Skip to main content
The Official Site of the Tampa Bay Lightning

Mishkin's Musings: Where I stand on hockey analytics

by Dave Mishkin / Tampa Bay Lightning

Several years ago, the Lightning were about to face an opponent in a mid-November game. It was their first contest that season against this club. In looking at the other team’s stats, I noted a certain player’s assist total. This player, who had been in the league for a couple of seasons, had been a former high draft pick. Much was expected of him. To that point in his career, he had not lived up to those expectations. At the start of this season, however, his assist numbers were solid – and, based on the stats, he seemed poised to have his first reasonably successful offensive season.

At the morning skate, I asked the other team’s radio broadcaster about this player, inquiring if his elevated stat numbers were indicative of an elevated level of play. In response, my colleague chortled. He then told me: “Do you know what (he) leads the league in? Second assists on five-on-three power play goals.”

He was kidding, of course. The league didn’t (and still doesn’t) track leaders in secondary assists on five-on-three goals. His point was that, despite the higher assist total, the player wasn’t necessarily performing better than he had in the past. It was a cutting (and funny) way of expressing it, though. Sometimes a player earning the second assist has an integral role in setting up the eventual goal. That isn’t always the case when a five-on-three goal is scored, though, since the attacking team has so much room in the offensive zone to make plays.

In this case, a hockey stat didn’t provide a full picture of what was happening with a certain player. I felt it was a useful example because of the recent attention given to “advanced stats” in the sport – and the heated debates that have accompanied them.

First, a little background. Many teams have been tracking various advanced stats for a number of years. But a couple of seasons ago, on-line bloggers covering specific teams began writing about them. They’ve gained momentum since then, to the point that the advanced stat terminology has become commonplace in the sport’s vernacular.

Not everyone in the hockey world has welcomed this new verbiage with open arms. Many longtime hockey writers have been skeptical about the value of these stats. What’s resulted reminds me of the movie “Moneyball”, in which Billy Beane (played by Brad Pitt) favors a “numbers” approach in choosing which players to sign over the his scouts’ traditional “eyeball” test. Except that this argument isn’t taking place within organizations. It’s happening in newspapers and in the blogosphere.

Their squabble makes for an interesting sidebar, but it’s not the point of this column. Rather, I’m going to give you my two cents on four of the most commonly-referenced advanced stats and how useful I think they are.

The “Corsi” number and “Fenwick” number both essentially track the same thing: shot attempts. Corsi accounts for all shot attempts – those on net, those missing the net and those that are blocked by the opposition. Fenwick does the same, excluding blocked shots. Corsi and Fenwick do account for the difference between five-on-five numbers and power play time, as well as whether a game is tied, close (within a goal) or lopsided.

Why are they considered useful? All teams want to possess the puck, but the league does not track puck possession. So this is an alternate method. One would figure that, in a certain game, the team with more shot attempts possessed the puck more than the opponent, so they had more opportunities to score. The advanced statisticians tally Corsi and Fenwick numbers not only for teams, but also individual players.

“PDO” and “Zone Starts” are two others. PDO adds shooting percentage and save percentage. Similar to Corsi and Fenwick, it’s used for teams and players. Most teams and players will and should reach a total of 100%, contend the statisticians. Let’s say a team scored on 8 percent of its shots and stopped 92 percent of opposition shots. Adding them together makes 100 percent. So if a team or player has a PDO over 100%, the statisticians argue, then the club or player may be enjoying some lucky bounces. Bounces that won’t continue. A PDO under 100% shows the opposite. PDO is tied to Corsi and Fenwick in that, if the percentages for PDO even out, then the best way to score more goals is to shoot a lot and the best way to keep pucks out of your net is to limit opposition attempts.

Zone Starts refer to which zone a player is in when a faceoff occurs. It does not include neutral zone faceoffs. Zone Starts show how a coach might use a player. Someone with more offensive faceoffs than defensive would be an offensive threat. A player with more defensive zone starts would be viewed as a more defensive player.

So what’s my opinion? Thinking about these stats is similar to me asking my kids about their day when picking them up from elementary school. Those parents with children that age know what I mean. I get an answer, an idea of what happened, but it’s not the complete picture.

When I prep for a game, I review a lot of stats, but not specific Corsi or Fenwick numbers (the NHL doesn’t include those in its official stat package or on nhl.com). I do look at team’s average shot totals, both for and against. I also examine the blocked shot number. A team with both a low SA/G and blocked shot number would figure to be a good puck possession team. But I’m also interested in the blocked shot number for another reason. A Lightning opponent that blocks a lot of shots is, well, good at blocking shots. I conclude that the Bolts may have trouble getting shots through to the net that night. I don’t think it means that the opponent definitively struggles with puck possession.

I agree that, generally speaking, Corsi and Fenwick do reflect puck possession. But not always. Look at last night’s game between the Lightning and Detroit Red Wings. For two periods, the Lightning held the Wings to just three five-on-five shots (Detroit netted two power play goals in the first 40 minutes). When the game was being played five-on-five in the first two periods, the Lightning did have the puck more than the Red Wings – and were able to generate more pressure. But at the start of the third period, the Wings enjoyed their best stretch in the game. For most of the first nine minutes, Detroit forechecked the Lightning effectively and forced a number of turnovers after failed Lightning clearing attempts. It was a grind for the Lightning, who couldn’t move the puck up ice with possession. Eventually, the Wings tied the score at three. The final 11 minutes were back and forth, as both clubs did good work to keep the puck in the other team’s end for extended shifts. But those first nine minutes of the period were lopsided in favor of the Red Wings. The shot numbers didn’t reflect that, though. They had only three shots on net (including the goal) and just four others blocked. Those totals weren’t appreciably higher than any other point in the game – the Wings finished the contest with 20 shots on goal, 16 attempts blocked and only eight missed shots. But it was the hardest nine minutes the Lightning endured in the game. Much harder than, say, if the Red Wings had fired seven isolated and unscreened point shots in those same nine minutes.

Still, team Corsi/Fenwick numbers can be helpful. I’m less rosy on individual Corsi or Fenwick numbers. A player on a good Corsi/Fenwick team should have positive Corsi/Fenwick numbers. And vice versa. Similar to the plus/minus stat. (Two exceptions. A player with a terrific Corsi/Fenwick on a bad Corsi/Fenwick team is noteworthy. As is the reverse.) But I don’t place nearly as much value on them as the team number. It’s similar to the example given in the first paragraph. Hockey is a fluid game. One player on the ice is dependent on the four skaters out there with him, as well as the goalie. Stat-heavy baseball, on the other hand, which is more static, can assign individual credit or blame very effectively. For example, a pitcher’s ERA is not affected when an error leads to a run scored. The player that committed the miscue is charged with an error, though. In hockey, one player’s mistake that leads to a goal results in a minus rating for the other four on the ice and it also against counts against the goalie’s GAA. Similarly, regarding Corsi/Fenwick, a player may be pinned back in his own zone for a long shift (one in which the opposition attempts many shots), but that player may have had little to do with why the puck came into the zone – or why it stayed there.

Regarding PDO, I agree that shooting percentages tend to even out. But not all shots are created equal. Some are harmless and others are dangerous. And some players are destined to score at a higher percentage because of the quality of their shot. Steven Stamkos’ shot is one of the best in the league, so it’s more likely to go into the net than the typical shot from most other players. Also, for goalies, not facing many shots can be more challenging than dealing with lots of them. That’s because they may endure long stretches without facing a shot, then suddenly contend with an opposition scoring chance. So like Corsi and Fenwick, I see how PDO can be useful, but don’t feel it tells the whole story of “puck-luck”.

Zone Starts don’t interest me much, because, in prepping for an opponent, I can already have an idea which players are considered “offensive” or “defensive”. During the Lightning’s Stanley Cup Championship season in 2003-04, head coach John Tortorella seemingly exclusively used Captain Dave Andreychuk and Tim Taylor on defensive zone faceoffs. (Andreychuk for faceoffs to the goalie’s right and Taylor to the goalie’s left). It’s because they were responsible defensively and very good on faceoffs. Broadcasters calling games involving the Lightning knew that. But maybe some viewers or listeners didn’t. So for a casual fan not clued into a certain coach’s preferences, I suppose Zone Starts can be helpful.

My conclusion: regarding the advanced stats debate, I’m not firmly in either camp. Corsi and Fenwick, with a side of PDO, provide a broad picture of puck possession. But not a detailed one. One final note on the player referenced at the beginning of this column. He has gone on to have a solid career. He just needed more experience before he hit his stride. And in defense of his assist total from years ago, at least he was on the ice when his team had a five-on-three.

View More