Numerology

Recall my earlier article back in August about the
changing world of sports journalism and something I had remarked on in regards
to the misconception amongst some in the media about the intention of advanced
analytics. Here is what I wrote: Nobody, at least not a single person I can find, is
suggesting that advanced analytics should be used as a sole source for all
hockey decisions. Instead, they are being suggested as a complement to aid in
separating the signal from the noise, when it
comes to reviewing games or making player asset decisions.

Well, in the spirit of the season (training camp and roster
projections, that is) I decided to try and do just that. I’m going to draw up a
roster made up entirely of some of the strongest players in various advanced
stats categories, and keep it all under the current salary cap, and see what
sort of team we end up with.The purpose of this is to illustrate what a numerically-chosen roster might look like and offer up some real names that we can then discuss. 

I’ll be drawing on a few different statistical categories
pulled from ExtraSkater (before it went dark), Hockey Analysis, and a Behind
the Net, then dropping the names into capgeek to use its model for calculating
cap hit and deployment. The categories include Fenwick (Fenwick = Shots +
Missed Shots), IPP (Individual Points Percentage – The percentage of goals
scored by players team while player is on the ice that the player had a point
on), and points per 60 minutes.

I’m prioritizing players who control the puck and send it in
the other direction. So, yeah, Mikael Backlund and TJ Brodie are on the team.

None of the players chosen were selected ahead of the
statistical categories. I let the numbers make the list and then went down
until I found names that I could afford so as to provide enough balance between
the forward, defense and goaltending. To be completely honest there were only
two players that I knew from the beginning I had to get on the roster and they
were the two Flames because of my familiarity with how well they handle the
possession side of the game and the value of their current contracts. Virtually
every other name that appears surprised me.

Hossa is surprisingly good as a two-way winger, where he ranks at the top of the league in regards to the Quality of Competition he faces and provides excellent offense. Toews is an excellent two-way center who can dominate play, win
possession and cover defensively against the best competition in the league while still posting strong offensive numbers. Taylor Hall is arguably one of the best
wingers and perhaps the best left-winger in the game today. His Individual Points Percentage, meaning the percentage of offensive production for his team
in which he had a direct part was a league-leading 98.1% last year, followed by
Ovechkin at 87.9%.

The second line has two strong possession
wingers in David Perron and Jaromir Jagr. Despite his age (or perhaps partially
on account of it) Jagr is ranked 7th for right wingers on the same
scale as Hossa. His Corsi Relative to Quality of Competition is an impressive
0.941, just behind players like Justin Williams, Hossa, Rick Nash and Corey
Perry. His contract is more affordable than those other players and his puck
protection and playmaking abilities would arguably complement the other two
rather well. Patrice Bergeron is the 2nd line center because of his tremendous
talent at zone entries, his high QualComp scores and faceoff percentages. David
Perron surprised me during my investigations. Exploring Hockey Analysis I
stumbled upon their HARO, HARD, and HART categories which are essentially Hockey
Analysis ratings of offense, defense, and the two combined. They adjust these
rankings with Fenwick data and then contrast them against quality of
competition and David Perron, in this case, leads the league as a player who
provides the most offense while minimizing changes against when playing some of
the toughest competition in the league. 

The third line consists of Saad, Nugent-Hopkins and Voracek and
provides a combination of speed, skill, scoring and defensive acumen. Nugent-Hopkins
is an emerging two-way center who delivers tremendous 
value defensively while contributing offensively at 5v5 and 5v4. He limits the opposition’s scoring chances as evidenced by the drop in
opposition Corsi numbers when he is on the ice to when he is off, and improves
the play of his linemates as shown in his WOWY numbers. Saad is an effective
scoring option on an ELC while Voracek is a strong option on the wing to
provide scoring who has performed well against more difficult opposition.

The fourth line is not a traditional 4th line
because of the high puck-possession metrics we see in the players here and the
overall talent level. As an example, rookie star Nathan MacKinnon is playing 4th line
wing. Backlund is a
possession monster, and his professed inability to score should be improved by
playing with Nathan MacKinnon. Antoine Roussel proved to be a very effective
depth player for Dallas last year with good possession numbers for a bottom-six
forward and a zone start differential from 41.3% to 49.2%. He also acts as
something of an agitator as reflected in his +0.3 penalty taken/drawn
differential. In short, he helps to move the puck in the right direction, adds
shots on goal, and draws penalties.

The extras are Gustav Nyquist and Ryan Garbutt, two
stand-out young players from last season who both had strong underlying numbers.

To this point I have used a number of statistical categories
that all reflect various aspects of the puck possession game, ranging from
Corsi relative numbers to IPP. For the defense, the categories aren’t always so
cut-and-dried.

My ideal role of a defensemen is that they are expected to
take the puck away from the opposition and then quickly pass it to a forward
for an offensive sortie. In some cases they are relied upon to skate the puck
out of the zone and create the rush and ensuing offensive zone pressure with a
shot on goal. As well, once zone possession has been established they are often
relied upon to hold the zone and recycle shots on net for rebounds and
deflection chances. All of these things require passing, skating mobility, and
agility, traits that I favour in blueliners. Size is a bonus in that it
provides a measure of physical range and strength to counter the forecheck, but
less so as an instigating factor in the deciding of a hockey game, measured by
hits. Hitting opponents has its place in the game as a means to separate the
player from the puck, but I’d rather have a defensive pairing who don’t have to
hit every shift in order to recapture possession.

This muddies the water and means that Quality of Competition
and straight Corsi ratings are sometimes necessary to piece together a larger
picture of a blueliner’s play. A defenseman with modest point totals and
unimpressive +/- numbers could be doing a fantastic job, but is limited by the
quality of his team or the opposition he is up against. TJ Brodie, for
instance, was even last season while collecting a modest 31 points, yet by more
discerning measures he was one of the best defensemen in the league last year.  

My criteria favoured defensemen who play tougher
competition, often start in their own zone as a measure of the coach’s trust in
their defensive abilities, and were able to record positive scoring chances
relative to their teammates.

I’ve also included the goaltending, which was selected for
players on affordable deals who have shown the ability to play at a consistent
level over a reasonable period of time. I’m of the opinion that if your team is
paying a goaltender more than $6 million a year to stop pucks, you are relying
on him too much and there are weaknesses in other crucial areas of your team
that could limit success.

Ekman-Larsson and Brodin make up the top-pairing while
Brodie and Giordano are the second pairing. Ekman-Larsson is a top pairing
defender and terrific player who just happens to wear the jersey of a team
nobody cares about. He has consistently been sent out against the toughest
competition in the league, playing a majority of his games in the Western
Conference, and ranks in the top three in the league in +/- QualComp with
weaker teammates than his competitors in that regard (Gunnarson and Phaneuf)
and produces offensive chances in his team’s favour while limiting those of the
opposition.

Brodin showed quite well in his first season playing nearly
20 minutes a night. There are better defensemen ahead of him Quality of
Competition list, however his underlying numbers as a nearly-raw rookie were
solid when paired with a strong defense partner in Ryan Suter and the economic
considerations of the exercise dictate that value be found where it can. Placing
him with another solid veteran, and countryman at that, is justified by his
play thus far in similar circumstance.

Those familiar with this site won’t likely need any
justification for Giordano and Brodie as a top-pairing defense. In this
exercise I have them in the 2nd pairing, but the difference between
1st and 2nd is virtually negligible. They log heavy
minutes against the best competition in the league and post positive possession
numbers while doing it. Brodie is listed 11th in Corsi Relative to
Quality of Competition at Behind the Net and 4th for straight Corsi
weighed against the Quality of Competition behind Roman Josi, Jan Hejda and
Shea Weber.

Mark Giordano led the league in Corsi Rel last year for
defenders playing 60 games or more, just ahead of Brodie, Marc-Edouard Vlasic,
and Jake Gardiner. He ranks 3rd in the league for straight Corsi
weighed against Quality of Competition at 1.908, tied with Shea Weber, and
behind Josi and Hejda, respectively. Giordano, like Brodie, is currently on a
very affordable contract, making the addition all the more attractive from a
mathematical point of view.

Jake Muzzin and Martin Marincin both had strong
seasons last year as largely unheralded rookies. They received vastly different
treatment from their respective coaches when it came to zone starts with Muzzin
getting the OZone push while Marincin was buried in his own end most shifts,
but in both cases they produced positive scoring chances, Marincin a larger
number relative to his team. In Muzzin’s case he was ranked 7th in
the league in Corsi Rel at 13.5, 0.2 ahead of P.K. Subban, for defenders
playing in 60 or more games with Corsi Relative to Quality of Competition
numbers also similar to Subban, 0.556 to 0.586.

Marincin was called up halfway through the Oilers’
season and stuck, being one of the few defensemen to post positive Corsi
numbers while being paired with the Oilers’ best defenseman in Jeff Petry and
playing deep in the defensive zone almost every shift. His CorsiRel shows he
has more offensive impact relative to his team than Jake Muzzin, played more
often in his own zone and against tougher competition. His Corsi relative to Quality of Competition ranks him in line with Victor Hedman , Drew Doughty,
Duncan Keith and Alex Edler, while his overall Corsi Rel for defensemen that
played at last 40 games (Marincin played 44) has him 3rd in the
league behind (drum roll) Giordano and Brodie.

Dougie Hamilton is a strong young defender who can play the
same level of competition as Muzzin and Marincin, often with better offensive
results relative to their team. In all truth, one could rotate any of the
bottom three defenders through with only a change in playing styles.

Would this team win?

Well, I’m not sure that it is necessarily the best-looking
team to an outside observer. I’m sure people would criticize the defense for
not being tough enough or the goaltending for not having an established starter.
But there are four strong centers, great wingers with size, speed and skill
mixed all over the roster, and one or two agitators to draw penalties without
being liabilities in other areas of the game.

On a more analytical level, there are a number of players
here who drive play in a variety of categories, virtually every skater scores
well in relative Quality of Competition scores, and there are several player
who are at or near the top of the league in point production categories.

The purpose of this exercise was to illustrate that while analytics
alone might not provide the definitive answer to all roster questions, they can
deliver some interesting information and can offer a counterpoint to
discussions about who is the best player at a given position.

Personally, I think it would be interesting to see All Star
teams picked by two groups, the Professional Hockey Writers Association, and a
collective of some of the leading figures in the analytics fields.

  • BurningSensation

    I had to laugh at you Rex, that’s an awful lot of Oilers for a team built on possession stats!

    I have no problem with Hall, but the Nuge doesn’t deserve to be ahead of Backlund (the Nuge is a big negative on Corsi), and while the Oiler fanbase is convinced the kid is going to be Pavel Datsyuk part deux, I’m still wondering if he isn’t Mike Ribeiro the sequel.

    And while I think Perron is a nice complimentary player, there are a lot of wingers I’d look to before him to fill out the roster, especially if I’m basing it on possession stats.

    • RexLibris

      Four Oilers and three Flames made the cut.

      Does this change one’s perspective on the teams or the stats?

      As I mention in the article, there were some surprises for me and none of this was done with fan bias in place.

      • seve927

        1/3 of a roster from the two possession challenged teams you follow, and no bias. Yeah, no. Where did you come up with using ipp as a criteria? Perhaps because Taylor Hall lead the league in this obscure category? Sven would have been on this team last year. Very weak.

    • RexLibris

      Out of curiosity I decided to compare the four names listed here on the Vollman chart adjusted for CorsiRel to account for team differences.

      http://www.hockeyabstract.com/playerusagecharts

      Feel free to do the same, but I think you’d see that Nugent-Hopkins has more in common with Datsyuk and Backlund than Ribeiro. The biggest differences between Backlund and Nugent-Hopkins is their zone starts and offensive production. Both play similar QualComps – higher than Datsyuk.

      Ribeiro gets the easier competition, better zone starts and produces less, although he, Backlund and Nugent-Hopkins all perform relatively well given their lacklustre teammates compared to Datsyuk.

  • NORRIS TROPHY! GET YOUR NORRIS TROPHY HERE!

    @burningsensation

    I’m with you on Perron but that contract is pretty darn good value for the immediate future. Other guys are better to build around but in the short term he’s a nice fit.

    Very interesting take Rex. A good read.

  • Byron Bader

    It’s a good thought and I enjoyed the article. However, a lot of the guys are on great value contracts but it’s not to say they’re under valued or being skipped over on their team. If a team had drafted all those players they would have the best team in the league without a doubt. However, a lot of those players are top player’s for their respective teams and you wouldn’t be able to trade for them for anything the world ( I realize this probably isn’t the the point).

    Hall, Toews, Bergeron, Hossa, RNH, Vorachek, MacKinnon and I’d even argue Nyquist, Backlund, Jagr and Saad are absolutely untouchable. Similar on the back-end.

    A few years ago there was probably a lot more undervalued and underappreciated guys but I’d say there’s only a few possession monsters left presently that are completely undervalued by their team. Most are signed to appropriate deals or very reasonable deals and I believe their teams know how good they are without stating publicly. The hockey world knows how good Justin Williams, Patrice Bergeron and Mike Backlund are … it’s no secret anymore.

  • mk

    I’d best interested to see a comparison vs a team built in the opposite fashion (i.e. take players with only bad ‘advanced numbers’).

    For example, take players who put up great ‘boxscore numbers’ but actively search for players who have terrible advanced stats.

  • Bean-counting cowboy

    Interesting read and approach to the debate. However, this team is soooo hooped on the cap for next year 😉 Re-ups to Toews, Saad, Backlund, Brodie, Hamilton & Nyquist.

    There’s at least $20 million or more to spent on re-upping all those players. In this sense, I don’t think this is a realistic NHL roster, as teams plan for the re-ups of their impact players.

    Hell yes this team could compete. They might win the whole thing – it’s just no team (even if they drafted them all) could secure and keep this many quality players.

  • RexLibris

    This exercise is a fantasy response to the common criticism that you can’t build a roster relying solely on advanced metrics.

    Of course there is no way you could actually collect this kind of talent into a single roster, but if you were to build something using only advanced analytics what would it look like?

    Something like this. Of course were others to run the exercise they’d come up with different names here and there because we are picking 23 roster spots from nearly 700 names.

    What I discovered from running through this exercise was that there are good to great players on very, very bad teams and that you can’t let just the numbers or the “winning” reputation cloud your view.

  • SavardianSpinorama

    This typlifies the lightning in a bucket, cheesy-assed, nacho approach that puts a pig mask on everything we have been taught about hockey.

    No hall of fame career for you, RexLibrus!

  • RexLibris

    I didn’t pick the names and then justify having them on the roster by searching for metrics.

    Jake Muzzin wasn’t even on my radar going in. Neither was Marian Hossa, David Perron, or a host of others.

    IPP is Individual Points Percentage. Meaning that when this player is on the ice, they account for a percentage of every point that their teammates produce. Ideally you would want a winger who can both capitalize and assist on production within your first line.

    Wingers on the second line and proceeding on down would need to be able to do the same but against strong competition in the event that the coach is unable to get favourable matchups.

    Those listed all managed to do exactly that in a variety of statistical categories.

    I’ve provided links to every category listed and the player’s ranks in every one.

    Anyone interested can do their homework with the information provided.