Firstly, this is the first blog that’s being posted exclusively at the new site. TheFootyGuys now points to Fantasy Insider, and all content will appear here from now on. We’ll find a way to use the old name somewhere as this moves forward!
One thing that’s interested me in thinking about lineup optimisation and projections for AFL and NRL is how much of the US theory for NFL, MLB and NBA can be applied. While a lot of things are implied in US discussions, such as the benefit of stacking, it’s mostly discussed without any factual data points. For example, it’s widely stated that there’s a correlation between WR and QB, but how significant is that correlation, and how does it compare (for example) to MD and FD in the AFL. So, for the last few hours I’ve been experimenting with NFLDB, an open source data collection of NFL stats, which – by the way – puts to shame the AFL and NRL’s proprietary stat collection. AFLTables does a great job of making available, for free, what they do, but there’s nothing on that level in the NRL, and this freely available NFL source lists the detail of every single play, meaning there’s no gatekeeper (*cough* Champion *cough*) to the advanced stats.
But, rant aside, what did I find?
Here’s NFL QB’s to WR’s, since 2009, plotted for each game, and each QB to WR pairing, with at least 5 targets for the WR, and at least 100 yards for the QB (to minimze the impact of backups and injuries), plotted according to their Fanduel score:
Across this dataset, the overall correlation is 22% – which is a little higher than FR and MF in the AFL, plotted here by Moneyball scoring, where the correlation is 14% (where playtime for each player is over 80%):
What may be interesting to explore further is the differences between QB’s, and between QB/WR pairings. For example here’s Peyton Manning and Eric Decker (31 games, 33%):
And Peyton vs Demayrius Thomas (46 games, 46%):
The other Manning has some much higher correlations though. Here’s Eli Manning and Odell Beckham (11 games, 55%):
And here’s Eli Manning and Victor Cruz (50 games, 61%):
And they’re certainly not the extreme examples (I’m currently iterating through the various combinations to explore further). So, early experiments, but there are a bunch of ideas worth exploring here, and theories worth testing. There’s no doubt stacking makes more of an impact in the NFL than the AFL, but in both sports it’s about identifying the right types of stacks to optimise your lineups..