Parity Podcast Episode #306

Player matchups are absolutely a confounding factor. So are the specific roles people play on the field (e.g. a lot of high-touch, low-possession players get stuck with the disc in bad places because their "high-possession" receivers bail it and don't get open again). Having better receivers reduces your throwaways even if you don't actually throw it better. Playing good defence doesn't matter if the rest of your team is bad.

There are a lot of valid reasons to criticize any statistical assessment of a player. The goal here is not to come up with a perfect number, but to get a sense of the relative contribution of different players. It's useful especially in a league like this where you're trying to get the best value players on your team, not necessarily the "best" (highest salary) players.

You're right that someone like Giulian's value is inflated by his weaker matchups, but since he's going to continue to get those matchups his rating is a fair indicator of his value in this league.

Analysis of player stats has definitely influenced my decisions by making me aware of things I wasn't otherwise seeing. One clear example is how much more playtime I gave Khalid Al-Zahrani on the Outlaws after my analysis showed he was one of our most effective defenders, which isn't necessarily obvious when you watch him play because of his unorthodox style.

On Carleton, I would often use stats for grouping/splitting players into practice teams (or sometimes for tournament rosters). You can use this kind of analysis either to make balanced teams or, alternately, to put all of one kind of player on a team, which forces people to take on a different role.

Not sure about predictive value. In the short term, you might be able to predict game results based on the relative value of their players, but I haven't tried. Longer term, it's very difficult to quantify things like "potential" when talking about player development.

Is there value in looking not only at points that start on defence, but also points where they play defence at all? 

For the defensive stats.

Seb's encouraging words notwithstanding, Tom has actually been the one taking on our toughest match-ups while still putting up great offensive numbers. Amos may tease, but there's no debate over who is the best Newman.

So there are two different offensive plus-minus measures I'm aware of: (#1) the shorthand individual summary calculation (goals + assists + blocks - turnovers) and (#2) offensive efficiency (how often a player scores when starting a point on O).

I plotted them both against an average of my three offensive indicators. The first kind of plus-minus actually correlates pretty closely with my offensive rating (R2 = 0.608), which is surprising since the shorthand method doesn't take into account the number of points played or anything like that.

The second kind of plus-minus (offensive efficiency) has a weak positive correlation with my offensive rating (R2 = 0.188), but it's pretty scattered. What that means is there are a lot of people who are individually effective with the disc, but their teams are failing to score (i.e., strong players are being let down by weaker teammates); and conversely, there are a lot of people who are not very effective with the disc but their teams are scoring anyway (i.e, weak players are riding the coat-tails of better players).

If offensive efficiency is such a poor indicator of offensive performance, than why do I use defensive efficiency in my ratings? The straightforward answer is a lack of alternative defensive measures, but I think there's a more fundamental difference here. On offense, you only need two people to be successful. As long as they play well, it doesn't matter how good your other four players are. On defense, on the other hand, everyone on the field has to play well (in the context of their individual matchup) in order to get a turnover. It doesn't matter if you have the two best players on the field playing defense if the four weakest players on the offensive team can win their matchups... poaches notwithstanding.

Basically, it's a lot harder to inflate your defensive efficiency number than it is to inflate your offensive efficiency number.

Just pick the slowest person on the field and if I'm covered, cover someone in my weight class