Sep 262011
 

A week or two ago I presented a prediction of the eastern conference using a purely statistics based analysis.  There were a number of limitations with the process which I outlined at the beginning of the post but I have fixed some of those so this is version 2.0 of the prediction algorithm.  Let me summarize the process.

  1.  I took each teams current rosters and estimated the amount of even strength, power play and shorthanded ice time each player on the roster would play.  For veteran players, the estimates were loosely based on previous years ice time which should give us a pretty accurate number for the majority of the players, serious injuries aside.
  2. I then combined the ice time data with my 3-year 5v5close, 5v4 power play and 4v5 shorthanded HARO+ and HARD+ ratings.  I used 3-year ratings because I think they more reliably reflect each players true abilities where as one year, and even two year, ratings have significant margins of error associated with them.
  3. For rookies and other relatively un-established players I had to take guestimates at their ratings and their ice times.  Most rookies or players with little NHL experience to develop ratings with I guestimated them to be below average players, except for players who are premiere prospects in which case I rated them more like an average player.   It is actually somewhat rare for rookies to perform significantly above average, especially defensively.
  4. Unlike my previous ratings, I did make adjustments for strength of schedule.
  5. Also, unlike my previous ratings, I did make adjustments for teams that might get more or less than an average number of power play or penalty kill opportunities.  To do this I used each teams total power play and short handed situations over the past 2 seasons and compared them to the league average.  For teams which more powerplays than the average team had their power play goal production increased and those with less had their power play goal production decreased accordingly.  The same was done for the penalty kill.  Of course, if a team changes their playing style to take or draw more or fewer penalties than in the previous 2 seasons the reliability of the predictions will be degraded somewhat.

As with the previous post, I haven’t converted goals for/against into points in the standings but this gives you an indication of how the numbers seem to view the teams talent levels.  So, with that said, here are your eastern conference predictions.

Predicted Last Season
Team GF GA GF-GA GF GA GF-GA
Boston 227.1 203.5 23.5 244 189 55
Pittsburgh 242.0 219.9 22.1 228 196 32
Buffalo 235.4 217.7 17.6 240 228 12
Washington 236.3 218.9 17.3 219 191 28
Philadelphia 239.3 222.1 17.2 256 216 40
Toronto 245.3 235.6 9.6 213 245 -32
Tampa Bay 233.6 224.4 9.3 241 234 7
NY Rangers 217.5 210.8 6.8 224 195 29
Montreal 226.5 225.6 0.9 213 206 7
Carolina 227.5 231.0 -3.5 231 234 -3
Florida 211.4 216.6 -5.2 191 222 -31
New Jersey 202.3 210.1 -7.8 171 207 -36
NY Islanders 227.4 240.5 -13.1 225 258 -33
Winnipeg 210.3 235.5 -25.2 218 262 -44
Ottawa 189.5 251.8 -62.3 190 245 -55

Before getting into some team specific observations, a first observation worth noting is that the goals for and against predictions seem to be more compressed than what typically occurs in the NHL standings.  The predicted goals for totals range from a high of 245 to a low of 189.  The low of 189 is perfectly reasonable (the lows from the previous 3 seasons are 171, 196 and 190) but the high of 245 is well below the high totals of previous years.  Last season the Canucks scored a high of 258 goals, the previous season the Capitals led with 313 followed by the Canucks with 268 and in 2008-09 the Red Wings led with 289 goals.  I am not sure if this is evidence of increased parity or whether it is a flaw within the ratings system and/or the prediction algorithm.

Team GF Team GA
Toronto 245.3 Boston 203.5
Pittsburgh 242.0 New Jersey 210.1
Philadelphia 239.3 NY Rangers 210.8
Washington 236.3 Florida 216.6
Buffalo 235.4 Buffalo 217.7
Tampa Bay 233.6 Washington 218.9
Carolina 227.5 Pittsburgh 219.9
NY Islanders 227.4 Philadelphia 222.1
Boston 227.1 Tampa Bay 224.4
Montreal 226.5 Montreal 225.6
NY Rangers 217.5 Carolina 231.0
Florida 211.4 Winnipeg 235.5
Winnipeg 210.3 Toronto 235.6
New Jersey 202.3 NY Islanders 240.5
Ottawa 189.5 Ottawa 251.8

 

The teams with the largest predicted improvements in goal differential are the Leafs (42 points), the Devils (28), Panthers (26), Islanders (20), and Jets (19) while the teams predicted to fall back the most in terms of goal differential are Boston (-31), Philadelphia (-23) and the Rangers (-22).   The predicted top 6 scoring teams in the east are Toronto, Pittsburgh, Philadelphia, Washington, Buffalo and Tampa while the lowest scoring teams are predicted to be Ottawa, New Jersey, Winnipeg and Florida.  The teams with the predicted worst defense are Ottawa, Islanders, Toronto, Winnipeg and Carolina and the predicted best defensive teams are Boston, New Jersey, NY Rangers, Florida and Buffalo.  While there are a couple of surprises in there, most of those seem quite reasonable.  Now for some team specific observations.

Washington Capitals – The Capitals played a different game last season from the previous two seasons.  In 2008-09 they scored 268 goals but gave up 240, in 2009-10 they scored 313 and gave up 227.  Last season they improved significantly defensively giving up just 191 goals but their offense also suffered as they scored just 219.  The predictions are predicting the offense will come back next season but will cost them a little defensively.  Mathematically speaking it makes sense, but in reality it is difficult to say whether they will change their playing style back to a more offensive game or not at the cost of defense.  We’ll have to wait and see.

Toronto Maple Leafs – One of the biggest surprises in these predictions is the offense of the Maple Leafs.  They are predicted to score the most goals of any team, eastern or western conference.  A big reason for this is both Joffrey Lupul (who played just 28 games with the Leafs) and Tim Connolly have very good HARO+ ratings as do many of the returning Leaf forwards including Kessel, Kulemin, Grabovski, and MacArthur.  Even projected third line players Armstrong and Bozak have solid HARO+ ratings.  If the ratings are true, scoring goals shouldn’t be a problem for the Leafs and in fact the late season surge last year was predominantly a result of increased goal production and not solely due to the play of James Reimer.  The Leafs problematic defensive ability is still an issue for the Leafs though.

New York Rangers – It is difficult to fathom how a team that added Brad Richards will see their goal production drop from 224 to about 217.  This is a little dumbfounding, but the Rangers did lose 16 goals from Frolov and Prospal and the algorithm is certainly not predicting another 21 goals from Brian Boyle (his previous career high was 4) so it is certainly possible that Richards won’t dramatically increase the Rangers offensive output.  We’ll see.

Philadelphia Flyers – Unless some of the younger players really step up their games it is difficult to see them being as good a team as the Flyers from last season.  They are predicted to score 16 fewer goals but give up 6 more (despite Bryzgalov).

New Jersey – The Devils will be a dramatically better team this year, but they still may not be a very good one.  They have some highly talented forwards (Parise, Zajac, Kovalchuk) but they depth is weak and they will produce very little offense from the back end and who knows what Brodeur has left in the tank.

Florida Panthers – They brought a lot of players in this past off season and they should have an improved team but like the Devils it might be a stretch for them to make the playoffs.

Tomorrow we’ll take a look at the predictions for the western conference standings.

 

  One Response to “Predicting the Eastern Conference – Version 2.0”

  1.  

    I have to say, regardless of the validity or accuracy of the predictions that are produced with your algorithm, it is always a great read. The amount of time that must go into producing these predictions must me exhausting.

    Good Job on all your articles.

Sorry, the comment form is closed at this time.