2005 - 2006 Frequency Standings

N long_name W L T ePTS PTS SW TL OL LW CB RW 1 Detroit Red Wings 69 23 8 123 124 3.11 55% 41% 56% 71% 0 2 Ottawa Senators 65 25 10 119 113 2.54 53% 41% 62% 62% 1 3 Dallas Stars 57 28 16 112 112 2.95 49% 43% 49% 62% 0 4 Carolina Hurricanes 58 30 12 111 112 3.32 49% 45% 48% 67% 4 5 Buffalo Sabres 62 26 12 116 110 3.35 54% 40% 47% 67% 2 6 Nashville Predators 57 32 11 107 106 3.30 52% 43% 46% 62% 0 7 Calgary Flames 59 30 11 110 103 2.96 57% 38% 46% 54% 0 8 New Jersey Devils 53 32 16 106 101 2.75 50% 42% 46% 52% 1 9 Philadelphia Flyers 55 33 12 105 101 3.30 52% 42% 43% 60% 0 10 New York Rangers 51 36 13 100 100 3.05 48% 45% 45% 56% 0 11 San Jose Sharks 57 34 10 105 99 3.37 49% 47% 48% 64% 1 12 Anaheim Ducks 54 34 12 103 98 3.24 49% 45% 45% 60% 2 13 Edmonton Oilers 48 32 20 103 95 3.71 46% 47% 37% 66% 3 14 Colorado Avalanche 54 36 11 101 95 3.23 50% 45% 44% 59% 1 15 Montreal Canadiens 52 42 6 92 93 3.29 50% 47% 45% 56% 0 16 Vancouver Canucks 50 40 10 94 92 3.47 50v 46% 41% 58% 17 Tampa Bay Lightning 50 38 12 97 92 3.29 47% 48% 44% 59% 0 18 Toronto Maple Leafs 51 36 12 99 90 3.32 50% 45% 42% 59% 19 Atlanta Thrashers 49 39 12 96 90 3.25 45% 49% 44% 60% 20 Los Angeles Kings 48 44 9 89 89 3.15 47% 49% 44% 55% 21 Florida Panthers 45 44 11 87 85 3.39 48% 47% 37% 53% 22 Minnesota Wild 44 46 10 84 84 3.15 41% 54% 46% 57% 23 Phoenix Coyotes 45 46 9 85 81 2.90 44% 52% 48% 52% 24 New York Islanders 39 46 15 82 78 3.20 43% 50% 36% 51% 25 Columbus Blue Jackets 38 49 14 79 74 3.43 39% 56% 37% 58% 26 Boston Bruins 38 50 12 77 74 3.53 45% 49% 32% 52% 27 Washington Capitals 32 51 17 73 70 3.53 42% 51% 28% 50% 28 Chicago Blackhawks 33 57 10 67 65 3.09 38% 57% 37% 48% 29 Pittsburgh Penguins 29 62 9 59 58 2.89 33% 63% 40% 47% 30 St. Louis Blues 25 59 15 60 57 3.40 42% 51% 23% 41%

I've added RW - rounds won. Reference in previous post, for the TL, OL etc.

If one assumes CB is the most important playoff stat, then it should be no surprise that Edmonton did well (66%) and Carolina won (67%). It's a bit of a shock that Edmonton beat Detroit (71% vs. 66%), however these are odds not guarantees. Of course there are other factors such as pre-post-season trades and injuries to consider. Of course this is one playoff dominated by two teams worth of data (Edmonton and Carolina are involved in 7/15 series)

N team1 TL1 OL1 LW1 CB team2 TL2 OL2 LW2 CB2 W1 W2 1 DET 0.55 0.41 0.56 0.71 EDM 0.46 0.47 0.37 0.66 2 4 2 S.J 0.49 0.47 0.48 0.64 EDM 0.46 0.47 0.37 0.66 2 4 3 ANA 0.49 0.45 0.45 0.60 EDM 0.46 0.47 0.37 0.66 1 4 4 CAR 0.49 0.45 0.48 0.67 EDM 0.46 0.47 0.37 0.66 4 3 5 DAL 0.49 0.43 0.49 0.62 COL 0.50 0.45 0.44 0.59 1 4 6 ANA 0.49 0.45 0.45 0.60 COL 0.50 0.45 0.44 0.59 4 0 7 NSH 0.52 0.43 0.46 0.62 S.J 0.49 0.47 0.48 0.64 1 4 8 CGY 0.57 0.38 0.46 0.54 ANA 0.49 0.45 0.45 0.60 3 4 9 CAR 0.49 0.45 0.48 0.67 N.J 0.50 0.42 0.46 0.52 4 1 10 N.J 0.50 0.42 0.46 0.52 NYR 0.48 0.45 0.45 0.56 4 0 11 BUF 0.54 0.40 0.47 0.67 PHI 0.52 0.42 0.43 0.60 4 2 12 OTT 0.53 0.41 0.62 0.62 BUF 0.54 0.40 0.47 0.67 1 4 13 CAR 0.49 0.45 0.48 0.67 MTL 0.50 0.47 0.45 0.56 4 2 14 BUF 0.54 0.40 0.47 0.67 CAR 0.49 0.45 0.48 0.67 3 4 15 OTT 0.53 0.41 0.62 0.62 T.B 0.47 0.48 0.44 0.59 4 1

So I did a binary logistic regression: (1/(1+exp(-(a+bx))) on the difference of each variable in the series. One did come out significant, however, there simply is too little data to be sure of any of the results. I'll have to go back a few years to fix that, but for now I'll work with what I got. So the percentage of concordant results is below with the p-value (probably it's a bad model) in brackets.
Difference in TL: 64% (20.2%) Negative (Fewer goals for when tied results in more winning)
Difference in OL: 59% (37.4%) Positive (More goals against when tied results in more winning)
The above two statements are nonsense. Although what it could be saying is that scoring goals when the game is tied is easier in the regular season than the playoffs so being good in the regular season may not help you that much in the playoffs. TL is almost directly correlated to OL (OL+TL ~ 1), so doing a regression on one or the other should have relatively the same results
Difference in LW: 52% (78.9%) Negative (Being better at winning when you have the lead doesn't result in more wins)
With a p-value of 78.9% anything here is nonsense and well as you can see it's nonsense. However, it's been said: "Offense wins games, defense wins championships", this result doesn't exactly support that claim as LW could be perceived as a defense statistic, although if you score a lot of goals when you're in the lead you won't need much defense.
Difference in CB: 80% (2.6%) Positive (Better ability to come back results in more wins).
Even with this tiny model the p-value falls below the "magical" 5%. The estimates (a,b) are not significant, but work out to about 6% increase in winning percentage for every 1% improvement in comeback frequency, so a 5% difference gives you 30% better chance of winning the series (that's good!). Of course since the estimate is wrong it could be anywhere from 0% to 12%. So my intuition about the last playoffs is correct: teams who can comeback did well, whether this is a trend or just the result of last playoffs is still up for analysis.

This shouldn't be too much of a shock as teams who can comeback demonstrate resiliency, determination and grit, much of what is needed for a successful playoff run. I'll run 2003-2004 some other time, but it's an interesting start. Also you have to win that elimination game to actually win the series, which isn't easy against a team that doesn't give up. Of course it's likely that an extremely dominant team could have a bad CB and still do well. Although it might be counter intuitive CB depends on goaltending as well as scoring, as the goaltender needs to keep the game close in order to give his scorers a chance to comeback.