clock menu more-arrow no yes mobile

Filed under:

Outlining the possession and overall player rankings for the 2014-15 NHL season

A project that ranks NHL players league-wide by combining various advanced statistics into a single score.

Dennis Wierzbicki-USA TODAY Sports

Late last season, I introduced a project that combined various advanced statistics into a score that could be used to rank players league-wide by their puck possession ability and overall quality. This was done separately for both defencemen and forwards. This season I will be continuing the project and providing regular updates over the course of the year.

If you are new to advanced/fancy stats and would like to learn more about them, you can check out EOTP's Fancy Stat Summer School series of posts detailing many of the stats used.

This project uses 5-on-5 data obtained from WAR On Ice. Scores for forwards and defencemen are calculated separately, but the same statistics are used in the same calculations for each position.

The statistics used and how they are combined to come up with the possession and overall scores are outlined in this article. All examples use data from a sample of players who had played at least four games in the 2014-15 season by the end of Saturday, November 1.

Possession rankings

Deployment quality (DQ)

Deployment quality is a measure of the difficulty of the situation a player faces during his ice time. It is made up of two parts: the quality of the players on the ice, both on the opposing team as well as the members of his own, and; in which end zone his shifts normally begin.

Quality of On-Ice Personnel (QoOIP)

This metric takes the average time-on-ice per game of the competition a player faces (TOI% QoC) and handicaps it with the quality of that player's teammates (TOI% QoT) he is given to oppose that competition.

The teammate handicap is calculated by subtracting the teammate quality from the quality of the competing players [TOI% (C-T)].  If a particular player's teammates are of lower quality (i.e., get less ice time) than the competition he faces, this will be a positive value.  If he plays with temmates of higher quality than his competition, it will be a negative value.

This value is then added to the quality of competition to come up with the quality of on-ice personnel metric

QoOIP = TOI% QoC + TOI% Qo(C-T)

So far this season the range of QoOIP (among players playing at least four games) for forwards goes from Vincent Lecavalier at 21.8% to Matt Halischuk at 33.1%.  This range is normalized to have all values fall within the range from 0 to 1.  Lecavalier's 21.8% becomes 0, the 33.1% of Halischuk becomes 1, and all other values fall proportionately in between.

Defensive zone start percentage (DZSt%)

This is simply the percentage of a player's total zone starts that occur in the defensive zone.

Connor Murphy has faced the easiest zonal deployment for a defenceman this season, lining up for only 14.0% of his total faceoffs in his defensive zone.  Mark Borowiecki has had the most difficult zone starts, with a defensive zone start percentage of 46.8%.

This range is also normalized to fall within the range from 0 to 1.

Deployment quality

The quality of on-ice personnel and defensive zone start percentage values are then combined--by adding the normalized values together--to come up with deployment quality.

Deployment quality = normalized value of QoOIP + normalized value of DZSt%

This value is also normalized, for use in the possession rankings.

The toughest deployment so far this season for a forward has been handled by Paul Gaustad, while Vincent Lecavalier has had his easy personnel quality combined with beneficial zone starts to have the easiest deployment quality among  forwards.

Team-quality-adjusted Corsi-for percentage (TQA CF%)

The question I am trying to answer with this stat is: what would a player’s Corsi-for percentage be if he played the same role on an average team? I would expect players on above-average teams to see their CF% drop if they were to start playing with an average team, while players on below-average teams would get a boost to their numbers.

For the purpose of calculating this stat, that 'average team' is one that would give up just as many shot attempts against as it would have shot attempts for, for a team Corsi-for percentage of 50%.  To obtain the quality of a particular player's team, I use the CF% of the team while that player is on the bench (off-ice Corsi-for percentage).

The Corsi-off percentage stat is subtracted from the 50% Corsi-for percentage of an average team.  Now it’s a matter of how much this difference between an average team and a player’s current team will affect that player’s personal Corsi-for values. It will probably be significantly more than a 0% change.  It is probably also quite a bit less than 100%, or else you’d just be flipping the league CF% rankings upside-down and having the best CF% players on the worst teams and the worst on the best teams.

The actual factor is something that would have to be looked at more in-depth, possibly by comparing the CF% of players who have had similar deployment quality on more than one team over the course of a season. For now, I will assume a factor of 50% of the difference between a player’s off-ice team and an even-Corsi team to adjust Corsi-for percentage by team quality.

Team-quality-adjusted Corsi-for percentage (TQA CF%) = CF% + [1/2 * (50 - Corsi-off percentage)]

Among defencemen to play four games this season, Jakub Kindl has the best TQA CF% at 63.5% while, at 38.7%, Jared Cowen has the worst.

This range is also normalized to become a range from 0-1.

Possession scores

The possession score is calculated by adding the normalized values of deployment quality and team-quality-adjusted Corsi-for percentage together.

Players who find themselves near the top of the possession rankings have performed well in terms of shot attempt differential given their deployment, whether that means having a respectable Corsi-for percentage with tough deployment quality or a high shot attempt differential with an easy deployment.

You can see in the tables below how the Montreal Canadiens have performed so far this season in that regard.

Montreal Canadiens defencemen


Jarred Tinordi has been spectacular to begin the season, putting up a team-quality-adjusted Corsi-for percentage above 50% despite the toughest deployment of all Canadiens defencemen.  P.K. Subban has a good shot attempt differential, but has achieved it with the easiest deployment on the team while being paired with Alexei Emelin.

Montreal Canadiens forwards


Lars Eller leads the way as the best possession forward, overcoming the fourth-toughest deployment to put up the fourth-best TQA CF%.  Jiri Sekac has also been impressive when he's been giving a spot in the starting lineup.

Overall 5-on-5 rankings

The overall rankings combine the two factors used in the possession scores and add in a point-production stat.

Primary points plus per 60 minutes of ice time (P1+/60)

Primary points plus (P1+) treats goals and primary assists (collectively ‘primary points’) as having equal value, and as being more impressive than secondary assists by giving a player only half credit for those secondary assists. So it is primary points plus half of secondary assists, or P1+ for short.

This P1+ stat is then divided by the 5-on-5 ice time of a player to obtain his Primary points plus per 60 minutes of 5-on-5 ice time.

Rick Nash leads all forwards this season, scoring 5.23 P1+ every 60 minutes, while Mike Green has a P1+/60 of 2.53 to top the chart for defencemen.  This early in the season, a lot of players at both positions have a P1+/60 of 0 to share the lowest rank of this production statistic.

The range of values for P1+/60 is normalized for use in the overall rankings.

Overall 5-on-5 quality, or Silver Score

The normalized values of deployment quality and team-quality-adjusted Corsi-for percentage are added together with the normalized value for P1+/60 to come up with an overall quality score, or the Silver Score as I've dubbed it.

The players with the highest Silver Scores are those who have performed well in terms of shot attempt differential given their deployment and also have the skill to convert that possession into goal-scoring for their team.

The Montreal Canadiens players are ranked by position below according to their overall 5-on-5 quality.

Montreal Canadiens defencemen


Tinordi has contributed a bit of offence that sees him currently at the top of this list.  I expect Subban to overtake him as the season progresses, but Tinordi has established himself as a player to watch this year.

Montreal Canadiens forwards


Max Pacioretty takes advantage of his offensive talents to jump to the top of the overall rankings for Canadiens' forwards.  Lars Eller's possession ability sees him hang on near the top of the list despite some difficulty in converting that possession into scoring.  The 'almost' effect of Rene Bouque is evident when comparing the possession and overall rankings, dropping to near the bottom of the latter despite being one of the best possession players on the team.

Schedule of updates

I will be putting together an article on the best players in the NHL to date at each position next week (for players who will have played a minimum of five games by the end of Saturday, November 8) and every two weeks thereafter, adding one game to the minimum games played criterion each time.

I will also include a list of the best defensive forwards and have updates on which players have risen or fallen the most since the previous rankings.