Analysis The Clarkson Rebuild Model - A Reason For Optimism

Jan 17, 2008
29,054
59,120
Brisvegas
AFL Club
Hawthorn
Proportion of Picks that Produce an AA.jpg

Using analysis completed on the draftguru website, I have plotted the proportion of players drafted which produce an AA player. The analysis was completed using uneven-sized bands. You can see its more or less linear (negative) after the first 5 picks. But you are still talking less than 1 in 5 players drafted 6-10 become AA players. @Absai you will note that pick 15 has a higher probability of producing an AA than pick 9 based on historical precedence ;)
 
View attachment 710691
Using analysis completed on the draftguru website, I have plotted the proportion of players drafted which produce an AA player. The analysis was completed using uneven-sized bands. You can see its more or less linear (negative) after the first 5 picks. But you are still talking less than 1 in 5 players drafted 6-10 become AA players. @Absai you will note that pick 15 has a higher probability of producing an AA than pick 9 based on historical precedence ;)
It’ll be good to see Sicily & Mitch Lewis **** with the system in a year or 2 & add to AA appearances for those picked a little lower down the line :)

Also, the model can’t account for HFC Players that should have been AA but weren’t, on account of their being HFC Players. Although, this is probably off-set by Geelong Players that were AA but shouldn’t have been; selected on account of their being Geelong Players...



Seriously though, does this model include Rookie Picks?
 
Jan 17, 2008
29,054
59,120
Brisvegas
AFL Club
Hawthorn
It’ll be good to see Sicily & Mitch Lewis **** with the system in a year or 2 & add to AA appearances for those picked a little lower down the line :)

Also, the model can’t account for HFC Players that should have been AA but weren’t, on account of their being HFC Players. Although, this is probably off-set by Geelong Players that were AA but shouldn’t have been; selected on account of their being Geelong Players...



Seriously though, does this model include Rookie Picks?
I’ll have to check about rookies. Breust and dean Cox are a couple I can think of that made it to AA. But they also appear in the national draft as when you get upgraded you get selected with the last pick of the club.

It shows the top 50 are generally similar outcomes after the first round. This is the first chart I have posted with data beyond pick 50.
 

flinchfree

Premiership Player
Jan 30, 2014
4,006
9,125
AFL Club
Hawthorn
Just get an A Grader back in Tom Mitchell and still be in the mix, injuries permitting.

Everyone drastically overestimates what is required to compete in the AFL at the moment and drastically underestimates what we have.


I think there is MUCH to this.
What people truly underestimate, is form.
Many of our great 80's teams had plenty of just good players, not great.
But the game plan was perfect for our elite players, and great confidence and form and belief arose from winning early and then the determination not to drop off.
That WILL, was as great as anything we did in drafting or trading or developing.

Our current crop doesn't play with that will, that comes from winning and form and playing near your absolute potential regularly.
Reckon it's getting enough players to that level, from our very good and above players, so that an almost unstoppable force of playing style is repeatable week in week out.
We either don't have the right players this minute, or the game plan needs further tweaking, to feed into getting the most of the talent we do have.

Just bringing in another A player like Cogs won't do it.
It will take an offseason, with some genuine coaching input and magic seasoned into what we're tyring to do, to get everyone up and not just trying - but achieving - close to each players best.

Reckon we could be a top 3 team next year with exactly the same 22 including Tommy Mitch, playing a very different game of footy.
At the moment it's just a grind that the team has to plough through, there's no energy to lift tired legs by scoring 8 goals in a quarter, or a 6 goal run on, that makes everything so much easier.
 
Last edited:
Jan 17, 2008
29,054
59,120
Brisvegas
AFL Club
Hawthorn
It occurs to me that the old father son rule allowed you to take eligible players in the third round which aligns with 30-50 band. Ablett, Scarlett, Hawkins etc. probably greatly distort some of the stats on their own! I’ll see if I can replicate the draftguru analysis and then control for fs selections to see if that drags that proportion back down to where the 20-30 band is. If it did then we can say outside the first round has almost no effect on probability of selecting an AA.
 
Also to add, Wright is known to have one of the longest running and most developed player lists in the AFL on all picks and player development and career progression. Over time that data is only going to become stronger to help make predictions that find diamond in the rough players from what is considered a low value pick but really pick 40-80 is the same value as most clubs do not have the same data/experience.

There are many crazy things I've heard about the team under Wrighty, my favourite is having lists of draftees they missed from certain years where they monitor progress on the traits they liked and target later down the line if they still see it in them. I'd love to work in that area.
 
Jan 17, 2008
29,054
59,120
Brisvegas
AFL Club
Hawthorn
For anyone interested, I played around with some modelling last night and being a father son selection couldn’t explain the additional frequency of AAs in the 30-50 range.

Pick alone isn’t particularly explanatory in terms of predicting games played or AA etc. once you add in a range of variables the effect is so small it is not statistically significant at the 95% level. What is missing are more independent variables. Data from under 18 champs, draft combine etc. would be needed to tease out some predictive indicators of future success. Getting that data will be a challenge.

I tried to model AA and Premiership players using clubs as an indicator variable. You will all be happy to know hawthorn had the largest effect on wether or not a player would become a premiership player :) but that is simply a function of us having won a lot of premierships and as such it isn’t really a good measure to model in this sense.
 
Jan 17, 2008
29,054
59,120
Brisvegas
AFL Club
Hawthorn
In your first graph it looks to me like only the first three rounds have picks that distribute normally. After that it looks like each round's picks form a Pareto distribution. In fact you can see that the average games played for each pick round forms a Pareto distribution. This leads me to believe that the top end talent is being picked correctly because their longevity in the game distributes as you would expect with one end of the data bound to a fixed value (round 0, the start).

Perhaps the likelihood of winning a premiership has more do with factors other than draft round, which aren't modelled, which is what you're arguing in effect. Maybe something like core group longevity regardless of round picked - perhaps the number of years the core of the team has played together is the determining factor. Draft could possibly influence that as we see the first three rounds gives you the best chance of picking long lasting players. Or maybe something seemingly unrelated to on-field performance is the driving factor, like board stability.

In fact, I wouldn't be surprised to see something external to the football department providing just that clue. Anecdotally, the Hawks dominated the premiership the last 55 years and since the draft was introduced in '86 the Hawks have been the undisputed kings. But everyone has drafted under the same rules so you would expect more even distribution of results. In fact, some clubs have been utterly, consistently abysmal year after year despite a very heavy finger on the draft scale.

The draft is undoubtedly the primary tool to drive player parity between clubs, but talent parity is not the only factor which affects performance, and perhaps not even near the top. All things being equal, the longevity distribution in the first graph would balance performance over time between clubs. But all things aren't equal and the draft appears to be the last lever of parity, not the first. I guess this is why the AFL introduces success taxes and hands out failure bonuses to clubs, but money is a tool not an influencer.

So, there you go. Does this help? I don't know. But factors like board or football department turnover would be interesting to explore for a correlation to premierships. Got any spare time?
As per my last post, I attempted to control for club but as the stability/management of the club is not a constant over time, it was never likely to produce definitive results. Could probably try and treat the data as panel data and see if I can tease out more of that effect. Maybe using a 10 year rolling average of ladder position or something similar might work as a proxy. I’ll have a think about that.

I looked into the Pareto distribution. It’s not one I have ever used before. Plotting histograms for individual picks in the top five showed the first few picks are flat distributions as opposed to normal. Maybe even a little skewed to the right depending on how many bins you choose. There is only about 16-17 records to work with once you prune early and recent drafts so it certainly could be normal. I don’t think a formal test will show it though. Grouping picks into variously sized bands does produce a distribution that looks Pareto for picks after 3 so you may well be right there. I attempted to download a program to fit a Pareto distribution however I wasn’t show how to correctly specify or interpret it and the only instructions appear to be a paywalled journal article :)
 
Jan 17, 2008
29,054
59,120
Brisvegas
AFL Club
Hawthorn
What do all these players have in common?

711626


Before I answer, lets look a linear prediction model which takes a players games, goals, if they got AA or B&Fs and predicts how many career Brownlow votes that player received.

Linear PRediction of Career Votes - 1991-2012 Pick < 51.jpg


Hopefully this is clear :)

Basically the lighter the colour, the lower the draft pick e.g. pick 1 is yellow while pick 124 is blue (noting that I attempted to limit the chart to the top 50 picks). As you would expect, the colour gets lighter as the individual gets more career votes implying there is a correlation between pick and votes but it just isn't very strong. The model does a good job of forecasting career votes explaining about 62% of the variation in votes with just those aforementioned variables. It struggles with players who scored more than 100 career votes so basically consider those players freaks! Pick taken is controlled for through a random effects model which just treats picks as randomly different to each other rather than ordinal. Games played is the important variable here and pick is weakly predictive of games played but really only for the first round and after that it flattens out.

Games Played by Pick 1991-2012 Pick < 51.jpg


Which brings me to the next chart. This is the same chart except the colour indicated premiership players (premiership player = 1). You will notice that players that score a lot of career votes tend to win premierships (also they play a lot of games). Only 7 players in history have recored more than 125 votes and not won a premiership. This is 30% of all players within the scope of the analysis. Can you guess who? :)

Linear Prediction of Career Votes - 1991-2012 Pick < 51 Prem.jpg


All 7 scored an AA. Only one (still active player) hasn't played 200 games. Only one didn't win a B&F.

What is all this saying? There is a lot of volatility of outcomes for ALL draft picks. After the first round the outcomes are fairly flat up to about pick 50. All those times Hawthorn traded out a late first round pick and got back a second round pick...you can bet we knew what we were doing. The question is, did other clubs? Has the landscape changed? Certainly future pick trading increases liquidity which allows more trades but also inflates prices. But still, when you look at the probability of getting an AA caliber player, two first round picks is a bargain. A draft focused strategy has to include picks in the top 5, and probably top 3. Throw a heap of picks at the board inside the top 50 and turn the list over quickly. The hawthorn strategy of finding better value through trades will remain a solid strategy until clubs start putting a higher price on exposed form.
 

CHUBBSS

Premiership Player
Apr 6, 2010
4,870
11,817
AFL Club
Hawthorn
Although it is highly unlikely it would be eye-opening to see behind the scenes in 10 years time of the data the good clubs are using to draft. The data shows the hawks have been ahead of the game with the trading strategy but that is not to say strong draft picks don't mean good players, just look at Kelly at Geelong.

I personally think the club has found significant value in later round drafting that it forces early round trading. Until that stops the clubs does not have change strategy
 

tige19

Cancelled
10k Posts
Feb 13, 2003
21,724
24,129
AFL Club
Hawthorn
James Sicily likely to hold off signing a new 2 year deal until after the trade period too, no interest in leaving but management have advised to hold off from signing (which he was happy to do).
 

SuperSic

All Australian
Oct 16, 2018
766
3,056
AFL Club
Hawthorn
What do all these players have in common?

View attachment 711626

Before I answer, lets look a linear prediction model which takes a players games, goals, if they got AA or B&Fs and predicts how many career Brownlow votes that player received.

View attachment 711629

Hopefully this is clear :)

Basically the lighter the colour, the lower the draft pick e.g. pick 1 is yellow while pick 124 is blue (noting that I attempted to limit the chart to the top 50 picks). As you would expect, the colour gets lighter as the individual gets more career votes implying there is a correlation between pick and votes but it just isn't very strong. The model does a good job of forecasting career votes explaining about 62% of the variation in votes with just those aforementioned variables. It struggles with players who scored more than 100 career votes so basically consider those players freaks! Pick taken is controlled for through a random effects model which just treats picks as randomly different to each other rather than ordinal. Games played is the important variable here and pick is weakly predictive of games played but really only for the first round and after that it flattens out.

View attachment 711665

Which brings me to the next chart. This is the same chart except the colour indicated premiership players (premiership player = 1). You will notice that players that score a lot of career votes tend to win premierships (also they play a lot of games). Only 7 players in history have recored more than 125 votes and not won a premiership. This is 30% of all players within the scope of the analysis. Can you guess who? :)

View attachment 711639

All 7 scored an AA. Only one (still active player) hasn't played 200 games. Only one didn't win a B&F.

What is all this saying? There is a lot of volatility of outcomes for ALL draft picks. After the first round the outcomes are fairly flat up to about pick 50. All those times Hawthorn traded out a late first round pick and got back a second round pick...you can bet we knew what we were doing. The question is, did other clubs? Has the landscape changed? Certainly future pick trading increases liquidity which allows more trades but also inflates prices. But still, when you look at the probability of getting an AA caliber player, two first round picks is a bargain. A draft focused strategy has to include picks in the top 5, and probably top 3. Throw a heap of picks at the board inside the top 50 and turn the list over quickly. The hawthorn strategy of finding better value through trades will remain a solid strategy until clubs start putting a higher price on exposed form.

Brilliant Bris, and once again backs up our list management strategy perfectly. I’m very confident that we know what we’re doing.
 

Bumps"N"Grins

Premiership Player
Oct 15, 2014
4,292
8,156
AFL Club
Hawthorn
Other Teams
Sell the team Dolan
James Sicily likely to hold off signing a new 2 year deal until after the trade period too, no interest in leaving but management have advised to hold off from signing (which he was happy to do).
Will get more should big names not come. Very good move by his management and James on a personal level.
 
James Sicily likely to hold off signing a new 2 year deal until after the trade period too, no interest in leaving but management have advised to hold off from signing (which he was happy to do).
He's still got a bit of time because he's contracted till the end of 2020.
 
What do all these players have in common?

View attachment 711626

Before I answer, lets look a linear prediction model which takes a players games, goals, if they got AA or B&Fs and predicts how many career Brownlow votes that player received.

View attachment 711629

Hopefully this is clear :)

Basically the lighter the colour, the lower the draft pick e.g. pick 1 is yellow while pick 124 is blue (noting that I attempted to limit the chart to the top 50 picks). As you would expect, the colour gets lighter as the individual gets more career votes implying there is a correlation between pick and votes but it just isn't very strong. The model does a good job of forecasting career votes explaining about 62% of the variation in votes with just those aforementioned variables. It struggles with players who scored more than 100 career votes so basically consider those players freaks! Pick taken is controlled for through a random effects model which just treats picks as randomly different to each other rather than ordinal. Games played is the important variable here and pick is weakly predictive of games played but really only for the first round and after that it flattens out.

View attachment 711665

Which brings me to the next chart. This is the same chart except the colour indicated premiership players (premiership player = 1). You will notice that players that score a lot of career votes tend to win premierships (also they play a lot of games). Only 7 players in history have recored more than 125 votes and not won a premiership. This is 30% of all players within the scope of the analysis. Can you guess who? :)

View attachment 711639

All 7 scored an AA. Only one (still active player) hasn't played 200 games. Only one didn't win a B&F.

What is all this saying? There is a lot of volatility of outcomes for ALL draft picks. After the first round the outcomes are fairly flat up to about pick 50. All those times Hawthorn traded out a late first round pick and got back a second round pick...you can bet we knew what we were doing. The question is, did other clubs? Has the landscape changed? Certainly future pick trading increases liquidity which allows more trades but also inflates prices. But still, when you look at the probability of getting an AA caliber player, two first round picks is a bargain. A draft focused strategy has to include picks in the top 5, and probably top 3. Throw a heap of picks at the board inside the top 50 and turn the list over quickly. The hawthorn strategy of finding better value through trades will remain a solid strategy until clubs start putting a higher price on exposed form.


My head hurts ...

.
 
Jan 17, 2008
29,054
59,120
Brisvegas
AFL Club
Hawthorn
My head hurts ...

.
I have another observation that is supported by the data. Among players drafted by Hawthorn since 1991 (I wanted to include Crawford), players drafted under Clarkson (I included 2004 draft given those players played their whole career under Clarko) are significantly more likely to have played in a winning premiership team than if they hadn't being drafted under clarko :)

The clarko effect remains when you compare all players drafted from 1991 to 2012. In fact, its even stronger! There have been 26 premierships from 1991 to 2018 (not including 1991) and of the 1,821 players drafted in the national draft in that time (up to 2012), 229 have gone on to become premiership players (12.6%). 1,779 players were drafted not under clarko (including Hawthorn players such as Hodge, Mitchell, Crawford etc.). Of those 210 have gone on to become premiership players (11.8%). 42 players were drafted to Hawthorn (2004-2012) of which 19 became premiership players which is a whopping 45.2%!!! If it wasn't obvious, the best thing a player can do for their career is play for Clarko! :) :) :)

As per below, Clarko has taken them from all through the draft. Only 3 players drafted under Clarko played more than 100 career games and did not win a premiership. Tom Murphy, Brendan Whitecross (injury restricted) and Shane Savage, with Savage playing most of his games for StKilda. Conversely, only 3 players drafted in this period in the National draft played less than 100 games in their career and won a premiership: Max Bailey (injury restricted), Brent Renouf and Stuart Dew. For anyone wondering, Will Langford is another who fits this criteria but only made it to the senior list in 2013 (via pick 71). So it seems that Clarko is very good at determining who can take the team to a premiership and invests games into them. Players who can't take the team forward appear to get around 20-25 games in most cases if they are even getting time in the seniors. So when you see the likes of Howe, Obrien, Hardwick, Sicily pass the 50 game mark on the way to 100 games we should all have a bit of confidence that Clarko knows what he is doing. He gets it right nearly every time.

Players Drafted Under Clarko.jpg
 
Last edited:
I have another observation that is supported by the data. Among players drafted by Hawthorn since 1991 (I wanted to include Crawford), players drafted under Clarkson (I included 2004 draft given those players played their whole career under Clarko) are significantly more likely to have played in a winning premiership team than if they hadn't being drafted under clarko :)

The clarko effect remains when you compare all players drafted from 1991 to 2012. In fact, its even stronger! There have been 26 premierships from 1991 to 2018 (not including 1991) and of the 1,821 players drafted in the national draft in that time (up to 2012), 229 have gone on to become premiership players (12.6%). 1,779 players were drafted not under clarko (including Hawthorn players such as Hodge, Mitchell, Crawford etc.). Of those 210 have gone on to become premiership players (11.8%). 42 players were drafted to Hawthorn (2004-2012) of which 19 became premiership players which is a whopping 45.2%!!! If it wasn't obvious, the best thing a player can do for their career is play for Clarko! :) :) :)

As per below, Clarko has taken them from all through the draft. Only 3 players drafted under Clarko played more than 100 career games and did not win a premiership. Tom Murphy, Brendan Whitecross (injury restricted) and Shane Savage, with Savage playing most of his games for StKilda. Conversely, only 3 players drafted in this period in the National draft played less than 100 games in their career and won a premiership: Max Bailey (injury restricted), Brent Renouf and Stuart Dew. For anyone wondering, Will Langford is another who fits this criteria but only made it to the senior list in 2013 (via pick 71). So it seems that Clarko is very good at determining who can take the team to a premiership and invests games into them. Players who can't take the team forward appear to get around 20-25 games in most cases if they are even getting time in the seniors. So when you see the likes of Howe, Obrien, Hardwick, Sicily pass the 50 game mark on the way to 100 games we should all have a bit of confidence that Clarko knows what he is doing. He gets it right nearly every time.

View attachment 711877
Hey man,

Can you flick this last one to Cogs - tell him the best thing he can do to achieve career success is to play under Clarko?

Thanks,

B
 

Topiary

Premium Platinum
Jul 22, 2016
6,972
26,923
AFL Club
Hawthorn
James Sicily likely to hold off signing a new 2 year deal until after the trade period too, no interest in leaving but management have advised to hold off from signing (which he was happy to do).

Can I say this is utter BS. I am unequivocally told, Sic, his management and the Club have already agreed to what his re-signing will look like. It’s announcement is “just PR timing”.
Sorry, but this is pure BS and mischief.
 

tige19

Cancelled
10k Posts
Feb 13, 2003
21,724
24,129
AFL Club
Hawthorn
Can I say this is utter BS. I am unequivocally told, Sic, his management and the Club have already agreed to what his re-signing will look like. It’s announcement is “just PR timing”.
Sorry, but this is pure BS and mischief.

2 year deal Tops.. will be announced after trade period. But as you say, things are fluid so could change? Certainly a 2 year deal right?
 

Carl Spackler

Club Legend
Jul 29, 2018
2,406
5,642
AFL Club
Hawthorn
2 year deal Tops.. will be announced after trade period. But as you say, things are fluid so could change? Certainly a 2 year deal right?

Where did Tops say things were fluid and could change?

What's up into the descent to trolling 24x7. I recall that you used to enjoy the football. Now you seem to more enjoy spoiling the joy of others. What's up?
 

tige19

Cancelled
10k Posts
Feb 13, 2003
21,724
24,129
AFL Club
Hawthorn
Where did Tops say things were fluid and could change?

What's up into the descent to trolling 24x7. I recall that you used to enjoy the football. Now you seem to more enjoy spoiling the joy of others. What's up?

Trolling? I’m pretty confident thanks
 
Back