Statistical Analysis of drafts

[Patty_Moz]

I got bored over this draft/trade period and thought I might look back over the draft.

This might have way too many numbers for most people but I've been back through the drafts from 1988-2011 and compiled the average games that are played from each draft pick just to get an idea of how important draft picks are compared to trading in players.

The results can be seen in a couple of different graphs in the pdf (hopefully its attached properly).

To sum up quickly
Top 5 picks in regards to average games played:
1. Pick 1- 140.2 games
2. Pick 2- 137.3
3. Pick 3- 115.0
4. Pick 5- 103.1
5. Pick 13- 97.8

Not many surprises there except for perhaps pick 13 sneaking into the top 5 but definitely goes to show how important those top 5 picks are.

Bottom 5 picks in the top 40:
1. Pick 35- 27.6 games
2. Pick 22- 39.0
3. Pick 28- 41.4
4. Pick 6- 44.0
5. Pick 23- 44.8

Big surprise is pick 6 which over 24 yrs just hasn't seemed to produce any real quality yet. Most games from this pick is 151 while all other top 20 picks have produced at least 1 200 game player.

I will post some more things to have a look at and think about and if anyone else is interested and they want me to get certain stats I can see what I can do.

 

[Knightmare]

Nice research. Always interesting these statistical findings.

 

[bombermick]

Pick 6 is an odd one. Often clubs choose good mids but they're not the absolute cream and go for riskier players in the 10-13 range - many whom end up coming better players.

 

[Igloo]

Tom Williams has beaten the average games for pick 6. I'm so proud. We've got that pick again this year, hopefully we get someone who can do better.

Nice work, Patty.

 

[BABYBOMBERS2007]

Pick 6 is a strange one.

 

[Aldo]

Goddamnit we have picks 6 and 22 lol

 

[Lana]

A problem with such analysis is for many reasons(zones, lack of draft camp etc) the early drafts results have little to no relevence for the value of drafts today and recent drafts have players whom have yet to finnish and thus skewing the results. I thought about doing a similar thing but there isn't enough years of relevent data to create a decent model.

 

[swansfan51]

This is fantastic, great work. Can you generate stats on the percentage of players that go on to play 100 games or 50 games? For example picks 1-3 might be 50%, picks 10+ might be 30% etc

 

[scottwade]

Interesting, thanks.

Is there any reason why you chose 1988 as a start point?
Just looking at the drafts it seems to me that from 1992 onwards the picks seem to have a more 'scientific' edge in relation to selection.

 

[dajesmac]

Why would you include 2011? Won't the fact that they've played a max of 20-odd games skew the average?

 

[Chrizzt]

Great work, have been wanting to see something like this for a while (but too lazy to do so myself). Would be interesting to see std devs etc. especially for the second round picks to see how consistent they are.
Wingard was pick 6 last year, so expect him to break the games record several years down the track

 

[Masto]

Its very even after the first few picks, all the way through to about pick 57. Of course games played is no measure of the quality of those games played. I wouldn't mind seeing a quality one based on champion data or something.

 

[Hawkinspired]

Beau Dowler and Mitch Thorp come to mind on #6 failures

 

[Patty_Moz]

A problem with such analysis is for many reasons(zones, lack of draft camp etc) the early drafts results have little to no relevence for the value of drafts today and recent drafts have players whom have yet to finnish and thus skewing the results. I thought about doing a similar thing but there isn't enough years of relevent data to create a decent model.

Yeah you are right and these things almost stopped me from doing it but I just got bored and thought I'd have a look anyhow and despite the drawbacks the data still paints a pretty reliable picture that the top draft picks are far and away the best. Also if you look at the recent drafts which have lots of players that haven't finished their careers the numbers still show teams get more service out of top draft picks so it doesn't really skew the results but simply decreases averages. The most reliable data is the decade from '92-'01.

This is fantastic, great work. Can you generate stats on the percentage of players that go on to play 100 games or 50 games? For example picks 1-3 might be 50%, picks 10+ might be 30% etc

Yep can definitely do that, may take a couple of days tho.

Interesting, thanks.

Is there any reason why you chose 1988 as a start point?
Just looking at the drafts it seems to me that from 1992 onwards the picks seem to have a more 'scientific' edge in relation to selection.

Mostly because prior to '88 lots of the top picks (who were the best players from their state leagues) didn't bother coming to play AFL, from '88 onwards at least most of the tops picks came to play even if just for a short while.

Excellent point about '92 from that point onwards recruiters seem to get things right. I will post some more info looking at drafts from '92 onwards.

Why would you include 2011? Won't the fact that they've played a max of 20-odd games skew the average?

Yeah it does but I just threw it all in there and will dissect it a bit more to exclude recent yrs.

 

[Suma Magic]

Awesome stuff Patty_Moz - can you be bothered doing it with brownlow votes?

 

[yaco55]

You have to go back to at least 1992, seeing that 1994 is in the 3 best drafts of all time.

 

[Tazwegian]

A lot of this really just boils down to statistical anomalies though. It is not as though pick 6 is actually less valuable than pick 20 or that pick 13 is really more valuable than pick 4. To smooth out the data a bit, maybe break the picks down into a range. Such as 1-10, 11-20-21-30, and so on. Even a range of five might be useful, but the data will be less smooth. The smaller the range, the more an anomaly can throw it out more easily.

 

[Lach72]

James Walker the pick 6 star

 

[Trav 20]

Bottom 5 picks in the top 40:
1. Pick 35- 27.6 games
2. Pick 22- 39.0
3. Pick 28- 41.4
4. Pick 6- 44.0
5. Pick 23- 44.8

Big surprise is pick 6 which over 24 yrs just hasn't seemed to produce any real quality yet.

Most surprisingly, Melbourne has never had pick 6.

 

[Lazar Goz]

Weird

 

[35Daicos]

Most surprisingly, Melbourne has never had pick 6.

Allen Jakovich was pick 6 for the Demons at the 1990 National Draft. Quite a talent and loved to put on a show, too!

 

[NUFC Tiger]

Conca well on his way to beating that pick 6 hoodoo.

 

[Lorenzo]

The last 5 pick 6's have not been too bad

the list is composed of Meyers, Yarran, Rohan, Conca and Wingard

It seems this group will break the hoodoo

 

[Patty_Moz]

This one looks at the drafts from 92-02 as they are the most reliable in terms of players having completed their careers or played a substantial part of them, while the recruiters start getting things right.

Top 5:
Pick 1 during this time averaged 208 games.
Pick 2 192 games.
Pick 3 170 games
Pick 8 155 games
Pick 10 140 games


Also looks at the 92-11 drafts which shows very similar trends but as a lot of players are just starting their careers the averages are substantially reduced.

Top 5:
Pick 1 averaged 157 games.
Pick 2 142 games
Pick 3 130 games
Pick 8 105 games
Pick 4 103 games

 

[Patty_Moz]

The next thing I looked at has been the most interesting (so far IMO).
These graphs separate the averages for blocks of 20 picks from 1 to 60 with trend lines.

What I found interesting is the evenness of picks 21-40. There actually seems to be an upward trend for this block of picks ie. the later picks have been performing better than the earlier ones in the 21-40 range. This to me shows how even all drafts seem to be in the 21-40 range while after the 40 mark the quality drops off as expected. Other factors may be that teams take best available with their first pick and select for needs with their second pick. (This upward trend extends extends to about 45 then drops away).

With this in mind and as I was originally thinking about trading, to me it would appear if you had an early 20's pick and could pick up a quality player in a trade while still keeping a pick in the late 30's early 40's chances are you won't lose too much in what you can pick up in the draft so definitely worth a crack.

I'll do that for 5 and 10 pick trends as Tazwegian suggests but probly won't get that done before the draft.