#fancystats

Like most of us, I go to home team games put on by the local roller derby league. Unlike most of you, I’m fortunate enough to be in the vicinity of a banked track roller derby league, and to go to home team games put on by that league.

That league is, of course, the L.A. Derby Dolls.

When L.A. moved into their new home, the Dolleseum, they jumped right into a 12-game home season over the course of just six months. It was the first time they played in a double round-robin format for as long as I’ve been around, not to mention that many games in such a short period of time.

It was also an extremely competitive season. Of all of those games played, I could only recall one truly “bad” game, and maybe one or two mediocre ones. But every single game past that, including a stretch of five or six in a row, was extremely close through the first half at least. Many more games went deep into the fourth quarter, and a decent number of them went all the way to the last jam, including the championship game.

Out of curiosity, I went back through all of the games played in the 2015 LADD home season, just to see what this level of parity in competition looked like statistically.

Banked track data tracking isn’t as rich as it is in the WFTDA, so there’s no realistic way to calculate PPJ or MJD without manually tracking it at every game. But we do have the final scores, and the average score gap of all games played throughout the season.

I was a little surprised at how many games weren’t as close on the final scoreboard as they felt watching them live, and that this average wasn’t lower than I thought it was in my head.

It was here that I realized something: Just what is a good home season supposed to look like, anyway?

Every year, I track the scores and stats at the WFTDA Championships to check up on the progress of interleague competition among the top WFTDA teams. More close games and fewer blowouts is ultimately what everyone wants to see, and a lot us use this as a gauge to measure progress in the competitive aspect of the sport.

However, the level of parity in high-level interleague competition means very little to leagues that rely heavily on having a popular home team season to get fans to come out and spend money on tickets, concessions, merch, etc. Especially in the places where home teams are more popular than all-star travel teams, which is quite a few places.

In reality, the home seasons are where we should be focusing on to determine whether or not roller derby is getting more competitive in general. Every league has full control over their teams. Through various means, they can balance competition within their league, like allowing the weakest team the first pick of graduating juniors or top fresh meat skaters every season.

Ideally, a well-managed home league should more-or-less have balanced rosters leading to consistently putting on good-to-great games for their fans a majority of the time. Certainly, a low amount of bad games or blowouts. There will always be unbalanced teams here and there due to injuries or league transfers, but among several leagues over several years we should expect that these bumps will be smoothed out due to the law of averages.

This has inspired me to undertake an absolutely massive analysis project, which I am calling the Home Team Challenge. I will be collecting and organizing the home team scores of a significant number of WFTDA leagues, and any non-WFTDA leagues (RDCL, MADE) big enough to support a full home season, to see what there is to see.

And not just current or recent home team seasons, either. I'm going to try and go back as far and as deep as I can, for as many leagues as I can. I’m talking as many as 40 or 50 leagues, at least as far back as 2009 or 2008. (This seems to be the limit as far as statistical robustness goes in some of the league websites and Facebook pages I’ve dug through already.) With all of this data we can start asking, and getting answers to, some very intriguing questions:

• Was the massive drop-off of fan attendance from the peak levels in 2009 and 2010 in any way attributable to less competitive home league seasons throughout the country?

• How did WFTDA rule changes over the years affect competition within the isolated and more consistent environment of a home league season?

• Provided enough data exists for them to make a comparison, do leagues playing non-WFTDA rule sets have intraleague seasons that are—at least according to the final score—more competitive, less competitive, or about the same?

• Which leagues consistently have the most competitive or uncompetitive home league seasons, and why? Can this knowledge help other leagues put on more competitive home seasons as well?

• Do league championship games always pit the two best teams against one another? That is, are they always among the closest games of the year?

• Do higher-ranked WFTDA teams produce more competitive home team seasons? Does an all-star teams’ ranking have anything to do with how well that league’s home team season (if it has one) is doing?

That’s just the tip of the iceberg as far as how useful this data may be once I finish collecting it all in one easy-to-analyze place. Something of this magnitude is going to take a lot of time, probably the rest of this year, to collect the data and make up some whiz-bang visualizations from it. For that reason, this will be a side project that I’ll try to work on here and there, hopefully having it ready to go during the 2017 pre-season.

Having just gotten started, I’ve got data from as far back as 2008 for three leagues: Gotham, Rose City, and Bay Area. To pique your interest in this analysis project, let me show you the numbers from the 2015 home seasons of Gotham and Rose City.

First, Gotham, which has four home teams play in a simple six-game season with a single Championship game at the end, and has stuck to this format since pretty much the beginning.

As it turns out, last year Gotham had its most competitive home league season in its recorded history, which since 2008—that’s over eight years and 56 games—had an average score difference of 45 points. Not too shabby!

How about Rose City, then? Recently, the Rollers switched to a huge 14-game home season with their four teams, including a championship double-header.

Holy shit.

The chart speaks for itself. But to just add to it, in 2014 the league average score difference in Portland was 75 points, so things up there are getting worse in a hurry. When half-games results are adjusted to full-game score differences, Rose City’s home teams since 2009 have averaged a final score gap of 65 points—or 20 points less competitive than Gotham.

Fascinating, ain’t it?

As senior writer for The Hockey News, Ken Campbell makes a living by channeling Skip Bayless, King of the Bridge Trolls, which is a good way to get page views and make a lot of people dislike you, so there’s a two-birds-one-stone thing going on in a lot of the columns Campbell mails in writes.

I won’t link to his latest one—ever-so-trollingly titled “The death of Corsi? It’s only a matter of time now”—because traffic is the currency of the Internet and it's better for a lot of us if Ken Campbell goes broke.

If you’ve happened across his latest work, then you have a good idea of why things like numbers and graphs and calculators are the things that haunt Ken Campbell’s nightmares. If you’ve been lucky enough to avoid having the column retweeted into your timeline, you probably still know about this nightmare fuel because “common knowledge” is, by definition, knowledge known by nearly everyone.

Still: I blog, therefore I am. Or something like that. Which is why I thought I’d blog the ever-loving heck out of this Ken Campbell gem. Because an assault on logic of this scale is well worth unpacking.

Let’s begin.

  The death of Corsi? It’s only a matter of time

OK let’s stop right there. That title is…interesting. Why not “NHL to implement player tracking technology” or “NHL to partner with Sportvision by 2015 season” or something a little less incendiary? Because that would be a cold take, which is not at all Ken Campbell, that’s why.

Remember: this is Ken Campbell’s world and we’re all just nerding in it. (And even if it was actually an editor who wrote the title, you know Ken Campbell was all “YEAH! TAKE THAT, PENCIL PUSHERS! BOOM! ROASTED!” when he saw it.)

Yup, we certainly know what we’re getting in today’s column. Yessir.

  COLUMBUS – The pocket protector crowd may be dismayed to learn this, but the Death of Corsi is on the horizon. Push up your spectacles and deal with it, people.

“Deal with it, people.”

Ken Campbell is pulling no punches, as is his wont. Today’s target? The hordes of bloggers who always wear glasses, choose to carry slide rules in pants pockets usually reserved for GRIT, and have the type of hockey knowledge I’d venture to guess is intimidating to columnists like Ken Campbell.

Also: Ken Campbell is [incorrectly] predicting the future here, which makes for a strong opening. A STRONG opening.

  That much has become crystal clear this weekend when the NHL, in conjunction with a company called Sportvision, rolled out technology that will track puck possession, zone time and a host of other data far more accurately and easier to digest than what is currently available.

Things that are guaranteed: that there will be a host of data compiled by the NHL as a result of the Sportvision partnership.

Things that are not guaranteed: that the data will be pristine, free, or easier to digest than what’s currently available on NHL.com.

Also, it’s worth mentioning that googling “fancy stats primer” if you’re still hazy on Corsi, Fenwick, score effects, PDO, or zone starts is a pretty simple exercise. Then again, why google stuff when you can hot-take it to death?

Exactly. Now where were we…?

  And really, isn’t that what Corsi is doing right now? The game is all about puck possession and Corsi is a proxy way of measuring it. And while it has helped, this technology will allow the league to track it far more accurately. And once the league and Sportvision perfect this system, you don’t really need Corsi anymore.

After summarizing how the Sportvision technology would work, Ken Campbell straight up deads Corsi. WORLDDDDD STARRRRR.

Right, because that’s a logical conclusion to draw based on the fact that Sportvision’s data will actually enhance Corsi, not render it obsolete. Corsi by another name, as it were.

What Ken Campbell is overlooking—or possibly not understanding?—is that Corsi is a metric based on the raw data currently collected by the NHL: shots on goal, blocked shots, and missed shots. With the advent of Sportvision, the amount of similar multifaceted stats will only increase as bloggers get their hands on the raw data.

Kind of like what they’ve done since 2008 (or earlier).

Fact is, the metrics aren’t going away; they’re being improved. But Ken Campbell isn’t interested in things like “facts” because “facts” are the kind of cold water you don’t want thrown on your nuclear-weapons-grade fire-flames takes.

With the implementation of Sportvision, there will still be numbers, percentages, and graphics involved in hockey.

  So now instead of wading through a matrix of numbers, percentages and hard-to-decipher (for older white guys, anyway) graphics, you’ll now be able to know exactly how long a guy like Alexander Ovechkin played, how much time he had the puck on his stick, how much time his team possessed the puck when he was on the ice and how many shots were directed from where while he was on the ice.

I’ll acquiesce here. Time of possession at the individual-, team-, and zone-level is an incredibly important stat that Sportvision will hopefully provide. These will be super cool to (hopefully) see on the league’s website one day soon.

Still, that Ken Campbell is assuming every graph is out to get him is pretty revealing. As a supposed authoritative voice on all things hockey, isn’t it Ken Campbell’s job to, you know, learn about hockey stuff?

  And it will all be done without any guesswork.

To clarify: there isn’t any guesswork in counting Corsi events. It’s literally counting. You can do it on your fingers, if that helps.

  [In the next 267 words of the column, Ken Campbell gives context and quotes surrounding Sportvision and how the NHL hopes to leverage its technology]

Pretty standard reporting by Ken Campbell in this section of his column, which is surprising, given Ken Campbell’s apparent lack of interest in understanding what Corsi is, how it came to be, or why it’s an important metric, even though it’s routinely mentioned on hockey broadcasts nowadays.

Unimportant details, I guess?

  Unfortunately, some of the more subjective statistics such as hits, takeaways and giveaways, which can vary wildly from building to building and often favor the home team, will not change.

This subjectivity isn’t worth worrying about, for Ken Campbell cares only about standardizing—or killing off—the type of witchcraft responsible for Corsi, again conveniently forgetting that it’s a compilation of real, actual, quantifiable, not-at-all-subjective numbers gathered by the NHL.

  And while Corsi appears to be on its last legs, this new technology will almost certainly open opportunities for all kinds of new statistics, some we may not even know exist yet.

Just so I’m clear: Corsi is a waste of time and effort and graph paper because it’s something developed by bloggers based on the data collected by the league, but the prospect of “all kinds of new statistics, some we may not even know exist yet” based on Sportvision’s data collection services is cause for celebration?

OK, got it. Please continue.

  “You can’t have a math lesson on the screen,” [Hank] Adams said. “But there will be enormously deep data that sabermatricians [sic] are going to dig into and come up with all sorts of great insights.”

Oh, so just like Corsi? (And Fenwick? And PDO? And zone starts?) Basically, what we clarified above? Cool.

  Speaking of money, the technology is not cheap. It costs “several hundred dollars,” according to Adams to insert the chip into each puck. And if that puck goes out of play, nobody can expect it back. “At this point, it will be a very expensive souvenir,” Adams said.

Ken Campbell drops the hammer here, but does it ever so subtly.

Maybe he doesn’t even know he’s doing it. Maybe it’s actually his hot-take reflexes bringing this one home. (A “Jesus, take the wheel” moment if ever there was one.)

Still: the clinching quote about how a Sportvision chip-embedded puck will be an expensive souvenir is one final shot at how useless Corsi is. You can’t catch a Corsi after it goes out of play. There is no Corsi souvenir to be had. That’s just a Corsi fact.