Professor’s Super(coach) Computer | Player Comparisons

Published by The_Professor on

We’ve been joined by the Professor and you’d better be taking notes because there WILL be a test!

The Professor uses patented Super Computer technology to calculate highly complex Supercoach algorithms. Today he has run a couple of player comparisons through his Super(coach) Computer and the results are in!

ADAM CERRA v CALEB SERONG

The Super(coach) Computer has run a comparison between Adam Cerra ($490k) and Caleb Serong ($451k) – both topical players right now, especially for coaches looking to find some bargains in the midfield.

“I ran two linear regression models, with Supercoach (SC) score being the output, and with different predictor variables…looking at Time on Ground (ToG) Percentage, Contested Possessions (CP), Uncontested Possessions (UP) and Metres Gained (MG).

We want to see if any of these numbers either weakly, moderately or strongly predict SC scores (either -ve or +ve) for either player.”

Caleb Serong ($451,000 – MID)

The model (and the variables within it) was able to predict 79% of SC scores…which is reasonable. Here’s the breakdown of the significant findings:

  • If he increases his yearly contested possessions by 1 (on average), there’s a 79% chance that he’ll increase his average SC score by 2.75pts this year.
  • If he increases his yearly uncontested possessions by 1 (on average), there’s a 79% change that he’ll increase his average SC score by 3.14pts this year.
  • MG and ToG showed bugger all.

Adam Cerra ($490,200 – MID)

The model (and the variables within it) was able to predict 81% of SC scores…which is v.good. Here’s the breakdown of the significant findings:

  • If he increases his yearly contested possessions by 1 (on average), there’s a 81% chance that he’ll increase his average SC score by a whopping 5.05pts this year.
  • If he increases his yearly uncontested possessions by 1 (on average), there’s a 81% change that he’ll increase his average SC score by 2.52pts this year.
  • Again MG and ToG didn’t show much.

I looked a little further at the graphs for Cerra, particularly at how his CPs stacked up to SC points (see below). The matching overlay of the CPs and SC scores – particularly the trend upwards from games 45 onward – is very interesting!

What does this comparison tell us…not all players score equally in CPs and UPs (1 CP/UP doesn’t equal 1 SC pt, obviously), but some score a lot more, and in this case a huge argument could be made statistically in favour of Cerra, who seems to have a better chance to average more SC pts for the season if both players increase their CPs by 1 (on average) this year.

ISAAC HEENEY v JORDAN DEGOEY

Isaac Heeney ($454,500 – FWD)

Really interesting!!

The model (and the variables within it) was able to predict 94% of SC scores…which is amazing!

  • If he increases his yearly contested possessions by 1 (on average), there’s a 94% chance that he’ll increase his average SC score by 2.45pts this year.
  • If he increases his yearly uncontested possessions by 1 (on average), there’s a 94% change that he’ll increase his average SC score by 2.03pts this year.
  • This was all without adding GOALS into the model itself…

When Goals kicked was added, the model went to shit…all of a sudden the only thing significantly related to SC scores for Heeney was Goals Kicked. Essentially if Heeney averages 1 more goal per game this year than previous (he averaged 1.6 goals/game last year) than there’s a 81% chance that he boosts his SC average by approx. 12pts overall!

Question is…does he kick more goals this year to equate an average of 2.6 goals/game considering he’s meant to be playing midfield? Effectively he needs to kick over 50 goals!

If he doesn’t kick more goals, it is hard to see where the points come from within the modelling. Relying on CD heavily rating his possessions, which they do tend to favour for Heeney. But, I don’t like the pick from a statistical perspective!!

Having said that, he could come out and average 33 touches and 15 CPs in the first 6 rounds and go up $250k!

Jordan De Goey ($463,500)

A very similar story here…

He plays forward, and he needs to kick goal to score SC points…across the last 2 years AND 4 years the models ONLY showed a statistically significant relationship between goals kick vs. SC pts for DeGoey.

None of the other variables mattered.

However, let’s say he plays midfield this year (as expected, and same as Heeney), if you take goals out of the model, then there was a significant relationship between CP…if De Goey averages 1 more CP this year (on average) then there’s an 83% chance that he averages a huge 6.9pts more than last year!!

If you asking me, there is a better risk/reward ratio in favour of Jordan DeGoey if he plays midfield in terms of what his CPs means for his SC pt averages, despite there being a lower chance (greater margin of error) for this not occurring compared to Heeney – probably due to injuries and too much variability in DeGoey’s scoring over the last 4 years!!!

What do you think community, are you taking on any of The Professor’s result? Let him know on Twitter – @Dale_Harris36


Subscribe
Notify of
guest
6 Comments
Oldest
Newest Most Voted
Inline Feedbacks
View all comments
Maverick

Can anyone tell me if we are still drafting tonight in the jungleland keeper league I think it is? I can’t seem too find the league anymore but I was in it only a few days ago…

Russty_

hi Mav, it had to be scrapped coz Hedski never showed up this is the code for the new one 583740 get in there mate, draft is at 9pm eastern 6 pm western time.

Maverick

Cheers russty I’m in mate.

Russty_

Thanks Professor, I like reading this kind of number crunching and then compare how it turns out at the end of the season, cheers! 🙂

Tracey3Tits

Averaged over the season Serong scored at 0.917 PPM (range 0.6 to 1.3) 1,823 points 1987 minutes
His TOG ranging from 77minutes to 103 minutes across the season with an average across 22 games of 90.3 and a mean of 89.
As the mean is the point at which half of his times are above and half are below I can not agree that a change in his TOG will be ineffectual to increasing his scores. Based on the data a mere 2 minutes extra each quarter would statistically result in 7.34 points extra p/game. = 1.45 pts p/game more than your possessions data
It is realistic to suggest that in all of his games this year he will have greater TOG than last years mean.
Offical game time is 100 minutes but ‘time on’ is included in a players official TOG stats consequently players can have more than 100 minutes TOG in a game that technically only lasted for 100 minutes

TheProf

Hey Tracy, good breakdown! The models adjust for all of the predictor variables and some are more compelling than others. So when we say that ToG showed bugger all, we really mean that when ToG is factored in with ALL other variables, it’s not as strong a predictor as the others. That’s certainly not to say that on its own, there wouldn’t be a relationship between ToG and SC pts…there most probably would be. Many predictor variables would probably share a relationship with SC pts. What we’re trying to do is figure out which ones are more (or less) statistically significant than others for certain players. Hope this helps 🙂