Summer Guest Column Series: Explaining Ottawa's Meteoric Rise - Zone Starts and Mean Regression
(Editor's Note: During the summer of 2012, the Senators website will welcome a number of guest columnists, who will write articles on items of interest to them and to Senators fans. The article below comes to us from Sens blogger Travis Yost. Visit www.ottawasenators.com and follow @NHL_Sens for future guest columns. Follow Travis on twitter @TravisHeHateMe)
The Ottawa Senators took the National Hockey League by storm last season, and by the fall of Game 82, both local and national hockey analysts were touting the club as one of the true surprise teams of the season.
After all, an 18-point improvement behind first-year head coach Paul MacLean was hardly expected — the team was almost universally projected to finish inside of the lottery in 2011-12.
While the media wasn't wrong in its analysis of Ottawa's achievement, one of the vaguely odd developments of last year's run was the relative inability of some analysts to rationalize both how and why this team made such a jump in the standings. Sure, Erik Karlsson and Jason Spezza scored a ton of points and Craig Anderson stopped a lot of pucks. But when truly pressed on the question, many shrugged their shoulders and would apply the same exhausted, unsubstantiated logic that appeals to the unknowns — grit, hustle, leadership, et cetera.
That's not to say the team didn't possess those qualities. Rather, it's that the 18-point improvement can much more likely be chalked up to other factors beyond such attributions.
Ottawa's play at even-strength — specifically, their ability to score goals at even-strength — was one of the main reasons why the team made such a positive dent in the win/loss column in 2011-12. Ottawa's +0.2 (GF/GA /60) at even strength was 12th best in the NHL, finishing as one of 13 clubs at or above break-even. Not surprisingly, every one of those 13 teams made the post-season.
Still, that only explains the how element of the playoff equation. Explaining the why half is a bit more difficult, but I'd point to two measurements — specifically zone starts and shot variance — to explain a large portion of the above.
The importance of zone starts can't be understated. Teams that correctly employ their talent in each third of the ice and manage time-on-ice (TOI) accordingly tend to yield the most positive gains over 60 minutes. Perhaps the most prominent example stems from the Vancouver Canucks, who continue to destroy inferior opposition during the regular season with extreme zone start percentages. Last season, head coach Alain Vigneault used both Henrik and Daniel Sedin — his two primary scorers — in more than three of every four offensive zone starts; conversely, he used shutdown centre Manny Malhotra in a complete defensive role, giving him just 13 per cent of offensive zone starts last season.
Rationale is simple: Scoring opportunities — and subsequently, goals — create wins and losses. The Sedin twins are second to none when it comes to creating; players such as Manny Malhotra are strong at preventing them. Vancouver's approach is to create a positive margin in the scoring opportunity department on a game-to-game-basis, and considering its track record and winning percentage over the past few years, that's almost impossible to dispute.
The 2010-11 Senators could also be considered as a compelling argument for importance of zone starts. Rather than apply the Vancouver — or Pittsburgh — coaching mentality of exploiting the opposition, Ottawa’s system appeared to share ice time evenly across the board.
In fairness to departed head coach Cory Clouston, he did have to deal with a number of external variables —injuries, roster moves, and superior opposition on a nightly basis, which didn't make his job easy. Still, his zone starts and player usage were interesting. Below, the '10-11 splits:
The first number you should notice is Jason Spezza's team-low offensive zone start percentage. Spezza almost certainly possesses the strongest offensive skill set on the team, but was forced to the neutral and defensive zones more than any other player on the team. Perhaps Clouston didn't have a choice but to trot out his No. 1 centre due to the dearth of talent down the middle.
Even the other forwards — Daniel Alfredsson and Milan Michalek, for example — didn't see a ton of high-leverage offensive zone starts. When your best goal scorers are forced to play defensive minutes, total team output tends to suffer.
One year later, Paul MacLean brought an up-tempo, puck-possession based system to the Ottawa Senators organization — a move I felt was the biggest positive change above all else last season. The approach was much like the one seen in Vancouver (referenced above), Detroit, and Chicago — goals win hockey games.
MacLean knew that his team was suffering in the depth department and fresh off a season where ittowed a -58 goal differential -- a combination of a weak offence and poor defensive play /goaltending. In other words, MacLean didn't have the same luxury many other coaches were afforded.
Instead, MacLean aligned his 12 forwards and six defencemen into three separate roles — point-scorers, two-way players and defensive specialists. The zone start percentages that developed over the 82-game sample size were incredibly indicative of this:
How was Spezza given this kind of flexibility? Well, MacLean entrusted Zack Smith, Zenon Konokpa and Jesse Winchester with more of the high-pressure defensive minutes — correctly playing to their strengths and weaknesses. MacLean quickly realized that Spezza's most valuable play would come from heavy TOI in the offensive zone.
The same can be said for Norris Trophy winner Erik Karlsson. While questions concerning potential defensive liabilities were present throughout the season, Karlsson's value also comes in the form of offensive zone minutes. The disparity between his presence and the presence of your league-average (or team-average) player in the same offensive zone minutes is massive. On the other hand, the disparity between Karlsson and the league-average ( or team-average) player in defensive zone minutes is much closer in margin. MacLean relied on his bottom four to eat up the tough defensive zone minutes, then sent out Karlsson to create havoc against the opposition's netminder.
Spezza finished with 84 points on the year, while Karlsson registered 78. Spezza moved from a .91 PPG player in 2010-11 to a 1.05 PPG player in 2011-12. Karlsson moved from a 0.60 PPG player in 2010-11 to a 0.96 PPG player in 2011-12.
You'll also note that there wasn't really one specific player carrying an incredibly low offensive zone start percentage. Plenty of that has to do with the already-referenced puck possession strategy brought forth by MacLean as he transitioned from Detroit to Ottawa. The goal was to control play, and as a byproduct of that, the team took some much-needed pressure off of Anderson's shoulders, playing behind a somewhat makeshift blue line that still struggled defensively.
A change in coaching strategy and mentality certainly played some role in the standings uptick, but that wasn't the only variable in play here. While MacLean certainly deserved his spot among the Jack Adams Award finalists, it should be noted that he was jumping to a team that was inevitably headed toward positive mean regression, following the team’s struggles in a non-playoff year in 2010-11.
Shot variance is the second measurement that generally goes unnoticed, although its role in the importance of winning percentage can't be emphasized enough. In the NHL, you will always have your extreme outliers — Sidney Crosby and Alex Ovechkin on one end; your enforcers/tough guys on the other. It's the endless total of names in the middle —- where talent disparity from one organization to the next just isn't as big as some perceive — that remains the focal point. Most players are simply interchangeable from one roster to the next.
Most of those who study hockey statistics tend to look at shot percentages, operating under the mentality that roughly the same number of chances should go in for each team over the course of a season.
QuantHockey has done a fantastic analysis of the above, and has found that shooting percentages among forwards and defencemen floated around 10.5 per cent and 4.6 per cent last year, respectively.
Why is this important? Well, during the difficult 2010-11 season, the Senators were brutally victimized by almost unfair shot percentages in both respects — teams tended to beat them at a higher rate than pegged against the league average, and the current forward/defencemen group didn't appear capable of buying a goal.
One statistic that attempts to measure shot variance and the role of luck in the NHL is PDO, which operates much like BABIP does in the study of baseball sabremetrics. PDO's equation is two-fold, and is described as follows:
PDO is the sum of "On-Ice Shooting Percentage" and "On-Ice Save Percentage" while a player was on the ice. It regresses very heavily to the mean in the long run: a team or player well above 1,000 has generally played in good luck and should expect to drop going forward and vice-versa.
The thought is that players with incredibly high PDOs — your all-world talents potentially withstanding — will regress to a natural mean of 1,000. On the other end of the spectrum, players with extremely poor PDO rates will either positively regress to a natural mean of 1,000, or simply wash out of the league due to a poor skill/ talent level.
Remarkably enough, not one single player from last year's roster (30+ GP qualified) maintained a 1,000 PDO or better. The on-ice shooting percentages were an abomination, and the on-ice save percentage wasn't any better. Players like Alfredsson (967), Smith (963) and Filip Kuba (949) never had a chance — shots didn't go in when they took them and when they defended them, they found back of the net regularly.
Part of it has to do with players collectively struggling on a bad team, but the fact that not one single player actually had a fair season is pretty terrifying. The two worst teams last season (Columbus and Edmonton) had 12 combined players at or north of a 1,000 PDO, and both teams were statistically worse than Ottawa in 2011-12.
Again, the driving point here is that even if some of the players on Ottawa's past roster may eventually wash out of the league, the majority were bona fide NHLers and mean regression — in a positive manner — was inevitably going to take place in the immediate future. In 2011-12, we saw major jumps on a player-to-player basis — some saw their on-ice shooting percentages skyrocket, and others saw their on-ice save percentages take flight.
Where the previous team had zero average PDO players in 2010-11, the newer brand essentially had 15, with only one returning player — Chris Neil — coming up on the unlucky end of things. With the link provided, note the wild positive movement in shooting percentages for and strong negative turn in shooting percentages against.
As a unit, Ottawa's normalized statistical output simply led to more wins. In graph form, the transition from incredibly unlucky ('10-11) to slightly lucky ('11-12) is almost surreal.
Thirteen (13) of 19 players ('10-11: three of 16) possessed an on-ice shooting percentage of 8.00 per cent or greater, and 12 of 19 players ('10-11: two of 16) possessed an on-ice save percentage of .917 or greater.
With so many players creating more chances and scoring more goals, Ottawa finished the season as the fourth-best offensive team in the NHL. Its up-tempo style of play often created high-quality scoring chances against, but the team still improved defensively on its goals-against margins, moving from a 2.99 GAA to 2.88 GAA in just one season.
Bring it all togetherand you have an impressive 82-game run that led to a playoff berth. Some of the lustre was undoubtedly created by the preceding team's inability to gain measured success, but at the highest level of professional competition in a sport where volatility and parity reigns supreme, these kinds of statistical anomalies are fairly commonplace.
In summation: The next time you're asked about how the Ottawa Senators managed to accomplish such a feat, respond with the following: "Paul MacLean, puck possession, and mean regression."
Just don't blame me if all you receive is a blank stare.
Thanks for reading!
Special thanks to Nice Time On Ice, QuantHockey, Gabriel Desjardins' Behind the Net and Puck Prospectus for providing statistics for this column