There are quite a few major inventions and discoveries that have been accidental, or at least serendipitous. I won’t put this into that realm, but I believe this does have a significant impact. And it simply happened to come about almost by accident.
I’ve been messing around with OPP, OPK, and OST over the past weeks. I believe these to be a more true indicator of the effectiveness of a team’s overall special teams (hence OST) than the traditional percentages, but that’s another story. Anyway, I was looking through the overall leaderboard from 1992-93 through 2016-17 when I noticed an interesting phenomenon. It’s one that I’ve seen before and speculated about, which is that nearly all leaguewide scoring in the NHL over time is most drastically affected by the number of power plays.
On the same spreadsheet that contains my OPP/OPK/OST data, I created a couple of additional columns that filtered out all special teams goals in either direction and left just even-strength (and shootout, for 2005-06 and later) goals. In an unsurprising twist, this flattened down the number of overall goals in both directions. Then I simply divided the number of even strength goals both for and against by the number of games to get a per-game average. And then, because why the hell not, I decided to then look at just the even-strength plus-minus, sorted from best to worst. There have been 711 team-seasons played since 1991-92.
The results were actually unexpected, both in correlation strength and in how drastic the whole thing is.
(NOTE: I am not claiming credit for discovering this metric. I’m quite certain that someone else has done so, and may have used it for various things. My contribution is contributing the data, for noting how strongly it correlates to playoff appearances, and then my ultimate conclusions. The link to the spreadsheet with data is at the bottom of this page.)
For starters, no team in the last 25 seasons has averaged 3 even-strength goals per game (ES/G). Think of all the great offensive juggernauts that we’ve seen in that time, plus two lockout-shortened seasons (which show more extreme ratios than full seasons), and none of them average 3 ES/G. The top teams were actually the 1992-93 Vancouver Canucks and the 1993-94 Detroit Red Wings, both of whom averaged 2.964 ES/G. The two worst were the 1997-98 Tampa Bay Lightning (to the shock of no one) and the 2002-03 Carolina Hurricanes, both of whom averaged 1.305. The clustering by year is largely unsurprising; the top teams in this list are mostly 1992-93, 1993-94, 2005-06, and 2006-07. The bottom teams are mostly from 2000-01 to 2003-04.
For ES goals against, the opposite holds true: the top teams in keeping the puck out of their net are mostly from 2000-01 to 2003-04, with the worst mostly being from the mid-1990s and the early-2000s. The best team in choking off even strength offense was 2002-03 Philadelphia, who allowed just 1.317 ESA/G. Only four teams of 711 allowed more than three ESA/G, those being the 1991-92 San Jose Sharks, 1992-93 San Jose, 1992-93 Ottawa Senators, and Ottawa again in 1993-94.
Obviously the time frames are heavily influencing the scattering, and I really didn’t feel like needlessly complicating things by determining an overall 25-year average of goals per game, normalizing to that, and coming up with “adjusted even strength goals” or some other such slop. I just re-sorted everything by overall ES+/-, which is nothing more than even strength goals per game minus even strength goals against per game. I figured that this, by itself, would automatically normalize the changing conditions of the game over time: the wide-open style of the early-1990s and mid-2000s would have more goals for but also goals against, and the more choked style of points on either side would have fewer goals for but also fewer goals against. And in the end, it would all wash.
Remember, there are 711 team-seasons in these 25 seasons. Of these:
- The top 200 team-seasons of this date range begin at only +0.2439. Of these, 199 made the playoffs. That’s 99.5%.
- The bottom 100, actually the bottom 103, starts at -0.4146 and descends from there. Of these bottom 103 teams, 102 missed the playoffs. That’s 99.03%.
The lucky and unlucky dogs? Lucky was the 1994-95 Toronto Maple Leafs, who came in at #615 overall and -0.41667 ESA/G, but had a 102.246 OST to help their case. They finished with 50 points in the 48-game shortened season.
Unlucky was the 2003-04 Edmonton Oilers, who rank tied for #130 overall with a +0.3659, but had a dismal 95.508 OST. Their 89 points in 82 games wasn’t enough to put them into the postseason.
At first glance, it appears that much like PDO, OST, and every other metric, all bets are off come playoff time. To use the 2016-17 season, the top five teams in ES+/- were Washington, Minnesota, Columbus, Chicago, and Pittsburgh. To this point, two have been bounced out of the playoffs already and a third one is down 3-1.
The three best even-strength teams of the last 25 years, by the way, are the 1993-94 New Jersey Devils (+1.119), 1995-96 Detroit Red Wings (+1.012), and 1994-95 St. Louis Blues (+1.000). The three worst are, to the surprise of no one, the 1992-93 San Jose Sharks (-1.750), 1993-94 Ottawa Senators (-1.714), and 1992-93 Ottawa Senators (-1.667).
Of the top 250 teams from that 25-year span (actually 255, since there’s an 11-way tie from #245-255), 249 of them made the playoffs. And those that didn’t, by and large, suffered from dreadful special teams. The six teams are:
- 2003-04 Edmonton Oilers (t-#130) – 89 points, +.3659 ES/G, 95.508 OST
- 1992-93 Philadelphia Flyers (t-#207) – 83 points, +.2262 ES/G, 96.286 OST
- 2005-06 Florida Panthers (t-#211) – 85 points, +.2195 ES/G, 95.677 OST
- 2001-02 Edmonton Oilers (t-#217) – 92 points, +.2073 ES/G, 102.294 OST
- 2006-07 Toronto Maple Leafs (t-#228) – 91 points, +.1951 ES/G, 94.149 OST
- 2006-07 Colorado Avalanche (t-#245) – 95 points, +.1707 ES/G, 100.870 OST
What’s perhaps most surprising is that the bar is extremely low. Every single team in the last 25 years that has been more than +0.5 ES/G has made the playoffs, and most of these rank among the overall best teams within a given season. With the exception of those 2003-04 Oilers, every team that’s been more than +0.25 ES/G (197 teams) has made the playoffs. And on the flip side, every single team that’s worse than -0.25 (of which there are 179 teams) has missed the playoffs with six exceptions. To be on either side of that ledger means that all one has to do is either score or allow one more even-strength goal than their opponent in a four-game stretch across an 82-game season. That’s a pretty low bar to clear.
By the way, of those six teams out of the bottom 179, we can almost see the opposite scenario unfold, with a poor ES/G plus-minus but mostly excellent special teams:
- 1995-96 St. Louis Blues (#538) – 80 points, -0.2927 ES/G, 100.002 OST
- 2011-12 Florida Panthers (#547) – 94 points, -0.3171 ES/G, 97.605 OST
- 1993-94 St. Louis Blues (#562) – 91 points, -0.3452 ES/G, 102.710 OST
- 2000-01 Carolina Hurricanes (#570) – 88 points, -0.3659 ES/G, 103.549 OST
- 1994-95 San Jose Sharks (#576) – 42 points [48-game season], -0.375 ES/G, 93.494 OST
- 1994-95 Toronto Maple Leafs (#596) – 50 points [48-game season], -0.4167 ES/G, 102.246 OST
For those who don’t remember the NHL of 1994-95 and 1995-96, the Western Conference went through a lull where multiple sub-.500 teams made the playoffs because the bottom of the conference was atrocious. 1994-95 had only five Western teams with winning record, plus Vancouver at exactly .500. 1995-96 was worse, with 4th-seeded St. Louis being two games under .500 on the season; all told, 10/13 Western teams were under .500 on the year.
The one that stands out is the 2011-12 Florida Panthers, who had both a bad record at even strength and subpar special teams, and yet finished with 94 points and won the Southeast Division. The Southeast was bad, but it wasn’t that bad.
Usefulness in Predicting the Playoffs
In prior analytics work, I’ve noted that there tends to be a clustering of playoff teams toward the top of a given metric, with non-playoff teams toward the bottom. The almost unbelievable extent of clustering with even strength +/- to playoff and non-playoff teams means that we may be able to get some amount of predictive usefulness out of it for the playoffs. PDO, shots taken/against, OST, and other metrics have all fallen completely flat when it comes to being able to accurately predict the postseason. ES+/- has by far the strongest correlation to playoff appearances, and if metrics can be used to predict future events, this should be the one that bears this out.
Let’s take a look, year-by-year, from 1991-92 to the current year:
1991-92 – Pittsburgh (8th leaguewide) defeated Chicago (9th) in the Stanley Cup Final. Of the top 8 teams, only Edmonton (5th) and Pittsburgh made it out of the second round.
1992-93 – Montreal (3rd leaguewide) defeated Los Angeles (15th) in the Stanley Cup Final; of the top 8 teams, Montreal was the only one to make it out of the second round at all. It’s also worth noting that six of those eight teams had OSTs of over 102.0%, with three of them over a staggering 106.0%.
1993-94 – NY Rangers (6th leaguewide) defeated Vancouver (17th) in the Stanley Cup Final. Of the top 8 teams, only the Rangers, Maple Leafs (5th), and Devils (#1 in the last 25 years in this category, by the way) made it out of the second round.
1994-95 – New Jersey (8th leaguewide) defeated Detroit (2nd) in the Stanley Cup Final. Of the top 8 teams, all four conference finalists came from there. One notable series saw St. Louis (#3 in the last 25 years) lose to Toronto (#596, by far the worst playoff team ever in this metric) in the first round.
1995-96 – Colorado (2nd leaguewide) defeated Florida (13th) in the Stanley Cup FInal. Conference finalists Pittsburgh (6th) and Detroit (1st, #2 overall in the last 25 years) were also toward the top.
1996-97 – Detroit (4th leaguewide) defeated Philadelphia (2nd) in the Stanley Cup Final. The other conference finalists were Colorado (8th) and the Rangers (9th).
1997-98 – Detroit (3rd leaguewide) defeated Washington (16th) in the Stanley Cup Final. The other conference finalists were Dallas (2nd) and Buffalo (4th).
1998-99 – Dallas (5th leaguewide) defeated Buffalo (3rd) in the Stanley Cup Final. Other conference finalists were Colorado (6th) and Toronto (2nd).
1999-00 – New Jersey (3rd) defeated Dallas (14th) in the Stanley Cup Final. Other conference finalists were Colorado (8th) and Philadelphia (6th).
2000-01 – Colorado (2nd leaguewide) defeated New Jersey (1st) in the Stanley Cup Final, with other conference finalists St. Louis (6th) and Pittsburgh (8th).
2001-02 – Detroit (3rd leaguewide) defeated Carolina (18th, behind three non-playoff teams) in the Stanley Cup Final. Other conference finalists were Colorado (11th) and Toronto (4th).
2002-03 – New Jersey (5th leaguewide) defeated Anaheim (18th, behind three non-playoff teams) in the Stanley Cup Final. Other conference finalists were Minnesota (12th) and Ottawa (3rd).
2003-04 – Tampa Bay (8th) defeated Calgary (15th) in the Stanley Cup Final. Other conference finalists were Philadelphia (6th) and San Jose (11th).
I want to pause here for a second, because you’ve undoubtedly noticed that the final four teams are mostly coming from the top-eight in a given season, but not always. This would seem to bear out that yes, there’s still a correlation between ES+/- and playoff success to a point. So far from 1991-92 to 2003-04, we have seen two teams that finished #1 in this category within that season even make it to a conference final; one lost there, and one lost in the Stanley Cup Final.
Of 52 conference finalists in this 13-year span, we have seen more teams that finished below non-playoff squads make it to the Cup Final than #1 overall teams. We’ve seen three teams ranked 8th in that season win Cups, and not one single #1 overall team.
It’s also worth pointing out that during this 13 seasons, the number of power plays plummeted while the amount of even-strength time surged. One would expect to find a larger scattering in the earlier years between a team’s rank in a season (since we’re measuring even-strength quality) and their playoff success and one that stratifies much more toward the top in the later years (as even-strength quality meets much more even-strength opportunity). Yet we find the opposite to be true.
Starting in 2005-06, ties went away for good and were replaced by shootouts. A shootout win is credited also as +1 goal in the goals for category, and is listed also as an even-strength goal. These were not removed or otherwise adjusted for when it comes to compiling this metric.
2005-06 – Carolina (8th) defeated Edmonton (21st, behind six non-playoff teams) in the Stanley Cup Final. Other conference finalists were Anaheim (5th) and Buffalo (12th).
2006-07 – Anaheim (7th) defeated Ottawa (2nd) in the Stanley Cup Final; other conference finalists were Detroit (3rd) and Buffalo (1st).
2007-08 – Detroit (1st) defeated Pittsburgh (2nd) in the Stanley Cup Final; other conference finalists were Dallas (6th) and Philadelphia (22nd, 16th among playoff teams)
2008-09 – Pittsburgh (5th) defeated Detroit (4th) in the Stanley Cup Final; other conference finalists were Chicago (2nd) and Carolina (15th).
2009-10 – Chicago (2nd) defeated Philadelphia (12th) in the Stanley Cup Final; other conference finalists were Montreal (16th) and San Jose (4th).
2010-11 – Boston (1st) defeated Vancouver (2nd) in the Stanley Cup Final; other conference finalists were San Jose (6th) and Tampa Bay (15th).
2011-12 – Los Angeles (16th) defeated New Jersey (15th) in the Stanley Cup Final; other conference finalists were the Rangers (6th) and Phoenix (10th).
2012-13 – Chicago (1st) defeated Boston (4th) in the Stanley Cup Final; other conference finalists were Pittsburgh (2nd) and Los Angeles (10th).
2013-14 – Los Angeles (5th) defeated the Rangers (12th) in the Stanley Cup Final; other conference finalists were Montreal (13th) and Chicago (4th).
2014-15 – Chicago (4th) defeated Tampa Bay (2nd) in the Stanley Cup Final; other conference finalists were Anaheim (11th) and the Rangers (1st).
2015-16 – Pittsburgh (5th) defeated San Jose (8th) in the Stanley Cup Final; other conference finalists were St. Louis (9th) and Tampa Bay (7th).
2016-17 – Still in progress, but #2 Minnesota, #3 Columbus, and #5 Chicago all lost in the first round while winning a total of two games between them. The rest of the top 8 are #1 Washington, #4 Pittsburgh, #6 New York (Rangers), #7 Montreal, and #8 Edmonton.
So there you have it: 25 seasons worth of playoffs and even-strength plus-minus, with a total of three top-ranked teams in a given season actually winning a Stanley Cup, and only seven top-ranked teams making it out of the second round at all. That’s a dismal record.
This seems to bear out pretty conclusively what I’ve already said: analytics do not carry any meaningful ability to predict playoff performance. That is a simple, 100% verifiable fact that too many hacks and Yahoos seem to want to ignore.
It doesn’t matter which way you want to predict it and based on what. The spreadsheet I compiled all of this with is attached to all of the OST stuff, so I can glance one column to the right and check the ES+/- against OST. It makes no difference; a team that’s lower in both can knock off a team that’s higher in both, just as a higher team in both can destroy a lower one in both, just as a team that’s higher in one and way lower in another, just as one that’s high in both may miss the playoffs completely while one that’s well below league average in both may make it…you get the idea.
Here’s what I can tell you about ES+/-:
- It’s directly related to the overall quality of your team, as the fact that 199 of the top 200 single-season teams making the playoffs shows
- A team that outscores their opponents at even strength at a rate of 1 goal every 6 games will make the playoffs 249 times out of 255. (If we include teams at exactly this rate, it becomes 252/258 teams spanning 25 seasons. That’s a 97.67% rate.)
- None of it means a damn thing once the playoffs start
Can we go back further?
Actually, yes. I sorted all of this for the years of the 21-team NHL as well, and reached much the same conclusion. There were 252 team-seasons that were played during these 12 years (1979-80 through 1990-91), and things most definitely got scrambled as a result of 16/21 teams making the playoffs every year.
The pertinent numbers:
- The top 91 teams out of these 252 all made the playoffs, and after #92 (1987-88 Pittsburgh), the next 16 also made it.
- The bottom 22 all missed. Only 60 playoff misses took place in these 12 seasons; 42 of these were in the bottom 51 single-season teams in ES+/-.
- The lowest team that posted a .500 record was the 1979-80 St. Louis Blues, who slot in at #180 and were exactly .500. The lowest team to have a winning record was 1990-91 Washington, which was #168 and had 81 points in 80 games.
- The highest team that missed the playoffs and had below a .500 record was the 1986-87 Penguins, who rank #109 and finished with 72 points in 80 games.
Could this metric show a greater amount of playoff predictability than it did once expansion hit? Well….one problem is that this era included two years of a 1 v. 16, 2 v.15 format; this was then followed by the divisional format that saw the top four teams in each division play each other in the first two rounds. And since this included the years where the Norris Division was possibly the worst in hockey history, the conference finals almost always finished a team that would have been destroyed in the first round of any other format.
Regardless, by Stanley Cup Final matchups:
- 1979-80 – NY Islanders (#5) over Philadelphia (#1)
- 1980-81 – NY Islanders (#3) over Minnesota (#11)
- 1981-82 – NY Islanders (#2) over Vancouver (#9)
- 1982-83 – NY Islanders (#6) over Edmonton (#1)
- 1983-84 – Edmonton (#1) over NY Islanders (#2)
- 1984-85 – Edmonton (#2) over Philadelphia (#1)
- 1985-86 – Montreal (#5) over Calgary (#4)
- 1986-87 – Edmonton (#1) over Philadelphia (#2)
- 1987-88 – Edmonton (#3) over Boston (#6)
- 1988-89 – Calgary (#1) over Montreal (#2)
- 1989-90 – Edmonton (#8) over Boston (#3)
- 1990-91 – Pittsburgh (#7) over Minnesota (#14)
In twelve seasons, the #1 team won three Stanley Cups and lost in three more Cup Finals. This was despite a hideously unbalanced league that regularly saw multiple teams finished better than 20 games over .500 and multiple ones finishing more than 20 games under .500. Even toward the end of this time period, a year like 1988-89 saw Calgary and Montreal both finish with over 115 points while the next best (Washington) had 92; six teams had 66 points or less.
The conclusion is unmistakable (no, I didn’t say “unsustainable”): metrics which correlate to regular season success to a staggering extent do not carry the slightest bit of useful predictive ability.
There’s still more!
Because there’s no sense in simply stopping with the NHL/WHA merger in 1979, I went back further through the entire NHL expansion era. So from 1967-68 to 1978-79, the following was noted:
- There were 185 team-seasons that were played during these 12 years
- The top 53 team-seasons by ES+/- all made the playoffs
- The top four teams that missed the playoffs were all in the first three years of expansion (1967-68, 1968-69, and 1969-70). Two of them (1967-68 Toronto and 1969-70 Montreal) would have been the division champions of the West, while the other two teams (1968-69 Detroit and Chicago) would have been the #2 seed in the West. That these teams missed the playoffs is due to the way that the league was constructed, which involved stacking the deck for the expansion teams to allow four to make the playoffs and one to make it through to the Stanley Cup Final.
- The highest non-playoff team that wasn’t in these three years was 1974-75 Atlanta, who are 76th
- On the other side, the bottom 24 team-seasons from this era all missed the playoffs. The handful of teams near the bottom of this ranking that made the playoffs were below .500, and in most cases well below .500. The worst team in ES+/- to make the playoffs and be at or above .500 in the regular season is 1975-76 Chicago, ranked 119th (of 185) and two games over .500. No other team below them is within five games of hitting .500
But then again, I’ve focused more on whether this is useful for predicting the playoffs. So the Stanley Cup Finals, by year, featured:
- 1967-68 – Montreal (#1) over St. Louis (#3 in West, #9 overall)
- 1968-69 – Montreal (#1) over St. Louis (#1 in West, #3 overall)
- 1969-70 – Boston (#2) over St. Louis (#1 in West, #6 overall)
- 1970-71 – Montreal (#4) over Chicago (#2)
- 1971-72 – Boston (#2) over NY Rangers (#1)
- 1972-73 – Montreal (#1) over Chicago (#4)
- 1973-74 – Philadelphia (#2) over Boston (#1)
- 1974-75 – Philadelphia (#1) over Buffalo (#4)
- 1975-76 – Montreal (#2) over Philadelphia (#1)
- 1976-77 – Montreal (#1) over Boston (#4)
- 1977-78 – Montreal (#1) over Boston (#2)
- 1978-79 – Montreal (#2) over NY Rangers (#6)
The issue is that this entire era was not exactly a balanced one. We saw the entrance of expansion teams that would have trouble going .500 in a minor league, while the cream of the NHL crop mostly got there and stayed there. This is reflected in the standings, which show staggering differences between the best and worst teams, whether simply on a first-to-last basis or when stratified further.
It has been argued that the disappearance of extremes over time is a clear sign of increasing quality. The fastest hockey player in a given age group will be several orders of magnitude faster than the slowest in his league, or have a harder shot well beyond what the worst players can do; as the caliber of play increases, these extremes flatten down. In the NHL, the fastest shot isn’t that much faster than what the softest-shooting player can muster up, relatively speaking. This is part of why we see youth phenoms who never amount to anything as time goes on, and whatever advantage they may have held first diminishes and then is surpassed by other players.
In the 25-year span of 1991-92 to 2016-17, just three single-season teams out of 711 had an ES+/- per game of +1.00 or higher, and only twelve of those 711 were -1.00 or worse. In the 12 seasons from the WHA merger up through San Jose’s entrance to the NHL (252 team-seasons), there were fifteen single-season teams at or above +1.00, and thirteen at or below -1.00. But we back up to the span of 1967-68 to 1978-79, where 185 team-seasons produced twenty at +1.00 or higher and eighteen at or below -1.00.
What we also see are the extremes on both ends. The highest single-season ES+/- per game in the last 25 seasons is +1.119 by the 1993-94 New Jersey Devils, followed by +1.0121 by the 1995-96 Detroit Red Wings. The bottom three are -1.6667, -1.7143, and -1.7500, and it goes without saying that these are the 1992-93 and 1993-94 Senators plus the 1992-93 Sharks. The next worst is -1.3625 by the 1991-92 Sharks, -1.2683 by the 1999-00 Atlanta Thrashers, then -1.0854 by the 1997-98 Tampa Bay Lightning.
From 1979-80 to 1990-91, we see a high of +1.4750 by the 1981-82 Montreal Canadiens, then +1.3375 by the 1981-82 Edmonton Oilers, then +1.3000 by the 1983-84 Oilers. There are eight teams in this span that exceed the top one of the last 25 years, and fifteen that exceed the second-place one (with a record 62 wins). On the bottom end, we don’t see the same type of extremes; the worst is -1.7500 by the 1980-81 Winnipeg Jets, then -1.6500 (1989-90 Quebec Nordiques) and -1.5000 (1982-83 Hartford Whalers). On the other hand, the sixth-worst team of the last 25 years (-1.0854) would only be 12th from the bottom in this group.
But by the time we get to the NHL/WHA Expansion Era, everything goes nuts. The top six teams (again, of 185) are all at +1.5000 or higher, a mark that hasn’t been touched since. Four of these six teams are Montreal, topped by the unbelievable 1976-77 team that posted a staggering single-game +2.2875. The second-place team is 1970-71 Boston with a +1.8462. Many pundits have said that the 76-77 Canadiens are the greatest team in history, and this may simply bolster that argument further.
From 1991-92 to 2016-17, only three teams were +1.0000 or higher. From 1979-80 to 1990-91, fifteen teams were in this range. From 1967-68 to 1978-79, twenty teams did. The 1993-94 Devils, highest of the last 25 years, would be 17th in the 1967-68 to 1978-69 time period. And then down toward the bottom, we have six teams that are -1.5000 or worse. It goes without saying that the worst of all-time is the 1974-75 Washington Capitals, at -2.6125. They’re the only team at -2.000 or worse for a season, although the 75-76 Capitals came close with a -1.9750. There were 18 total teams, just under 10% of the team-seasons in this time span, with a -1.000 or worse.
A possible study could end up being done that further analyzes this group of data from the last 50 years and finds some way to look at something like competitive balance over time. I believe that what we’ll find is a significant tapering at both the top and the bottom as time progresses with some flaring in the expansion years, although I believe this flaring would be much smaller in the last expansion group compared to the previous ones. And the flattening of the extremes would speak to the idea that the league itself is more balanced than ever, which may in itself have significant implications toward the actual state of the game and its ability to handle further expansion without negatively affecting the product on the ice.
Can we go back even further still?
No. I mean, we could because official power play and penalty kill records exist for 1963-64, 1964-65, 1965-66, and 1966-67. But there’s not really a point in compiling and analyzing the 25 years of the six-team NHL. It would be interesting to do so, but I don’t believe that it would actually be meaningful.
Where’s the data?
Right here. This will require Microsoft Excel, and the version I uploaded here is an .xls so it’s compatible with the 1997, 2000, 2002, and 2003 editions as well as the newer ones. Feel free to download and play with as you see fit. Playoff teams are indicated with an asterisk next to their name and with a mint-green background in their cells.