Discussion Tip a Top 30: Using last year's points to predict the future
- Thread starter KLo30
- Start date
Defenders
What can we take away from this for the class of 2014 in 2015?
• Four players (McVeigh, Swallow, Hanley and Bartel) have had position changes,
• Malceski is highly unlikely to maintain a 100+ average (correct)
• The remaining top 10 are no guarantees to remain in the top 10 of defenders despite losing 4 to other positions (only Shaw remained)
• Look for 2 (or more this season) players from 11-20 to make a notable improvement (Hodge, Rance, Enright)
• Look for 1 player from 21-30 to make a notable improvement (Houli)
• Hurley (FWD), Picken and Laird also made notable improvement
What can we take away from this for the class of 2014 in 2015?
• Four players (McVeigh, Swallow, Hanley and Bartel) have had position changes,
• Malceski is highly unlikely to maintain a 100+ average (correct)
• The remaining top 10 are no guarantees to remain in the top 10 of defenders despite losing 4 to other positions (only Shaw remained)
• Look for 2 (or more this season) players from 11-20 to make a notable improvement (Hodge, Rance, Enright)
• Look for 1 player from 21-30 to make a notable improvement (Houli)
• Hurley (FWD), Picken and Laird also made notable improvement
Rucks
What can we take away from this for the class of 2014 in 2015?
• There will be a new top 2 (correct – Jacobs & Mumford to Goldstein & Martin)
• One of the top 2 won’t play the full season (Mumford missed 11 & Jacobs missed 1)
• One of the top 2 will likely come from outside the top 10 rucks (incorrect – Goldstein 5th & Martin 3rd)
• Ruckman get hurt (correct)
What can we take away from this for the class of 2014 in 2015?
• There will be a new top 2 (correct – Jacobs & Mumford to Goldstein & Martin)
• One of the top 2 won’t play the full season (Mumford missed 11 & Jacobs missed 1)
• One of the top 2 will likely come from outside the top 10 rucks (incorrect – Goldstein 5th & Martin 3rd)
• Ruckman get hurt (correct)
Forwards
What can we take away from this for the class of 2014 in 2015?
• Three players (S.Martin, Parker and Dangerfield) have had position changes,
• S.Martin is highly unlikely to maintain a 100+ average (incorrect)
• The remaining top 10 are no guarantees to remain in the top 10 of forwards despite losing 3 to other positions. (only Martin & Zorko remain)
• Look for 1 or 2 players from 11-20 to make a notable improvement (Dahlhaus)
• Look for 1 player from 21-30 to make a notable improvement (Mitchell)
What can we take away from this for the class of 2014 in 2015?
• Three players (S.Martin, Parker and Dangerfield) have had position changes,
• S.Martin is highly unlikely to maintain a 100+ average (incorrect)
• The remaining top 10 are no guarantees to remain in the top 10 of forwards despite losing 3 to other positions. (only Martin & Zorko remain)
• Look for 1 or 2 players from 11-20 to make a notable improvement (Dahlhaus)
• Look for 1 player from 21-30 to make a notable improvement (Mitchell)
It's easy enough to do in Dec/Jan/Feb. Holding the conviction in March is the hard part as there is always a reason why you shouldn't pick someone being eloquently expressed by others, often with tables and graphs. 
Too many mid-pricers, don't like their role, someone else is going to take their midfield time, etc.
Too many mid-pricers, don't like their role, someone else is going to take their midfield time, etc.
It's easy enough to do in Dec/Jan/Feb. Holding the conviction in March is the hard part as there is always a reason why you shouldn't pick someone being eloquently expressed by others, often with tables and graphs. Too many mid-pricers, don't like their role, someone else is going to take their midfield time, etc.
Too many mid-pricers, don't like their role, someone else is going to take their midfield time, etc.
I see the key message of this work being that we shouldn't "just" chase last year's points and base our teams on 2015 performance - critically, because that is also what determines prices, so it's hard to find value using that approach. In working through the chasing pack, though, I am still inclined to simply look a year or two further back ... Ie, look for a fallen premium, not speculate on someone that has never been a premium before. I think the odds are much better, and it worked well for me last year, so I'll be doing it again this year ... with help from your tables. Thanks again.
- Joined
- 9 Dec 2015
- Messages
- 1,727
- Likes
- 2,336
- AFL Club
- Hawthorn
KLo - fantastic work on rebooting this thread. I had mentally earmarked it last season and worked through it again on the weekend - there is some great data and analysis on here from you and Rowsus, and some interesting discussion from others. I was going to give it a 'bump' for everyone else, but that now seems unnecessary
I see the key message of this work being that we shouldn't "just" chase last year's points and base our teams on 2015 performance - critically, because that is also what determines prices, so it's hard to find value using that approach. In working through the chasing pack, though, I am still inclined to simply look a year or two further back ... Ie, look for a fallen premium, not speculate on someone that has never been a premium before. I think the odds are much better, and it worked well for me last year, so I'll be doing it again this year ... with help from your tables. Thanks again.
I see the key message of this work being that we shouldn't "just" chase last year's points and base our teams on 2015 performance - critically, because that is also what determines prices, so it's hard to find value using that approach. In working through the chasing pack, though, I am still inclined to simply look a year or two further back ... Ie, look for a fallen premium, not speculate on someone that has never been a premium before. I think the odds are much better, and it worked well for me last year, so I'll be doing it again this year ... with help from your tables. Thanks again.
its then more likely you will get value by choosing at most one or two "can't miss" young (3rd-5th season) breakout contenders from the 2nd/3rd tier as well - but choose wisely of course
One of the stats I find interesting is the number of top 10 forwards, defenders less so, who are dual position. It makes sense that midfield time results in more disposals, contested possessions and inside 50s, and that playing as a lead up forward or small forward role becomes reliant on opportunity, and hence, is inconsistent from game to game.
Therefore in 2016 you take Montagna over Deledio, Zorko over Wingard and Barlow over Westhoff.
(NB. these player examples are simply paired on the table. No other attributes have been taken into consideration)
Forwards
- ------2015----------------2014-----------------2013------------------2012----------------2011---------------2010-----------------2009----
You can certainly see the evolution of the game from 2010/2011 to 2015. Approximately half were forward only in 2010/2011 but in 2014 (2) and 2015 (1) dual position players ruled the top 10.
Defenders
------2015----------------=2014-----------------2013------------------2012----------------2011---------------2010-----------------2009----
Therefore in 2016 you take Montagna over Deledio, Zorko over Wingard and Barlow over Westhoff.
(NB. these player examples are simply paired on the table. No other attributes have been taken into consideration)
Forwards
- ------2015----------------2014-----------------2013------------------2012----------------2011---------------2010-----------------2009----
You can certainly see the evolution of the game from 2010/2011 to 2015. Approximately half were forward only in 2010/2011 but in 2014 (2) and 2015 (1) dual position players ruled the top 10.
Defenders
------2015----------------=2014-----------------2013------------------2012----------------2011---------------2010-----------------2009----
Last edited:
^^ I was thinking the same thing. Most of the top-ranked forwards and defenders are mid-eligible, and only a small proportion of defenders and forwards have this DPP ... so the chances of getting a top-ranked fwd/def are *much* higher if you choose one that is mid DPP. Something to bear in mind.
Ruck-fwds also seem to score surprisingly poorly relative to straight rucks, albeit perhaps on a small sample.
Ruck-fwds also seem to score surprisingly poorly relative to straight rucks, albeit perhaps on a small sample.
Great reading KLO, one thing i'm taking out of that defenders table is that in any given year your top defenders will be somewhere in that 95-105 average range and anything outside of that, or even near the top, is a freakish outlier and should be avoided in your starting lineup the following year, eg Malceski in 2015 and Shaw next season.
I hear ya! Switch between "Rate my team" and position threads back to this and it can do your head in. Couple it with Mid-price Madness and the crazy really sets in.
Treloar came pretty dam close!
Others that failed to break the curse for the 2015 season includes S Jacobs, S Mumford, S Martin, Sandiland (only by 0.2), D Swallow and T Green. That's another 8 players who failed to break the curse last year.
There are another 6 players who qualifies for this double curse for the upcoming 2016 season:
H Shaw: 92.6 (2014), 112.6 (2015)
L Montagna: 89.4 (2014), 111.0 (2015)
T Mitchell: 83.8 (2014), 104.5 (2015)
M Blicavs: 69.2 (2014), 104.3 (2015)
M Gawn: 63.2 (2014), 102.1 (2015)
M Bontempelli: 78.6 (2014), 103.2 (2015)
That brings the total to 49 players who have achieved the cursed double to date, will any of the above join Swan, Ablett and Libba in breaking the curse? Apart from Shaw and Montagna, I think the others all has a 'genuine' chance of adding to those who have succeeded to breaking the curse!
The Class of 2016 does looks a promising group, could 2016 be the year for 'smashing' this double curse to pieces...only time will tell.
Note - D Swan and J Steven just misses out on qualifying again...LOL!
Others that failed to break the curse for the 2015 season includes S Jacobs, S Mumford, S Martin, Sandiland (only by 0.2), D Swallow and T Green. That's another 8 players who failed to break the curse last year.
There are another 6 players who qualifies for this double curse for the upcoming 2016 season:
H Shaw: 92.6 (2014), 112.6 (2015)
L Montagna: 89.4 (2014), 111.0 (2015)
T Mitchell: 83.8 (2014), 104.5 (2015)
M Blicavs: 69.2 (2014), 104.3 (2015)
M Gawn: 63.2 (2014), 102.1 (2015)
M Bontempelli: 78.6 (2014), 103.2 (2015)
That brings the total to 49 players who have achieved the cursed double to date, will any of the above join Swan, Ablett and Libba in breaking the curse? Apart from Shaw and Montagna, I think the others all has a 'genuine' chance of adding to those who have succeeded to breaking the curse!
The Class of 2016 does looks a promising group, could 2016 be the year for 'smashing' this double curse to pieces...only time will tell.
Note - D Swan and J Steven just misses out on qualifying again...LOL!
Last edited:
Not really sure where to place this; it's clearly not on Rowsus's level in terms of advice, more just looking something based on a post by Grainfedbeef in another thread to see if it makes sense. Feel free to move / delete as appropriate and apologies in advance in that case.
Paraphrasing a bit, the basic gist seemed to be that it is better to accept the outlay for the previous years' top scorers (so M1-M3, D1-D3, F1-F3, and I guess essentially R1, R2 given the gulf to the pack of late) all else being equal because if you lock them in, you get their production and can speculate on an (M4-M8) / (D4-D6) /(F4-F6) type from a larger pool of options losing enough value to be a cost-efficient upgrade target to fill out your squad rather than having placed all your eggs in the basket of very specific topliners having to experience a similar drop to become reachable.
Thought it would be interesting to have a look and see how this has panned out in recent years in a generalized (i.e. fast) way.
Only had 2015-2019 data, so it's not all-encompassing, but in terms of ranks year to year, plotting the rank of the top 30 scorers from year n (I didn't have full position data across the years, so I couldn't just take F1 - F6 etc. without doing that manually) vs their rank in year (n+1) - and omitting players who had no value change over rds 1-13 of the following season (reasoning: simply applied screen for long-term injury), what it looked like was as shown below.
The % of players from the top 10 retaining that status the following year wasn't that high, around the 25% mark.
The % of players from the band (11-20) retaining their rank in that band or pushing into the top 10 was considerably higher.
(The rank data were based purely on yearly starting price with no manual adjustments for injury discount, i.e. they're not perfect.)
Then had a look at price differentials in year (n+1) for players based on rank in year n:
The price differential was based on (minimum price reached between rd 1 and rd 13 in year (n+1), based on the reasoning that coaches would look to bring the player in at their base price over that run in an ideal world and would ideally look to complete their side by rd 14-odd.
The result is below for bands (1-10), (11-20), (21-30). Ignore the regression line in quantitative terms, I just wanted to see in terms of +/-.)
It looked like there was a pretty reasonable chance of a player in the (1-10) tier shedding 140k+, certainly higher than in the other bands.
But in general, there were significant numbers of players in bands (11-20) and (21-30) shedding something in the 100-150k range.
Here, you see the average result across the years per rank (again, some higher losses amongst ranks 1-10) - I guess it's noticeable that the spread is higher in the top band, but that's logical enough, it's basically just the price formula equivalent of gravity:
Same thing in % terms, some higher % drops / larger spread for the top band:
Looking at the average price differential, that top band overall sheds value at an above average rate (but to be honest, not quite as markedly as I would have thought / hoped). You can see a number of bars clearly above the average in the band 1-10.
Then had a quick look at the scoring output in year n+1 of players by rank in year n (again, banded, so 1-10, 11-20, 21-30 etc.)
It was interesting that the top 10 from year n overall seemed to produced slightly less scores in the 115+ range than players in the band (11-20), but ideally captainable scores at a much higher clip in year n+1 (around 40% more often in relative terms). Both bands gapped the rest of the field on that front (I guess, as expected), bands 21-30, 31-40 and 41-50 are a fair bit behind the top 20.
There are a lot of factors not accounted for here (also things like the change back to longer quarters, or changes in other rules like kicking in etc.), but for me at first blush based on a really simplified look:
(i) The loss of value of the topliners was there but was not as pronounced compared to the next tiers as I would have thought
(ii) The topliners did tend to yield a higher % of topline captainable scores (which I guess is the argument many use - you're not buying one of them, you're buying two).
(iii) It is definitely possible to pick up fallen topliners at a good discount, the trick is knowing which ones, I guess...
(iv) To do this more properly, you'd have to look at stuff like average score as a % of top score in the line to see rate of drop-off, pool of hopefuls for those M3-M8 type spots etc.
Might have missed something / have plenty of flaws in the thinking, so grain of salt etc., but figured this would be a place where folks might be able to point out flaws etc. Either way, was good Spotfire practice for an hour or so an a more interesting topic than work applications.
Paraphrasing a bit, the basic gist seemed to be that it is better to accept the outlay for the previous years' top scorers (so M1-M3, D1-D3, F1-F3, and I guess essentially R1, R2 given the gulf to the pack of late) all else being equal because if you lock them in, you get their production and can speculate on an (M4-M8) / (D4-D6) /(F4-F6) type from a larger pool of options losing enough value to be a cost-efficient upgrade target to fill out your squad rather than having placed all your eggs in the basket of very specific topliners having to experience a similar drop to become reachable.
Thought it would be interesting to have a look and see how this has panned out in recent years in a generalized (i.e. fast) way.
Only had 2015-2019 data, so it's not all-encompassing, but in terms of ranks year to year, plotting the rank of the top 30 scorers from year n (I didn't have full position data across the years, so I couldn't just take F1 - F6 etc. without doing that manually) vs their rank in year (n+1) - and omitting players who had no value change over rds 1-13 of the following season (reasoning: simply applied screen for long-term injury), what it looked like was as shown below.
The % of players from the top 10 retaining that status the following year wasn't that high, around the 25% mark.
The % of players from the band (11-20) retaining their rank in that band or pushing into the top 10 was considerably higher.
(The rank data were based purely on yearly starting price with no manual adjustments for injury discount, i.e. they're not perfect.)
Then had a look at price differentials in year (n+1) for players based on rank in year n:
The price differential was based on (minimum price reached between rd 1 and rd 13 in year (n+1), based on the reasoning that coaches would look to bring the player in at their base price over that run in an ideal world and would ideally look to complete their side by rd 14-odd.
The result is below for bands (1-10), (11-20), (21-30). Ignore the regression line in quantitative terms, I just wanted to see in terms of +/-.)
It looked like there was a pretty reasonable chance of a player in the (1-10) tier shedding 140k+, certainly higher than in the other bands.
But in general, there were significant numbers of players in bands (11-20) and (21-30) shedding something in the 100-150k range.
Here, you see the average result across the years per rank (again, some higher losses amongst ranks 1-10) - I guess it's noticeable that the spread is higher in the top band, but that's logical enough, it's basically just the price formula equivalent of gravity:
Same thing in % terms, some higher % drops / larger spread for the top band:
Looking at the average price differential, that top band overall sheds value at an above average rate (but to be honest, not quite as markedly as I would have thought / hoped). You can see a number of bars clearly above the average in the band 1-10.
Then had a quick look at the scoring output in year n+1 of players by rank in year n (again, banded, so 1-10, 11-20, 21-30 etc.)
It was interesting that the top 10 from year n overall seemed to produced slightly less scores in the 115+ range than players in the band (11-20), but ideally captainable scores at a much higher clip in year n+1 (around 40% more often in relative terms). Both bands gapped the rest of the field on that front (I guess, as expected), bands 21-30, 31-40 and 41-50 are a fair bit behind the top 20.
There are a lot of factors not accounted for here (also things like the change back to longer quarters, or changes in other rules like kicking in etc.), but for me at first blush based on a really simplified look:
(i) The loss of value of the topliners was there but was not as pronounced compared to the next tiers as I would have thought
(ii) The topliners did tend to yield a higher % of topline captainable scores (which I guess is the argument many use - you're not buying one of them, you're buying two).
(iii) It is definitely possible to pick up fallen topliners at a good discount, the trick is knowing which ones, I guess...
(iv) To do this more properly, you'd have to look at stuff like average score as a % of top score in the line to see rate of drop-off, pool of hopefuls for those M3-M8 type spots etc.
Might have missed something / have plenty of flaws in the thinking, so grain of salt etc., but figured this would be a place where folks might be able to point out flaws etc. Either way, was good Spotfire practice for an hour or so an a more interesting topic than work applications.