Twitter Was Made For Hockey Talk

I had a great talk on Twitter with former Ranger beat writer and current Sporting News writer Jesse Spector about advanced hockey statistics. Being a huge proponent of advanced stats, and a guy who loves to talk hockey in general, I jumped right in and overloaded Jesse with tweets. Since Twitter is pretty tough for explaining things in detail, I'll try to expand on the discussion here.

It started with this tweet from Jesse:

Here's an expansion of how I responded to Jesse on Twitter:

Relative Corsi, of course, measures the difference in shot differential when a player is on the ice vs. when he isn't. This allows us to see who really makes their team's shot differential better, and who makes it worse. Brent Burns' current mark of 33.2 (at 5 on 5) leads the league among all skaters who have played at least 10 games. The next closest is Colin Wilson at 26.8, so Burns is in fact obliterating the NHL in this stat. 

Burns' on-ice Corsi is 26.93, and his off-ice Corsi is -6.28. This means that, when Brent Burns is on the bench, San Jose (a team who outshoots opponents by 4.7 a game) is a negative shot differential team. This tells us one of two things: Brent Burns is either incredible (he is) or San Jose gets an insane boost from their power play to assist in their shot differential. Without having specific shots for and against numbers with the man advantage, I'd have to guess it's a solid mixture of both.

However, as I mentioned to Jesse, when we combine this stat with Burns' offensive zone start percentage of 58.8%, the picture becomes slightly clearer. Burns starts nearly 60% of the time in the offensive zone. This, by definition, means that we should expect a positive shot differential if Burns is doing his job. It's difficult for opponents to rack up shots when the puck is starting 200 feet away from their net. Regardless, Burns' figure is still very high and very impressive. He excels at what he does: moving the puck, generating shots, keeping possession, and being a stud defenseman for the Sharks.

I then compared Burns to Ryan Nugent-Hopkins as an example. RNH has an even higher O-Zone start percentage: 71.3%. Based on the great start to his season, we'd expect him to have good Corsi numbers too, right? Turns out that he has a -6.02 Corsi and only a 2.5 relative Corsi. This tells us that A) Edmonton is a bad possession team 2) RNH can thank the power play for his production. There is some blame for his teammates, but overall, RNH should have a stronger gross Corsi number given the amount of time he starts in the offensive zone.

Jesse wrote this tweet shortly after:

I wrote a great deal of tweets responding to this one. Here's how I'd answer the question:

The stat referenced is Relative Corsi Quality of Competition. The title is a bit misleading - this stat measures the relative Corsi of Player A's opponents while Player A is on the ice.

Let's use Joe Pavelski as the example. Pavelski has a league-leading CorsiRelQoC, which means that  he is constantly being matched against strong opponents who are putting up high relative Corsi numbers. 

Here's the most important part: because relative Corsi is used, this doesn't mean that Pavelski's opponents are dominating him in shot differential. It just means that the opponents' relative Corsi is higher. It's possible, if not common for low shot differential teams, to have a negative gross Corsi and a positive relative Corsi. Therefore, Pavelski could still be holding opponents to a negative shot differential---that number could just be "less negative" for that opponent than it typically is. This means that Joe Pavelski is facing the toughest competition in the league, the players most likely to increase their team's shot differential to a number more positive than it typically is.

CorsiRelQoC doesn't tell us much about Pavelski's performance, because we don't know anything about his own Corsi. All we know from this stat is that he faces the toughest competition in the league. CorsiRelQoC tells us his role. Based on the fact that the Sharks and Rangers have the top 5 players in this stat, it tells us that those teams give these players heavy minutes against the opponent's top players, and stick to those rigid, line-matching systems. However, to get any insight into performance, we need to look at the players' own Corsi numbers.

Sure enough, Pavelski has a strong relative Corsi of 11.3, which I think is extremely impressive given the competition he plays against. These two stats, combined, summarize that Joe Pavelski logs minutes against the opposition's best players (more so than any other player in the NHL), and yet he still maintains a double digit relative Corsi. 

This is especially impressive because, by definition, Pavelski's teammates are playing against weaker competition than he is. Pavelski's off-ice Corsi is 1.33, and his on-ice Corsi is 12.67, meaning that Joe does roughly 9.5 times better than his teammates do, all while facing much tougher competition than those teammates.

I brought up Michael Frolik on Twitter as an example of the opposite. Frolik's CorsiRelQoC is 6th in the league, meaning he plays against the 6th best competition in the league. His relative Corsi, meanwhile, is -8.5. Since Frolik is playing against incredibly strong competition, it's okay to expect him to have weak gross Corsi numbers--he is, after all, playing against the best, and having a on-ice Corsi of -0.80 is holding water. 

While the situation is not dire, however, his team has strong shot numbers when Frolik is on the bench---they're a much worse shot differential team with Frolik on the ice. This tells me that Frolik isn't capable of these types of minutes. The Blackhawks would be better served decreasing his ice time against the strong competition he's been seeing. This is especially considering that his offensive zone start is only slightly below 50%. It's not like Frolik is constantly logging defensive zone minutes, or is a defensive specialist. 

Again, maintaining a on-ice Corsi number around 0 (or slightly below it, in Frolik's case) against this type of competition is not horrendous--many players would be below zero. However, given the extremely negative relative Corsi, I believe the Blackhawks have better players elsewhere to handle these types of minutes.

In my opinion, this also highlights how strong of a season Pavelski and similarly situated players such as Nicklas Lidstrom, Joe Thornton, Ales Hemsky (when healthy), and Radim Vrbata are having. These players are all in the top 14 of CorsiRelQoC, yet they all maintain double-digit relative Corsis. That tells me that these players are rising above that strong competition, unlike Frolik, and excelling no matter who is on the ice against them, relative to their teammates, who are facing weaker opponents. That's a quality to admire, and a quality that we can detect through these advanced stats.