Friday, October 3, 2014

The Year of Analytics? Using Advanced Statistics to Break Down the Sabres Preseason Opener

The hockey analytics movement has been gaining steam for some time but one could argue it reached a crescendo earlier this summer.

Many key players in the community have shut down their blogs and expansive websites and taken their talents to NHL clubs, with Edmonton, New Jersey and most notably Toronto among the teams investing heavily in analytics. The Leafs went so far as to hire 28 year old Sault Ste. Marie Greyhounds general manager Kyle Dubas to not only build their analytics department but gave him the coveted title of assistant GM.

Pretty heady stuff.

I'll admit that much of this talk about Corsi and Fenwick numbers made my eyes gloss over last year. I didn't need a statistician to tell me that the Sabres were bad - it was pretty apparent to anyone watching their games.

That said, I'm a huge numbers junky so it was natural that at some point, I'd want to learn more and the "aha" moment came when I bought Rob Vollman's Hockey Abstract, an incredibly detailed yet very approachable look at analytics. Inspired by Bill James' Baseball Abstract, Vollman delves into numerous topics throughout the book, with most starting with a question (i.e. "What makes good players good?") before he uses advanced statistics to reach potential conclusions.

It's a wonderful piece of work and to Vollman's credit, there isn't an ounce of pretension with his analysis and he's the first to say that analytics are meant to compliment traditional statistics and scouting methods as opposed to renounce them.

While I'm still very much at the beginning of the learning curve, I hope to learn much more about these key metrics that have been developed in order to help break down and analyze the Sabres and their players throughout the season.

A good starting point would be to look at the Sabres first preseason game this year, which took place on September 21st against the Washington Capitals...

September 21, 2014

NHL Preseason
Buffalo Sabres at Washington Capitals
Verizon Centre, Washington, DC

On the surface, there wasn't much to this game, which ended up finishing 1-0 in favour of the home Capitals. Andre Burakovsky, a highly-touted 19 year old winger, scored the game's only goal midway through the first period.

The traditional stats are pretty even across the board, as you can see below.

Shots were even at 25. Both teams got two power plays and both teams got exactly 3 minutes of power play time. Hits were also dead even at 26 apiece!

The Sabres fared a bit better at the face-off dot and blocked seven more shots than Washington. The Caps were also giveaway prone, with 12 compared to the Sabres 2.

While Washington elected to start a pretty strong team at home, with superstar forward Alex Ovechkin in the lineup, along with stalwart centre Nicklas Backstrom and one of the big offseason signings on defence, Matt Niskanen, Buffalo went with a very young and inexperienced lineup. In fact, the top scoring returning player in the lineup for Buffalo on this night was winger Brian Flynn, who produced a whopping six goals and seven assists last year. Ouch.

Here's how the Sabres lines looked that night, in the order that they came out in the first period:

Luke ADAM - Phil VARONE - Matt ELLIS



In the future, I plan on doing something similar to Joe Buscaglia's "Upon Further Review" write-ups over at WGR550's website on Bills games. There, he goes through every single Bills player that took part in the game, includes each player's "Plus-Minus" rating of positive plays versus negative plays and also assigns each player a game rating.

Here at Bleeding Blue and Gold, I'll break down each player by Corsi and Fenwick numbers, along with their Zone Start percentages and perhaps other key metrics if so desired by my readers.

In order to gain an understanding of all the underlying numbers, I decided to build my own spreadsheet and chart out the entire game myself as opposed to using one of the great analytics websites out there, some of which I have linked over on the right sidebar.

The NHL has comprehensive Play By Play sheets of every game, including those in the preseason and from these detailed game sheets, I can chart all of the shots, whether they are scored, saved, missed or blocked and I can get all of the face-off details, including zone start information, all of which add context to the possession and scoring chance metrics.

So I guess I should first say that since I'm new to all of this and still attempting to learn more about Corsi, Fenwick, PDO and all of the other metrics, there may be a chance my numbers are incorrect. If you spot something you feel is erroneous, let me know - I want to learn!

First, lets take a look at the traditional stats for the Sabres on this night.

Like I said, in the future I intend to provide analysis on every player in the lineup but in this first post, I want to narrow my focus on perhaps the most intriguing training camp battle this year in Buffalo...


While Ted Nolan or Tim Murray haven't said as much, many observers are under the impression that Reinhart and Grigorenko are fighting for one roster spot, with the "loser" of the competition either being sent back to junior hockey, in the case of Reinhart, or to Rochester in the AHL if it's Grigorenko.

The traditional stats above show Reinhart was on the ice for the Burakovsky goal and therefore was credited with a -1 on the night. Reinhart had no shots on goal, one hit, one blocked shot and one takeaway while winning 60% of his face-offs. He also received almost 17 minutes of ice time, including just over 14 minutes at even strength.

Meanwhile, Grigorenko had two shots on goal, three hits, one blocked shot and won face-offs at a fine 88% clip. Grigorenko only played a little over 12 minutes, all but a few seconds of it at even strength.

So Reinhart played close to five minutes more than Grigorenko, but had no shots versus Grigorenko's two and was also on the ice for a goal against.

Those numbers still don't say an awful lot but most observers of the game noted that Grigorenko played a very strong game while Reinhart was just "average". Lets look a bit deeper to see if we can find more...

For those new to advanced statistics, a Corsi number gives an approximate measure of puck possession by taking into account not only shots on goal, but shots that missed the net and shots that were blocked.

In essence, the rationale behind it is that if a team has a significant advantage in total shots directed at the goal, the team will naturally possess the puck that much more than their opponents.

Corsi, like most other key metrics, is most often calculated for play at even strength for obvious reasons. So with that out of the way, here are the Corsi numbers for Reinhart first:

Corsi FOR: 8
Corsi AGAINST: 14
Corsi +/-: -6
Corsi %: 36.4% (Corsi FOR divided by (Corsi FOR + Corsi AGAINST)) --> (8/22 = 36.4%)

Ok, where did I get those numbers and what in the hell do they mean?

Reinhart, as mentioned, played just over 14 minutes at even strength and while he was on the ice, the Sabres had five shots on goal, all of which were saved, one shot that missed the net and they also had two shots blocked. That adds up to 8 shots directed at the goal by the Sabres while Reinhart was on the ice, the "Corsi FOR" number.

The flip side saw Reinhart on the ice for 1 goal against, another 7 Washington shots on goal that were saved by Lieuwen, 2 Capitals shots that missed the net and 4 shots that were blocked by the Sabres. That adds up to 14 shots directed at the Sabres goal by the Caps while Reinhart was on the ice.

So Reinhart's aggregate Corsi plus/minus number is -6 on the night and the 36.4% number equates to the amount of puck possession the Sabres had while Reinhart was on the ice. Not great numbers obviously but not much worse than the Sabres put up on average last year.

Small sample size aside, that's pretty cool, huh?

Now, lets see if Grigorenko's numbers are similar:

Corsi FOR: 12
Corsi AGAINST: 6
Corsi +/-: +6
Corsi %: 66.7%

So that's interesting. Grigorenko played two minutes less at even strength but was on the ice for 12 positive Corsi events, including 9 shots on goal, 2 shots that missed the goal and 1 shot that was blocked. He was only on the ice for 2 even strength shots against and 4 Washington shots that were blocked, giving him the Corsi AGAINST number of 6.

So Grigorenko was a +6 in Corsi events and had a whopping 66.7% Corsi rating, meaning that Grigorenko and his teammates essentially controlled possession for a full two-thirds of the time he was on the ice, which isn't uncommon for such a small sample size but would be absurd, in the most positive way imaginable, if sustained over the long-term. Which would be highly unlikely.

Now, naysayers will say that stats don't tell the whole story and that is definitely true in almost all cases, including Corsi and Fenwick, which I'll get into another day. There are a few things that can be considered to put things into context for both Corsi and Fenwick:

1. The quality of his line mates has an effect on his stats
2. The quality of his opposition has an effect on his stats
3. How he is deployed or used has an effect on his stats

There are formulas out there to help calculate both the strength of a players line mates (#1) and the strength of his opponents (#2) and at some point, I'll go into that a bit deeper.

#3 is a bit easier to understand and quantify and that's by charting zone starts.

As an example, last year, a guy like Zenon Konopka would be utilized heavily as a face-off specialist and as such, Nolan would give him the tough defensive zone face-off assignments. He rarely got the chance to start play in the offensive zone, mostly due to his lack of scoring ability and those assignments were given to guys like Cody Hodgson.

So with Konopka starting so many shifts in the defensive zone, it's only natural that there was a greater likelihood of him and his line mates to have to immediately start defending and the chances of having shots directed at the Sabres goal were much greater than if he got the majority of his shifts in the offensive zone.

Hopefully that's pretty clear!

So lets look at Reinhart and Grigorenko to help provide a bit more context to their Corsi numbers.

Offensive Zone Starts: 3
Defensive Zone Starts: 2
Neutral Zone Starts: 0
Offensive Zone Start %: 60.0%

Offensive Zone Starts: 3
Defensive Zone Starts: 3
Neutral Zone Starts: 2
Offensive Zone Start %: 50.0%

So not much difference there but even then, the numbers are a bit different than we might have seen in past years, as Grigorenko was given sheltered minutes, meaning he was given the plum offensive zone assignments to a much greater extent in order to mask his perceived defensive deficiencies. But not in this game, as Nolan gave him just as many defensive zone starts as offensive zone starts. As the preseason continued, it's become quite apparent that Nolan is giving both of these talented youngsters every opportunity to impress, giving them key minutes at even strength and on the power play in addition to giving them talented line mates in the last few preseason games.

In this game, Grigorenko won all three of his defensive zone face-offs and both of his neutral zone face-offs while going 7 of 8 overall.

It rubbed off on his line mates for this game too, as both Brian Flynn and Vaclav Karabacek had similarly good possession numbers and the three of them were the only positive possession players up front for Buffalo out of the 12 forwards dressed.

We'll get into different metrics in the future, like the subtle difference between Corsi and Fenwick numbers but for one game at least, it's quite clear that Mikhail Grigorenko was more dangerous and had a much more productive game compared to fellow top prospect Sam Reinhart.

Let me know what you think of all of this, either here in the comments section or give me a buzz on Twitter.

Thanks for reading.

No comments:

Post a Comment