much progress. When I referred to 'subtle unquantifiable things that make players better than the stats' - I wasn't trying to be glib or condescending. The somewhat unquantifiable aspects of the game may well me meaningful. Your example of Martin sharking one and kicking a goal from a long way out might have some intangible effect on the game - for example maybe he has to be defended differently knowing that he's capable of that, so that in turn opposition reacts differently when he's in the middle and this impacts in a way that doesn't show up or is hard to measure with stats. I have explained to you the way I come to my conclusions - if you think that's flawed or flat out wrong then draw your own conclusions. As I say I've tried to be transparent - I have a certain logic which is based on my mathematical evidence which I believe is reliable. I don't know where else to take this with you, it just seems a bit like my results don't mesh with the way you evaluate players and so no matter how I defend the premise you will still conclude there are things I don't take into consideration. In particular you seem pretty unimpressed by tackles as a stat, which is fine but if I tell you that I completed a regression analysis and there was a somewhat significant correlation between teams that tackle and teams that score or don't score against will you say 'oh I didn't know that maybe tackling is somewhat important' or will just go 'meh, correlation isn't causality - that's just a coincidence' ? You're entitled to your opinion mate - if you think the eye test plays better than my stats then that's for you to decide.

Cheers - thanks for the feedback. The total RPV is in fact a combination of offense and defense - I only didn't include it because I didn't have space so the fact that those mid grouping don't correlate necessarily with high scoring teams is partly because it's not an entirely offensive stat. I have tried in the past to incorporate a distinction for long, short and backward kicks but I found that it wasn't statistically significant to make the distinction. As in long kicking can short kicking sides one wasn't necessarily better than the other on average. Even disposal efficiency had lower correlation with offense and defense than simply having more of the ball in the aggregate. So I don't know, draw from that what you will :-)

1. That tells me Mitch Robinson might just be better than a lot of people think - don't' think there's a problem there necessarily. 2. We didn't establish that at all. You extrapolate that instance or that specific comparison as being indicative. It isn't. These clean goal kickers vs fumbly ball accumulators don't really exist in the way you're imagining. The point I made back to you was how often does Martin or any 'clean goal kicker' shark a tap and boot A 60m goal? How often does a midfielder literally carry the ball one end to the other? You make it sound like there's a Mick McGuane goal every match. This discrepancy simply doesn't occur often enough to impact the data heavily. 3. You say Freo's midfield is top 3 no matter what then immediately qualify that by saying even you could get a clearance with Sandilands as your ruck. Which is it? Are they actually good or only good because of Sandilands? Clearances are important - you are right some mids get more opportunity than others - but that is accounted for in the weighting a good mid in a low clearance team will do other things and if they do those things in good measure their RPV will reflect that. Look it's not flawless I mentioned the correlation between my RPV and scores for or not getting scored against is strong but not perfect. Sometimes players will defy my expectation but usually they don't. You're obviously not a believer which is ok. I understand you believe there are subtle unquantifiable things that make players better than the stats say. Sometimes that's true but my belief is that usually the stats, and my RPV are a good reflection of where value lies.

Incidentally - although I'm probably just talking to myself at this point - it does seem that the people who have issue with this ranking have issue because their team isn't as high as they think it should be. This seems to be spurious reasoning to me. I think I have been pretty thorough in explaining myself on these issues. You can disagree that my quantitative process is accurate, that is fine and hard to defend because if I go through a math maze to come to an answer and someone else goes 'I just reckon you're wrong' then it's hard to break down that position. Nevertheless I have established that there are certain criteria that impact my decision making - knowing that you can make adjustments however you feel necessary - if you think your good ruckman will make your rating better, throw him in there. if you think someone will play more this year than list year, inflate their numbers. If you think someone will change position, give him an extra 10 touches. If you think 20 touches from player A just 'feels' better than 25 touches from player B, go ahead and rate that better than I did. At the very least I have tried to be transparent and thorough in my explanation. BUT it is interesting that fans of the teams ranked high on this list didn't have a problem with anything but the fans of teams ranked lower than they expected are questioning the methodology. MAYBE and it's a big maybe, those 9 players just aren't as good relative to the other teams 9 players as you think? Maybe... ?

In theory what you're saying is correct. Statistically the mangled play you describe would register more value that the crisp clean play you give in comparison. My first thought is to say go ahead and record the number of times Martin sharks the opposition tap runs out of the middle and goals from 60m. I mean I know he's capable of it but if he's doing that even once a game I'll concede and say he's the best player in the league but my metric doesn't show it. But I think we both know that specific play does not occur THAT often. But I guess that's not the point exactly.... Tackling is a component of performance that I believe I have weighted appropriately. Not the only component as you point out but as I believe I have weighted it correctly I think it should balance out. So yes teams like Hawthorn and Richmond don't necessarily tackle lots but they are in possession of the ball. Incidentally my regression analysis suggests that total possession is actually a better indicator of defence than offence (not by much but still) which would fit with your understanding that teams who possess the ball don't get scored against much. You are right to say that tactical considerations are valid. Some teams - like your Tigers - deliberately play a low tackle type of possession game with a zone defence. Sydney by contrast play pretty much the opposite of this. Is one better than the other? I don't think so necessarily but I think you can value tackles relative to possession. This way you can go 'if we are a low tackle team how much do we need to out-possess the opposition if we know we are going to lose the tackle count' - So if the Tiges get 'out tackled' and break even on possession this is probably not a good sign. By quantifying the value of a tackle we can have an idea of where the parameters are for those things depending on our game style. So to your point I don't think good defensive teams have to be high tackling teams - which we agree on. But I wouldn't dismiss tackles as an irrelevant by product of game style - they contribute to scoring or not getting scored against in some measure it's just a question of how valuable are they relative to other statistical outputs.

I can talk to you about how I've developed the predictive models - it's not that complicated but it will take up a lot of comment space as brevity is not my strong suit - it has nothing to do my personal views.

Cheers - so you say that you like quant modelling because it provides objectives results and eliminates inherent biases but then when the results don't mesh with your subjective understanding you dismiss the modelling as influenced by the developers (my) personal views? I'm confused - only results that match with your initial beliefs are correct? Otherwise I must just hate the Hawks? You're welcome to look at the data for those players - obviously there is weighting for each statistical category that I have not revealed that might skew things in certain ways that aren't clear. But go compare Isaac Smith and Lewis Taylor. Statistically not that different. And Taylor is trending up in a lot of categories while Smith is more likely to plateau. It might be worth noting that Smith is pretty much the definition of 'outside runner' and while he plays his role at Hawthorn and has carved out a niche he might not be as successful at another club. Then there's the notion that what Smith does (runs, carries, kicks long) is more valuable than the things Taylor does (handball more, sometimes plays closer to goal as a small forward). Well when I did my regression analysis there was no significant correlation between teams with the most long kicks and scoring. Which is to say I don't know that long kicking is inherently preferable. But it is difficult to overcome subjective assessments - you think Isaac Smith is better than Lewis Taylor, he's certainly been more successful and has more highlight reel plays. Does it make him better? I don't think so but despite having a quantitative methodology for coming to this conclusion because your 'eye test' says different you conclude the methodology is wrong. I think what is more likely is that the sample of 9 mids and the criteria I've used limits in such a way that the midfield ranking as I've put it together doesn't necessarily reflect the entire teams quality. For what it's worth I think Hawthorn this season have the 5th best first choice 22 in the comp. This is partly due to modelling that suggests some of those midfielders are on the wrong side of their peak form. but on the other hand many of their most valuable players are distributed in other parts of the ground (Gibson, Gunston Burgoyne, Rioli, Roughead) - in some cases although Champion Data designates them as general forwards or defenders their influence really occurs all over the ground. My point being only that I have used a criteria to select a midfield and that does not necessarily indicate how good the entire team is. You can go through and find player comparisons that don't mesh with your reading of the game Smith v Taylor for example but this hardly scotches the entire process. You would need to examine the underlying methodology more closely than some random player comparisons that you don't think are right to shake me of my confidence in my process or 'send me back to the drawing board' as it were. I can assure you I have no personal views and rely only on the numbers - the fact that you conclude I must be imposing personal bias to get my results is actually just evidence that you start from a non negotiable premise (those 9 Hawthorn players are better than x) and when results differ conclude that the methodology is wrong because it doesn't fit your pre-existing bias.