Consistency in AFL Dream Team

Consistency is a big part of AFL Dream Team. All of us know the pain of winning an eliminator by 10 points because player A gets 140, then losing by 30 when he gets 70 next round. That’s why when you pick your teams; consistency and ceiling combined have to come into it. The team that wins the competition at the end of the year is the team that, week after week, pumps out 2200+.

However, to the casual DTer, players that are consistent are quite hard to pick out. Most of the time they go with average and average doesn’t tell all. Consider Player B gets 140 then 70 every alternating week for 10 weeks. Then consider Player C getting 90-110 every week for 10 weeks. They both have the same average. From a casual glance they both look just as good. However, a more detailed analysis tells you a different story. You need another couple of measures of data to see which is better.

That’s why I decided to put together a couple of tables (one for each of the positions) to show you (after round 9) which are the players that are the most consistent. I am going to use 5 key measures of data. Before I launch into the tables, I will explain what each of those measures of data means.

I will not however explain how they are calculated, because while some of them are quite easy, a few of them are a little tricky. I got the values in the tables from a graphics calculator, but you can look up how to do them on the internet to get values from other players of interest I have not included.

  • The Mean-The mean is a fancy word for the average. The average is basically the scores added up divided by the number of scores. This takes into account every score the player has received.
  • The Affected Mean-This is a mean that takes into account only the games where the player in question is not subbed out early or injured. I believe this one to be the more accurate one where these players are concerned because it is not often where premiums get subbed out of games, and therefore does not reflect truly on their consistency and scoring potential. I do not omit scores from games where the player was just bad (Hartlett and Birchall I’m looking at you!)
  • The Standard Deviation (THE IMPORTANT ONE)-Standard deviation is a measure of how far the data is spread out. Lower values are better, i.e. the lower the value, the more consistent the player. This is the set of values I rank the players on.
  • The Range-The difference between the highest and lowest scores, i.e. Max take away Min. Lower is better.
  • The IQR-This is short for the Interquartile Range. This value is given to you in the form A-B (the dash does not mean minus, it means ‘in between’, i.e. between A and B). This gives you the range the player scores 50% of his scores in. For example, say Player D has an IQR of 100-130. This means that player D scores between 100 and 130 fifty percent of the time he plays. The lower the number between the values I give you, the better. This means you want the values to be as close together as possible.

 

General Info

Okay! Now we have got that sorted out, we can move into the tables! Remember, they are ranked by standard deviation not average, and the green numbers on the left side are their places if you were placing them by average. They are there to contrast between the highest averaging players and the players with the lowest standard deviation. The underlined players are the players that have a different affected mean to their actual mean. Lastly, I will comment on a couple of players of interest in each category.

 

Defenders

cons_def

The average standard deviation (SD) for the defenders is 23.79, with the top 25% having a value below 17.63. A value above 23.79 is not ideal; it means they are quite erratic in their scoring. One notable exception to this however is Brandon Ellis. Even though he has a SD of 29.58, it comes about because he had scores of 86 and 155, which are fairly spread apart. This is why I have included the affected mean and normal mean. With Ellis you can see he has an affected mean of 113.5. If you combined this with his SD, you could come to the same conclusion as me. (*Note: the scores of 86 and 155 come from memory. There is not a way (that I am aware of) to get those values from the table)

You can see the inconsistent Hartlett and Birchall are the bottom ranked players on the graph. They both have massive SDs, as well as big ranges and IQRs. This is also reflected in their means, with both of them having averages far below the top ranked Goddard and Hanley.

The three most consistent players in defence it seems are Hibberd, Goddard and Enright, in that order. They all have extremely low SDs and IQRs, particularly Goddard.

 

Ruckmen

cons_ruc

The average standard deviation for the ruckmen was 18.75, with anything below 14.69 being in the top 25%. Although Matthew Leuenberger is out for 4-6 weeks with his sore thumb, I have included him and his stats to show you just how consistent that man is. He has a SD of 12.2. I’ll repeat that again. 12.2! That is incredible for a group of fantasy scores. To remind you again, that means most of his scores are clustered relatively closely around the mean. His IQR of between 68-90 means that 50% of his scores come between these two values.

Minson and Cox are the next two most consistent ruckmen, with Roughead and McEvoy rounding out the rear. For consistency’s sake, the stats are telling you to stay a fair bit away from these two, but in real life you can cope with a SD of around 24. It just means you’ll get some scores around 55-60 and some around 100-105.

 

Forwards

cons_fwd

The average SD for the forwards is 20.5. Anything below 17.5 is an excellent score. There are three players in the above table that have less than 17.5. Those are J. Kennedy, J. Lewis and M. Stokes. All of those players are excellent recruits, with their scoring falling between around 80 and 100 50% of the time.

Kennedy surprised me a little bit; I thought it would be someone like Stevie J who’s just gone BANG BANG BANG with 5 consecutive scores above 100. However, Stevie’s first score was 82 and his highest was 144. This means his range is fairly high along with his SD. It is funny that in Martin and Riewoldt, you have the second and highest DT scorers, but they are the second and third worst for consistency. Cloke rounds out the rear, with his SD of 28.83 being horrible. Stats say stay away despite his average of 93.89.

 

Midfielders

cons_mid

Finally! The Midfielders! The average SD for them was an incredible 18.94, with 25% of them having a score below 13.91. People, there is a reason these people are beasts. However, some are still better than others. The three most consistent are Mitchell, Priddis (who’s surprised?) and Moloney. However, Pendlebury and Swan have scores in the 13’s as well.  All of these players will give you a score generally within 14 points of the mean every about 70% of the time they play. You can see Swan, Mitchell  and Pendlebury have the highest averages, so they are the players you would look to trade in.

The players around the middle (Ablett, Jack, Stanton, and Cornes) are still great players, it’s just they might have had one extremely high score, or one extremely low score. A couple of those and it affects the SD. However, the stats are telling you to steer clear of Selwood, Watson and Dangerfield if you want your sanity at the end of the year.

 

Conclusion

Whew! Okay, now that’s done, I can show you how you would use these tables when trading. If you are looking for higher averages when trading players in, then ignore this completely. However, if you are smarter and want consistency over flashy scores now and then, you would do well to study these tables. Now, for example, I am looking to trade out Leuenberger to a premium ruck this week. I would go and study the ruck table. *looks at ruck table*. Hmm, Minson seems to be quite consistent, with 50% of his scores coming between 79 and 107. He also has an average of 87.78, that’s not bad for a ruck. Okay, trade him in.

I am using my second trade to get rid of Terlich to a premium defender. If I look at the premium defender table I can see that Hibberd is the top. However, as much as I want him, I don’t want 3 Essendon defenders. It messes up my bye structure. I have got Goddard; I don’t want Enright with rumours of a rest, so Hanley it is. It is literally as simple as that (it helps Hanley is the third highest averaging defender as well). I have now got 6 premium backmen, although currently I’m not sure you can count Hartlett as a premium.

 

Okay, so now you know how to use this, you can go forth and use it. Remember, just because I haven’t included a player in this table, doesn’t mean you don’t want to trade them in, it just means they weren’t included in my list because I just picked the ones that mostly everyone has.

Cheers, and any questions, just hit me up on Twitter @ZeusODea and I will try my best to answer them, although I really should be studying for my literature exam tomorrow!

Zeus

DOWNLOAD: consistency.xlsx

102 Comments

  • Ah that’s great! Cheers Warnie, you have officially made my day!

    • Haha. No worries mate! Thank you! Get back to studying… and good luck for tomorrow!

      • Haha yeah probably should! I’ll keep this open though if anyone has any questions along with twitter. I expect there’ll be a couple about the IQR and Standard deviation… they are the most complicated ones.

        Cheers, Zeus

        • Very interesting, thanks Zeus.

          Perhaps add another column for “Affected Mean less vs Melbourne” scores? Thats likely to affect the scores of certain players significantly.

          How apt that a consistency article comes out just prior to the bye rounds. More consistant players would definately be a boon during these “best 18” rounds.

          • No probs red, glad you liked it. Yeah the players scores would significantly increase against melbourne. They are struggling!

            I thought of your second paragraph when I first came up with the idea, I thought it might be quite helpful to people! Glad everyone agrees!

    • A few people are asking about Kreuzer, here are the stats: Mean: 71.5, affected mean: 81.6, SD: 23.95, Range: 74. IQR: 61-103

    • I agree whole heartedly, excellent article! Really appeals to my inner nerd. Has given me food for thought re this weeks and fwd trading. Hopefully win get me a few more W’s in the work league.

      • haha thanks mate, I hope you win them, if you do thanks to the above research, then that is my job done. Good luck!

  • I Like it! I am such a sucker for a high ceiling, but will definitely use this next week to plan for consistency through the byes. Cheers O Zeus

  • nice list, I used this type of thing to swing me to Cotchin preseason but unfortunately he turned off his Mr Consistency this season so far, but expect him to bring back his Mr.Consistent game for the rest of the year!!

  • wow… great article warnie…

    this would be by far one the best things i have seen on this site.

    (3 year DTer)

    Well done… i cant imagine how long this took you.

    • The actual article didn’t take me long at all, it was the calculation of the stats that took the longest, i.e. about 40 players worth.

      Great to see you like it, cheers!

    • i mean Great article Zeus…

  • Well done Zeus. This is fantastic info!

  • Thank you for the article, but I can’t agree that lack of consistency is necessarily a bad thing. Assuming the averages are the same, I would want my team be sprinkled with a couple of players with high SD, like Cloke. This would allow for a solid base of scores to be potentially boosted into the extreme category sometimes, while remaining solid at other times.

    I mean players like Cloke of course, who mix 70s with 110s, not Hartlett’s who mix 30’s with more rubbish.

    • That is true mate, but look what happens when they all go the opposite end of extremely great… extremely bad, just look at (round 6) I think when they all got subpar scores and everyone struggled to pass 1800

      I mean in a perfect world you’d love players to do what you said, but when they stuff up collectively, it’s not pretty

      • Yeah, I would stick with a number of ‘deviants’ that shouldn’t hurt you too much when they score low, maybe 2. I like having one or two players like Cloke, as long as they average well.

        • Yeah that is alright, maybe having maximum 2 or 3 i.e. Dangerfield, Cloke, D. Martin, just not too many!

    • (Assuming averages are the same) if you’re going for the weekly prize, or you’ve got a relatively shit team (and need some luck to win league games) then, yes, a higher SD is better. But if you’ve got a good team and want to be consistently good then the more consistent the better because for every match you win by a couple of fluky good scores you’ll lose one from a shit score.

      • Exactly mate, and I probably should have said that in the article, but I assumed everyone on this website would probably have pretty good teams because the amount and quality of info is second to none, I’m sure you’ll all agree…

  • Great article man!. Been waiting for someone on this site to post a ‘hardcore’ stats article. I’m really into that.. bit sick in the head really. Any chance you can point me in the direction of how to calculate standard deviation?

  • Perhaps the “Affected Mean” should be renamed the “Unaffected Mean”

    • Not trying to sound like a troll here, but stevie J’s first score was 92, not 82 — if this gets included in the re-calculations, Stevie J might be closer to Stokes in SD

      • Unless it was just a typo in the article. And if you exclude Stoke’s sub affected 60, then JJK’s the only guy in the fwds (of those listed) who’s anywhere near him (See my post below).

      • Ah another one, sorry mate, rushed calculations the culprit there…

  • Great article Zeus. Just my cup of tea.

    I don’t normally bother with calculating these, but rather just a have a look over scores and try to avoid players with unacceptably low scores that little bit too frequently. Although, SD are calculated by FFGenie (http://ffgenie.com) for those who are interested. Of course, it doesn’t exclude sub affected games which makes it virtually useless for players such as Ellis.

  • Great article Zeus, good read!

    I used this kind of logic a few weeks ago when I had to get rid of Hartlett. Was tempted by Hibberd & Ellis, but thought better be reasonable and get in Scotland – someone I won’t have to worry about and just get that guaranteed 85-95 every week. Nek minnit, I get a 74 & 68… Win some, lose some I guess!

    • Haha yeah to be fair I went for Scotland as well, so he’s one of my back 6 ‘premiums’ along with hartlett and birchall so thats probably why I am ranked 13000 something. I would have got in Hibberd, because as I mentioned in the article, he is the most consistent, but it messed up my bye structure. Probably not one of my best decision, but the value of hindsight!

    • Stats are an excellent guide but must be interpreted with your DT nous.
      eg
      Scotland new coach, new role???
      Hanley’s low score of 62 when he sat out the last quarter.

      Great article Zeus. Appreciate the effort.
      Was looking at Minson or take a punt on Maric for a Berger replacement
      to suit my bye structure.
      Mincin’ Will may get the call up.

      • Yeah you should never look at stats alone. I agree with Scotland, he generally is pretty consistent, thats why I went with him two weeks ago, but he is stinking it up currently, and I’m not sure why.

        Thanks for the kind words, Minson also fits my bye structure, as he also has a round 11 bye, same as leuey. It all depends on your individual team.

  • Oh and Stokes had a sub affected 60 early in the year (Rd 2) which would make his values even better:

    Affected mean: 102.25
    Standard Deviation: 12.93
    Range: 40
    IQR: 95.5-107.5

    • That’s true Tim, I must’ve missed that one. I’ve probably missed a couple of others as well, I did it all this morning and last night, so thats probably why.

      Stoke’s stats are all good except for the SD, I think you might have used the n-1 (population) SD instead of the normal SD, but not that it matters, it only makes it 0.80 something higher.

      Thanks for pointing that out though!

      • No worries mate, easy to miss since it was still a respectable score.

        For the calc, I just excluded the 60, which made n=8. Pretty sure my SD is correct with the population SD being 12.09.

        • yes that second SD is correct, it was just the first one was a little off. It is easy to do, though I had to correct some of mine because I did the same thing! On my calculator, tehy are listed one after the other, so it is easy to get the wrong one!

  • Great article mate :)
    I’d say all good Dream Teamer’s do consider consistency when selecting players, it’s great you’ve gone to the time and effort to analyse the numbers in more detail for us !
    I usually just have a look back through all the players past score’s, so this table is a real time saver. Should make it a monthly or fortnightly article on the site I reckon !!!

    • Thannks Killaz, I appreciate the kind words. I came up with the idea while walking my dogs yesterday (I was doing a bit of DT reflection) and just ran with it. I would love to make it a monthly article if Warnie said so, but until that happens (probably not likely) its probs a one off.

      I think fortnightly might be too much though, because barring a shocker from one of them, the SD isn’t going to change that much in two rounds. The monthly one is a much better idea.

      Cheers!

    • Or a really short weekly article looking at the consistency of a couple of popular options to trade in for the week.

      No pressure by the way Zeus. Just something you could do if you want to continue to contribute to the site.

      • Yeah that is also a good one Tim, i.e. maybe the top 5 scorers in each position or something. That would be a good one! Don’t worry about the pressure Tim, I like it! It is a good way to express my creativity! I enjoy helping other people!

        • As you said that would remain fairly similar each week. I was thinking more along the lines of picking an issue that’s affecting many DTer in a particular week. So this week you might consider Leuenberger’s injury and look at potential replacement rucks. Or maybe do a comparison between Boyd and Cotchin, something like that.

          • ahhhhh I see, sorry I misunderstood your first point. Yes that is a good idea! I would love to take on the challenge!

  • luenberger for vardy?????????

    • I’m not sure mate? Vardy hasn’t played this year, so I can’t really comment on his consistency or his ceiling. He is highly rated though

    • alf.com.au dump and grab seem to think z.clarke or vardy will be a good replacement for him…

      up to you if you take the chance… personally i am benching him for the next 2 weeks, as blicavs will post sub par decent scores in his place as he is versing GWS and GC.

  • Really takes time to write a huge and important article like this one Zues! Goon on ya! What are everyone’s thought on Staker not being a crucial pick at this time, as he will not fit my bye structure, so if its bad that I don’t pick him up? One other thing, I need to get rid of Hodge or Fyfe, as they are both causing havoc for my byes, what is everyone’s thought on Montagna?

    • seems to be a top form at the moment… but i can see the tag shifting from Dal Santo to him very soon as he is becoming more damaging.

    • Cheers mate, I appreciate the support.

      I would say Staker is probably the best of the rookies this round, hutchings doesn’t have great JS, and that therefore affects his consistency (couldn’t resist-I’m sorry). Staker on the other hand has great JS and solid scoring.

      I was looking at montagna as a potential pod myself, just couldn’t justify trading out one of my playing players to him when some of my players are injured.

  • hats off to Zeus, not only writing and doing this article but also has replied to every post on here… not many writers do that…

    well done… again

    • Haha thanks mate! It’s a pleasure to help you all out, I’m glad you enjoyed it!

  • Great article mate, love my stats.

    I must question what everyone’s love of “consistency” is.
    I understand that is frustrating watching the roller coaster that is some players scores, but does it really matter?

    If you are going for overall ranking, consistency means very little.
    As long as the player scores the most points come seasons end that is all you should be concerned with.

    (For example I would much prefer player A to go 150 then 60 than Player B to go 100 and 100)

    I would much prefer to look at your “affected mean” when looking at trading in players as that is quite a useful stat.

    • In saying all that, it really is a well written article!

    • Thanks for the feedback mate. It is more the frustration when you know they are capable of getting premium scores every time they play, and yet they still pump out 60’s. It is what made Swanny a DT Pig, for years he just kept pumping out the 130’s and that is what made him a great captaincy option, because people could trust that he wasn’t going to stink one up. sure, 150’s are great, but if a player is going to get a 50 the next round, then I’d rather not, although, it is personal preference, if you don’t mind copping the odd week where you get 1800 scores.

      I’m not having a go at you btw, just explaining in general. As I said, I appreciate the feedback. I’m happy to clarify anything if anyone needs it!

    • It gives you tough choices. Say your player gets 2 60s in a row. Say he’s a 50/50 for 60/150, then he’s a 50% chance of getting a third 60. Really hard to hold onto a player like that and bleed the cash, when you can’t be sure that it’s part of his rollercoaster and not a drop in form.

      Compared to the 100 every week, which saves you headaches and sideways trades, I’ll take the 100/week player thanks (85 and 120 on the other hand, I might be able to deal with).

  • Warnie I want to trade out moloney, for Stevie j do and coming up against easy teams as well as a flexible round 12 bye. I also want to trade out leunberger for a cox or natinui which do you think is better i could trade left out next week and play blicavs

    • Mate I would be looking at getting SJ in ASAP, so whether that means playing blicavs, I would do that

  • Leunberger*

  • Yes, I made this mistake this year. 2 trades per week made me lose sight of some of the basics.

    I traded some players in, well known for their inconsistency (Dangerfield anyone) when I really should have started with players known for their consistency.

    Players like Barlow, who was week in week out getting 100 for my team last year, and a few other stewards of the game, rather than unknown quantities like Hartlett, Zorko, Wright, who all didnt perform. Im trading players in for consistency from now on if possible.

    • Yeah that’s true Shags and I think we all did a bit of that. For example, I started with Zorko (who I dumped round 4)and Riewoldt (who is the human form of a machine).

      Consistency is what separates the good teams from the bad, the boys from the men.

  • ffgf

  • It’s funny that I haven’t traded looking at byes, but now I’ve figure what trades I’m going to do during byes, I will have 19, 18, 18 players in each round.

  • This is a superb anaylsis and something I will use now and in coming seasons, especially at the start of the year.

    • cheers albert, good to know people appreciate it! I’m thinking of doing one every month or so maybe, definitely one at the end of the year because there has been quite a positive response to it…

  • 2nd year DTer.

    Guesswork has been my mainstay in the past.

    Think I need to go to an adult class and brush up on my Excel skills and be ready for next year.

    There is a lot more to this gig than I ever thought.

    Thanks, Zeus.

    • Nah mate excel doesn’t’ have much to do with it, it is just the medium I typed my stuff in. If you want to do further research, just learn how to do the 5 things I show you above, and you’ll be set! :)

    • Even saying that, if you’ve got a graphics, just chuck player x’s scores into the stats mode and it will calculate for you

  • I wanna get Stokes in but do people think he is gonna get rested?

  • should i get kruezer or minson.

    they both have easy games this week but kruezer is 100k less then minson so im leaning towards kreuzer but i have this but felling.

  • A few people are asking about Kreuzer, here are the stats: Mean: 71.5, affected mean: 81.6, SD: 23.95, Range: 74. IQR: 61-103

  • many thanks Zeus for a terrific article – very well done.

    Is it valid statistically to (say) multiply the Affected Mean by the Standard Deviation or the IQR and get a meaningful statistic? If so would it indicate risk of the player going high (or low) but tell you something about what that number may be?

    Thanks, don’t want to keep you from the literature revision!

    • No problems Chels, it was a pleasure.

      I am not sure you could do that Chels, because I don’t know whether you want a high number as a result (indicates a higher mean) or a lower number as a result (could indicate a low SD).

      Haha its ok, ive done a bit of lit and I’ll do a bit more tonight. Sorry for the late response, I had work. I am going out for dinner as well, so I will respond to the rest tonight.

      Cheers!

      • You could multiply the mean by 1/SD – then the bigger the number the better. Could be a good way to compare two players, one of whom is consistent and one who has a higher mean.

        The fact that you are celebrating Leuey suggests that you can get carried away favouring consistency – coz anyone who brings Leuey in from here on in has rocks in their head.

        Presti was consistent…

        Basement non-affected scores are are better guide IMO – if a player averages less than 50 from their 3 worst games, pass… I they average 70 from their three worst scores, then they are either consistent or they have a massive ceiling – either way, I am buying

  • Thanks Zeus, best article i have read in a very long time! Have you considered making a mock team with regards to consistency and seeing how well it would go?

    • Thanks Brodie. Ummmm no I haven’t although that would be a good idea, just to see how well that would go against the rest of the competition! Maybe I will!

      Cheers mate for the idea

  • Thanks for the article mate. Great read. Just to step it up a nerdy notch…

    You could use a stat called the coefficient of variation, which is standard deviation divided by mean, to rank the lists as well. The smaller the number the better.

    For example, in your defenders, Goddard has a smaller COV than Hibberd,as his slightly higher std deviation is more than offset by a higher mean.

    Anyways, keep up the good work.

    • many thanks General Soreness for your post. The process of division rather than multiplication means one just inverts the resultant?

    • Very valid point General Soreness. I was not aware of that one. If I do this in the future, I will definitely include that. That seems like a good measure of both mean and SD. Cheers!

    • Could you possibly even go mean divided by SD, where the bigger the value is, the better?

  • Who to play out of Ben kennedy or Kane Mitchell? both sub dangers

  • Good article Zeus, and good to see more coaches taking an interest in stats.

    For those who aren’t aware NixTrader spreadsheet calculates standard deviation for all players on your team and your watchlist, for last year as well as current year. Plus from next week also includes a skew statistic to show whether the biggest variations are more lower or higher than the mean.

    And thats a good suggestion from General Soreness, maybe I’ll see if I could add a COV as well.

    • Cheers Nix,
      I was not aware of that, but in any case I couldn’t use that because I had to calculate the IQR and range and I’m not sure the values in your spreadsheet do that.
      If they do though, let me know, so if I ever do another one I can use it.

  • Great work Zeus! Gettin my nerd on here, lovin it!

    I did wonder though whether the STD DEV as a measurement of player value would be better represented as a percentage, rather than points. For example, as Gary Ablett or Stevie J are higher point scorers their STD DEV currently puts them down the rankings a little…I guess though that’s all covered in the IQR…maybe it would be worth then attempting to come up with a formula of IQR vs STD DEV and rank these lists…let me know what you think or if I’m just talking sh*t on a Friday!

    • Hi TBone, great to hear you enjoyed it!

      Regarding what you said about SD as a percentage, I do agree with you, and if I do it again maybe later in the year, I will include another value that relates the mean and SD, because a couple of other people have mentioned that.

      For this one I thought it would be better to give you the mean, IQR and standard deviation and you can make the judgement yourself, i.e. if Ablett has a average of 120 for example, and an SD of 20, then you don’t mind copping 100 as a low score? It really depends on the player, but like I said, you aren’t the first one to mention it, so I definitely will add another value in next time.

      Thanks for the feedback!

  • Pearce Hanley was a good trade haha.. Enright will get 110 tonight

  • Pfft, traded in Sam Mitchell this week. Seems “Mr Consistant” couldnt even crack 70 against Melbourne.. Same has happened when I traded in Hanley & Birchall.. I must just be jinxing these guys. Getting close to deleting my teams really.