Thursday, March 15, 2012

Breaking Down the Bruins Breakdown

What seems like years have passed since the Bruins were a winning team. In that time, I - like most other fans, I'm sure - have evaluated every possible explanation for this nightmare. How did it go from being so perfect, to being almost irreparable?

Do the answers lie in the numbers on the stat sheets? The video replays? The players themselves? Or is it something we can't see - something wrong inside the locker room? A problem that can be traced back in time? Or something even deeper? Or maybe something far simpler and more obvious? If the answer is a little bit of everything, doesn't it seem more impossible to overcome? Searching for answers doesn't yield solutions. We can only sit and helplessly wait for the results.

When rendered helpless, I like to pretend I can put myself to good use. This is my feeble attempt to make sense of the situation and shed some light, as dim as it may be. I go into uncharted territories, personally, by exploring advanced statistics and the many different types, uses, and even practices in the NHL.



Where to begin in assessing the current state of the Bruins? One point of departure is numbers. But advanced statistics used more and more around the NHL and among it's fans has reminded us just how unreliable numbers are in this game. Not just unreliable, but largely inaccurate. The first example of this is the percentage columns of the stat sheets, such as shooting % and save %, along with the frequently misleading and meaningless +/-. Even basic indicators like shots, goals, and saves are arguably more deceiving than informative. Hockey is a type of game that cannot be even remotely accurately depicted with any numbers, let alone simple ones. So when we look to characterize complex situations or answer multifaceted questions such as 'why is my team so terrible?' or 'why are the hockey gods out to get me?', we are just as clueless as we were before we pulled up the score sheet.

The statistics were incredibly fun to analyse when the Bruins were blowing every category out of the water back in the glory days of November and December, but now those numbers further spread panic and uncertainty. And the truth of the matter is that most traditional statistics are meaningless and deceiving anyway, whether they are good or bad, so rambling off the numbers is an exercise in futility. But for those of us who refuse to accept that, newly developed, advanced statistics are becoming available, providing new insight despite being incredibly difficult to comprehend for the mathematically challenged. An overview of the basic ideas, confusing as it may be, was enough to convince me of their considerable merit.

Shooting % is one statistic that is incredibly misleading. Several factors must be taken into consideration, many of which serve to provide meaningful explanations of the NHL ecosystem and the endless stream of cause and effect relationships. For example, "shots on net" as the NHL currently counts it, is a misnomer. A system of considering all shots attempted, including blocked shots, missed shots, and shots that hit the goal posts is far more telling; in this case, tallying the "shots attempted" differential is a statistical practice called Corsi. The mothership of new NHL stat machines, Behind the Net, says that this particular stat counter "is the single-best predictor we have of future team winning percentage". Another statistical calculator, called PDO (save % + shooting %) "captures several complex and powerful concepts", mainly the significance of shooting % and save % as both being "primarily luck-driven".
To calculate PDO, you simply add together a team’s even-strength shooting percentage and even-strength save percentage, multiplied by 1000 to give us a nice looking number to work with. For the entire NHL, the PDO number must equal 1000, because every shot is either a goal or a save. (Source)
PDO is quite intriguing because of how broadly and accurately it can be applied. Additional explanation from Behind the Net describes the stat as quite important and in this case, quite applicable to the situation the Bruins were in both at the time of winning streaks and current losing streaks. If the math doesn't quit click, it's still a good stat to look at.
In the long-run, PDO regresses very heavily to 1000; rare is the team that is able to stay consistently above this level, though every year, people claim that some team can out-finish their opponents while failing to out-shoot them. Examples include Colorado in 2009-10, Dallas in 2010-11 and Minnesota in 2011-12 - all three teams came crashing down to earth way before the season ended. When you want to know how a team is going to do going forward, out-shooting is way more important than out-finishing.
One Pens Blog adds some insight to some of the fascinating findings compiled at mc79hockey:
The Top 20 teams saw an average PDO decrease of 2.6% over the following three quarters of hockey, while the bottom 20 teams saw a 2.8% increase in PDO over the same time period. More specifically, 19 of the 20 top teams saw a drop in PDO and all but two of the bottom 20 teams saw a rise in PDO. The correlation wouldn't be as strong if you added the rest of the teams because you'd be adding randomness when looking at the average teams. But the two points to take away from this are that PDO is a great predictor of future performance, and that it works even better for those at the extremes.
When a team comes roaring out of the gate at the start of the season, this stat in particular exposes those who are premature in crowning Stanley Cup Champions. As Backhand Shelf puts it:
You would assume that most people would know by now not to get overenthusiastic about an early season scoring streak, but then you read about how Kessel is “on pace to put up Gretzky-like numbers” and you are proven wrong.
Like individual numbers, individual events do not explain complicated series of events. As the elaborate new statistical data collected and calculated exhibits, it is not only impossible but also foolish to attempt to confine the understanding of hockey and all it's complexities to any simple and individual concept. This is why the idea of the Tim Thomas White House "incident" is so ridiculously inappropriate in trying to assess anything that may have gone wrong or is going wrong with the Bruins. It's easy to pin troubles on a highly publicized drama fest like that because it serves as a mental timeline for the general period of "decline" in the Bruins play rather than look at the solid data.

But the "decline" in mind was more like a return to normalcy after a universally unbalanced explosion of offensive production and defensive shutdown in November and December. Although the Bruins have fallen far from grace, the beginning of the "end" - the end of winning streak - was already happening before we realized it. Things have a way of evening themselves out like that, and statistics that seem too good to be true usually are too good to be true. This variation applies to teams and players alike, and the following explanation can be applied to teams as a whole as well:
The highest PDO amongst players who played at least 60 games last season was 1062, while the lowest was 934...Several of the players who have jumped out to a hot start, including Phil Kessel, have PDO numbers that are much, much higher than that. That’s not a criticism, by any means. It’s just a way of understanding how a scoring streak happens. The reasons for a high PDO are often attributed to luck, but they can also be a result of choices made on the ice, and these factors can play into each other. A player who faces some bad luck, for instance, and hits a couple posts or has a bad bounce go against him may start second-guessing his instincts and try to force shots from low percentage areas of the ice, causing his shooting percentage to further plummet. Or a player who sees almost every shot slip past the goaltender may start handling the puck with more confidence and head to the front of the net with more authority, trusting that the puck will find its way to the right spot on the ice. (Source)
The combination of luck and its direct effect on decision-making is too obscure a reality to simply accept. Yet, we so urgently cling to and rely on and search for answers in these misleading stats. These stats, in the Bruins early season instance, "made the team look just a little bit better than it really was". While the Bruins could be a superior team throughout a season, the sad truth was that it could never have been realistic to expect them to sustain those numbers all season:
Since the beginning of the “Behind The Net” era...only two teams, the 2009 Pittsburgh Penguins and the 2010 Washington Capitals, have finished with shooting percentages of higher than 10%.

Boston’s 9.8% shooting rate by the time they hit the White House break wasn’t just due to come down: it already was dropping. A wild month of November and December saw the team shoot 11.4% and we were in turn fooled by their wild success, putting up six goals on regular occurrences and blasting through the competition, blinding us to the fact that they weren’t outshooting their opponents as drastically as the scoreboard decreed they had.

After another game in early January, the Bruins’ PDO...hit a season high of 105%, well above a sustainable range and nothing we’ve ever seen before in hockey. (Source)
In assessing the individual performances of the "underachieving" Bruins, many other stats explored at Behind the Net and other similar sites show that maybe it's not as bad as we think. NHL games are more accurately portrayed on paper with the adjustments to +/-, for example, also bring inflated statistics such as Tyler Seguin's league-leading +/- into perspective.

Unstoppably, player's results will gravitate towards the Corsi numbers as the season goes on. A few good bounces here and there can make the early stats misleading. To pick on an old favourite, Lupul is EV+/- +3, and Corsi -41. The drop is inevitable, just is. (Source)
This reference to an outdated Bruins team still holds the same weight:
One of the weaknesses of traditional +/- is that it tends to favor players on good teams - Bruins Ryder, Blake Wheeler, David Krejci, Phil Kessel and Marc Savard were all in the top ten in the league, which is not surprising given that Boston was the highest-scoring team at 5-on-5 in the NHL. At the same time, it penalizes players on bad teams - six New York Islanders were in the bottom 10.
This also brings David Krejci and his first-line-mates unusually negative ratings on an overly positive-rated team into a better perspective. In line matching, many of these complicated statistics break down the reasoning behind certain player combinations or roles (although, as each link indicates, ascribing a role to a player may produce numbers that do not necessarily reflect the talents of that player overall). While the Krejci line is the top scoring line for the Bruins (at least they are now - for a while it was the Bergeron line), they still rank low or negative in the +/- because their role as a line is purely offensive. In contrast, the Bergeron line often has the main purpose of shutting down the opposition, keeping the minus penalty minuscule if they are good at shutting them down (which they are). The opposing team will put their Bergeron-equivalent checking line against the Bruins top line of Krejci, Lucic, and whoever else rounds the trio out. In many instances where the opposition has been successful in keeping the Krejci line from scoring, the minus is magnified for the Bruins top line. Furthermore, top power play scorers like Lucic don't receive that plus point for power play points - and the Krejci line sees a larger proportion of power play time than any other line.

Measuring the Quality of the Competition as well as Zone Starts are additional stats worth looking into as they properly adjust +/- perspectives or display detailed and meaningful information. The aforementioned Corsi number is the most reliable and telling number. Essentially, scoring chances are more accurately displayed in terms of the Corsi as opposed to the actual tallied number of shots and scoring chances; it is even more telling of how a game is played than the goals scored in the game. The Corsi number is also worthwhile to get acquainted with - even NHL GMs are employing professionals and applying the data in carrying out the jobs of trades and contracts, for example.
“I’ve seen people use Corsi [Numbers] to make trades,” Desjardins said. “I’ll put it that way.” (Source).
Ironically, the Bruins recent acquisition, Greg Zanon, was crowned with the worst Corsi in this article written over the summer, and details how it explains why he is the worst player in the league. The claims made in that article are refuted here, however, further delving into the nature of the Corsi in relationship to environment, role, and the countless other variables. The argument presented in this specific case is intriguing because of it's relation to the Bruins and exposing the nature of advanced stats in general. Furthermore, Bruins GM Chiarelli is said to use some sort of advanced statistics without giving specifics.
The Bruins characterize quality of chances for and chances against and come up with different kinds of plus/minus numbers, Chiarelli said.
More advanced advanced statistics seem to be coming out, too! And certain GMs like Chiarelli turn to alternate/advanced statistics because they acknowledge how unreliable the NHL statistics are. A few articles compare the stance taken by Chiarelli as opposed to Leafs GM Brian Burke, who considers these statistics and their roots in the practices chronicled in 'Moneyball' to be ridiculous. I think this highlights why Burke is kind of ridiculous, and is quite interesting in assessing why or why not one GM has a successful team and one GM does not. The likelihood that Chiarelli's utilization of advanced statistical analysis - in whatever manner he may formulate it - is applied actively can be said to be a positive. For example:
What, for example, led the Bruins to think that Rich Peverley could help their team better than Blake Wheeler, or to believe that they could get significant player value out of Chris Kelly? On draft day 2010, the Bruins pulled off the biggest trade of their upcoming season, shipping Dennis Wideman and a pair of picks for Nathan Horton and Gregory Campbell. While Campbell is a tough minutes checker, he also was counted among last year’s 10-goal scorers. Horton filled in as one of the Bruins’ most key offensive pieces (and you can see how much the team misses him as he sits out with a concussion) and scored 20 even strength goals last year at a salary cap cost of just $4M.
Even if Chiarelli's intelligence use of numbers doesn't seem to be particularly helpful to the Bruins current state, there's a strong argument that he is, in fact, one of the smarter GMs out there. I know I think he is an evil genius. There is a lot more exploration into these advanced stats, but I'll stick with the basic research I've accumulated so far to lay the basic foundation for now...

The 10-Part FAQ also explores the value of numbers assessing a +/- in drawing/taking penalties, shooting accuracy in relation to distance from the net and thus defensemen and forwards variations in scoring, as well as positional scoring differentials and averages on the power play and even strength, quality of scoring chances, and quality of teammates. Any or all of these forms of statistical analyses can be applied to any team, such as the Bruins, and be a valuable tool in trying to understand trends of players or the team. I picked up some interesting information and perspective that I would never have considered otherwise, and certainly would not have picked up from any official NHL data.

In this report, the author takes a different approach to traditional statistics - accepting them and elaborating on them, but coming to many similar and insightful conclusions. One such conclusion supports the above claims about the ebb and flow of the season, a team, and its players and the inevitability of change given the polarity of the situations - in this case, the sharp rise and then fall of shooting percentages, and the natural - if not random - variations from year to year, are more commonplace than anything:
The Bruins shooting percentage went from a league best in 2009 to a league worst in 2010 to slightly above average in 2011. Washington went from above average in 2009 to league best in 2010 to below average in 2011
These numbers don't promise a great team, or an awful team; these statistical anomalies really don't paint an accurate picture of any team. In any given category shown in NHL statistics, only a few teams or players have numbers that stand out, for whatever bizarre reasons. But we misinterpret those numbers to have more weight than they actually do. It's easy to get worked up over something being great, or on the cusp of being great, when there seems to be factual evidence supporting said greatness. In retrospect, we should have seen the nosedive coming all along - and perhaps some of us did, but didn't want to accept it, because they were so great.
Not to say that the Bruins downward spiral in recent weeks was inevitable to the extremes we see now, or even can be understood or explained in this context, but perhaps the inertia of moving in a direction other than straight to the top was enough to rattle the team's psyche or the universe's distribution of karma. Simply put:
Had Tim Thomas gone to the White House, we might have looked at the team’s 4-3 loss to the Vancouver Canucks as the time that all began to fall apart for the team.
Now, as I continue to look into the confusing mathematical maze of these new statistical concepts, there is also evidence that whatever is going on right now is bound to pass just as likely as the good times. For example, above links and quotes talk about how extremely inaccurate percentages are and how shot volume is far more important in the long-term for success rather than goals scored in the short-term. In many of the Bruins recent losses, they have out-shot their opponents by a wide margin. Sure, it has hurt the save % and goals against averages, as well as the egos and scoreboards, but like the illusion of the Bruins early-season dominance, perhaps this rock-bottom scenery surrounding our team is a sort of illusion as well. The relatively injury-free Bruins of last season are experiencing what seems like a massacre of injuries lately but what adds up to only a league average of total man-games lost. More compelling is the notion put forth in Alan Ryder of HockeyAnalytics.com's 2011 NHL Review:
Wins are about 94% predicted by goals for and against, marginal goals totals or goal differentials. When teams like Phoenix, Anaheim and Tampa Bay win in spite of a low marginal goal performance they are either very skilled at winning close games or just plain lucky. “Hockey people” will tell you that winning close games is about character. Historical analysis suggests that this is mostly luck. I would not completely rule out some intangible, but nobody has found it yet. Lucky teams tend to regress the following season (and vice versa). 
Ryder makes many fascinating points in this report that apply just as thoroughly and logically to current situations as they do to the statistical ones explained in the findings. One of his most emphatic beliefs is that of the importance of goaltending and talent of Tim Thomas. He explains the idea that strong goaltending - phenomenally fantastic goaltending in the case of the Bruins, especially earlier this season - results in exchanging defensive-mindedness for offensive pursuits. Ryder reported that last season, the Bruins previous-season habit of keeping shots against relatively low, the 2010-11 season saw the Bruins abandon shot prevention - and other healthy defensive habits - because of that inflated confidence in goaltending. He explains the Bruins 29th-ranked shot prevention by saying "Teams tend to open up (trade defense for offense) in front of good goaltending". Nobody could possibly expect Thomas to be that good that consistently for that long, could they?

Although Tim Thomas' record-breaking year last year would lead you to believe that the Bruins were a shut-down team successfully implementing Julien's "defensive style system", Boston's defensemen last year were actually increasingly sloppy which resulted in over 20 unnecessary - or marginal - goals that resulted from these errors. But it is easy to understand how they could make these mistakes:
There is little doubt that greater confidence in goaltending leads to a more open approach to the game and, by implication, less defense...Boston won a Stanley Cup with a defense that was as productive as that of Edmonton (don't misunderstand my comment – the Bruins chose to invest skating energy in offense because of their goaltending).
The Bruins defense caused 24 "marginal" goals, as Ryder calls them, last year, but the Bruins goaltending gave up the fewest "marginal" goals of any goalies in the league. Truthfully, Thomas was there to bail them out whenever they needed it. Goaltending is an immeasurable piece to a winning hockey team, and the rarity of true, quality goaltenders in the league is far less than numbers and fans would have you believe. The truth is, years like last year when Tim Thomas reminded everybody how sensational goaltending could lead to such a thrilling Championship come and go - just like every other trend in hockey. To continue to quote Ryder's findings, he makes the incredibly compelling point that:
In 2010 “hockey people” concluded that goaltending was over-rated. This is what happens when two teams with no-name goaltenders compete for the Stanley Cup [Flyers and Blackhawks]. The 2011 Stanley Cup will go down in history as one of those goaltending years, of which there are many. It may surprise you that over 40% of all Conn Smythe Trophies have been won by goalies. 
Fellow Bruins fan Andrea (@androxin - who was also kind enough to link me to the above article from the blog Backhand Shelf) asked an interesting question after reading that quote: have the Bruins lost faith in Thomas?

Now a departure from the statistical investigation of the Bruins troubles, and into a philosophical, psychological analysis (or bunch of jibberish, you decide). Have the Bruins lost faith? They certainly don't have the confidence they used to have. Instead of taking chances offensively in recent months, the Bruins seem caught in a lingering hesitation to make the decision between defense and offense. Decreased confidence in your goaltending would explain some of the offensive disasters of late, such as frequently getting shutout and a complete inability to score on mediocre defense and goaltending. A crisis of confidence in their own defensive capabilities certainly is seen in all 3 zones.

Defensemen seem increasingly unwilling to - even afraid to - take the shot, especially on the power play; when they do, it seems to always gravitate to the defending teams shot blocker. The defense is caught in limbo when making the decision to pinch and more often than not put themselves in danger due to that moment of uncertainty. Turnovers and confusion in the neutral zone show a lack of execution, inability to communicate, and deterioration of team chemistry overall. Forwards seem to take charge individually in many instances, charging in on a one-man wrecking crew only to be stripped of the puck before he can fully cross the blue line. Passes are just a little off. And when it comes to those forwards getting back on defensive coverage, the confusion multiplies as the defensemen are already out of position and out of whack.

Recently, these trends of breaking down seem to happen after letting in a weak goal - or letting in a goal at all in the aftermath of early-game domination in the offensive zone that miraculously doesn't result in a goal for the Bruins.  They seem to enter the game with the mindset that they must go out and take the game by the throat early, generate those chances and desperately bury them as soon as possible because they are walking on eggshells the moment the puck is on the stick of an opponent in the defensive zone.

Such a trend of outshooting opponents but getting outscored nonetheless was discussed above as a fluky trend that is not indicative of anything in the long-term; as long as they continue to fire off a high volume of shots, they will find success...soon. But every time we think we've seen them grasp that concept of "success", they fall even flatter the next time out. Over-thinking things, as they say, doesn't help; and as they fall further from "their game", it seems more impossible to grasp exactly what that "game" was. But, just as an aforementioned statement about the nature of falling into patterns of streaks by way of chance and it's subsequent effect on decision-making, there's a lot to take into consideration that we can't possibly know. So we speculate.

Statistics and trends aside, the Bruins are capable of being one of the best teams in the league. Whatever psychological, mental, or even physical and physiological barriers are currently inhibiting their success may make them seem permanently unstable are, just as everything else, temporary. Maybe their faith is shaken - in their goaltending, in their teammates, in themselves - but slipping into the second-half "slump" was a gradual decline, so naturally the team's restoration will be a gradual one as well. Perhaps we've already seen signs of it, but we just don't realize it yet. Just like we were deceived by the illusion of being unbeatable at the start of the season, we could just as easily be disillusioned into thinking we are doomed for the rest of the season.

We mark our lives in terms of individual events or milestones - "the White House incident", "the Stanley Cup Championship", "the Horton injury", "the Boychuk contract extension", "10-straight games for Thomas" or "the 22 game unbeaten streak". Even broad events over time, such as the 2-month Cup run, unbeaten streak, or back-to-back winless streak are compartmentalized and viewed in our minds in terms of one single perspective rather than the complex, crazy conquests and intricate details that actually took place. To say "since the Horton injury" or "since the White House incident" and then finish with some sort of measure of time such as "the Bruins have gone 0-50-0" is to already misinterpret the gravity of reality. The truth of the situation is that no single, isolated event was the trigger point for the Bruins fall from grace, and no single, identifiable event will end it.

No comments:

Post a Comment