# 2014-2015 Season Predictions

Hockey is almost back! Hurray! With the pointlessness of the preseason finally behind us, and with the start of the real games less than 24 hours away, I thought I’d throw my hat into the ring and offer my best attempt at crystal ball gazing. I wanted to come up with a methodology that was relatively straightforward for 2 completely selfish reasons: 1) I ran out of time to do a more complicated methodology that I was planning; and 2) I’ll hopefully be able to blame the simplicity of the model when everything goes wrong. The basic approach that I took (which is certainly full of holes that I’ll elaborate on below) was:

• I downloaded each team’s current roster from Wikipedia (possible source of error #1)
• For each player, I estimated their even-strength time on ice by taking their historical TOI Pct over the past 3 years and adjusting that for their current team. In other words:

Est TOI = Lg Avg Tm TOI * (Ind. Avg TOI Percent)/(Total 3-Year Avg TOI Percent for Current Team)

For players without a TOI prediction, I used the average forward or defenceman rate in the calculation above (possible source of error #2)

• I then classified each team’s goalie as either a starter, 1A, 1B or backup. Based on that classification, I assigned each goalie an expected games played (possible source of error #3), and used the projections from Hockey Graph’s Marcels system to estimate a team level Save Percentage for the year.
• Lastly, I took each player’s weighted average xGD20 (possible source of error #4) over the past 3 years (using a 5-4-3 and TOI based weighting), adjusted it for the team-level Save Percentage I calculated above, and used it to calculate an expected goal differential. For players without a prediction (rookies, for example), I just assumed they’d be league average players over the coming year. I then ranked the teams by expected goal differential, and presented the (sometimes somewhat unbelievable) results below.

As I said, there’s a lot of areas where this prediction could go wrong:

• First, the rosters I used are likely incorrect, and if not, they almost certainly will be within the week.
• Second, my goalie playing time projections are going to be off, and possibly significantly so. And while the Marcels system provides us with a good estimate of where we expect a goalie to be, single season save percentages are incredibly variable (and hence, incredibly difficult to predict).
• Third, while xGD20 is a metric that tends to persist fairly strongly season-over-season, it is based on relative statistics rather than absolutes. This is an issue as putting together a group of good-player-from-bad-teams won’t necessarily yield a good team, and vice-versa.
• Fourth, none of this takes into account special teams or shootout ability. A project for next year I suppose.
• And lastly, of course, predictions for a whole season of hockey are really, really hard to do, so cut me some slack with the output. What’s presented below are the results of the model I described above, with no other tweaks or inputs from me whatsoever. I hope that over the course of the year I’ll be able to improve upon my methodology to the point where I can actually make predictions on the number of points a team will earn, but at this point you’ll have to live with my duct-taped together solution.

So without any further delay, here goes:

Western Conference (* denotes wildcard team)

Pacific

1. Los Angeles Kings
2. San Jose Sharks
3. Vancouver Canucks
4. Anaheim Ducks*
5. Edmonton Oilers
6. Arizona Coyotes
7. Calgary Flames

Observations

• The Kings are predicted to be the top team in the west and are favoured to advance to the Stanley Cup finals for the 3rd time in the past 4 years. Also, Anze Kopitar is really good.
• For all those Sharks fans worrying whether San Jose’s decision to revamp what was a very good team from last year would hurt them, this prediction should give you a bit of comfort. In spite of their offseason, the Sharks still go into 2014-2015 looking like they’ll comfortably be in a playoff position come the spring.
• The Ducks are lower than one might expect of the reigning President’s trophy winners for one reason: goaltending. Since Marcels doesn’t have predictions for either of their predicted goalies, the model is assuming both Frederik Andersen and John Gibson will perform at replacement level. If they manage to perform at a league average level, the Ducks jump up to 2nd in the division. If they’re better than that, they could easily compete for another conference title.
• Good news, Oilers fans: you’re probably not going to have a lottery pick this year. Bad news: the playoff drought looks likely to continue.

Central

1. Chicago Blackhawks
3. Dallas Stars
4. Nashville Predators*
5. Louis Blues
6. Minnesota Wild
7. Winnipeg Jets

Observations:

• Chicago is likely going to win this division, and possibly by a lot. You did not need a model to tell you this.
• Nashville and St. Louis are extremely tight in the predictions, so I wouldn’t be surprised (nor would most people, I’d guess) to see the Blues ahead of the Predators at the end of the year. The Blues seem intuitively too low by this method to me, and I suspect that there are likely team/system effects that we’re not capturing that may give them and edge over the Preds. Or perhaps the system (like everyone else) is just underestimating Brian Elliot, again.
• Minnesota, however is way behind both of those teams, primarily because the model doesn’t like they’re goaltending and/or is confused about who is actually going to play most of the games for them (ok, that might be me and not the model).
• The Avs spot as the number 2 team in the division is likely surprising to most people – as this years “test case” for the analytics revolution, Colorado has been predicted by many to fall back down to earth in 2014/15. While the model generally agrees that the Avs skaters are due for some heavy regression towards the mean, Marcels has Semyon Varlamov remaining as one of the top goalies in the league for the coming year, projecting him to put up the 5th highest Save Percentage of all predicted goalies at 0.9162. If we roll that number back ever so slightly to 0.915 it’s enough to drop the Avs down to 3rd, while if we move it down to a league-average 0.914 the Avs fall out of the playoff picture all-together. Godspeed, Semyon-Patrick Roy is going to need you.
• On the other end of the spectrum, the Winnipeg Jets are being held back almost singlehandedly by Ondrej Pavelec. If we move Pavelec up to league average the Jets skyrocket up the standings to second in the division! Second! And if they simply get replacement level netminding, our model still thinks they’re a playoff team. Kevin Cheveldayoff, you’re a gluten for punishment.

Eastern Conference (* denotes wildcard teams)

Atlantic

1. Boston Bruins
2. Toronto Maple Leafs
4. Ottawa Senators
5. Tampa Bay Lightning
6. Florida Panthers
7. Detroit Red Wings
8. Buffalo Sabres

Observations:

• Well let’s start with the obvious: this is a very bold prediction for the Leafs. And no, it isn’t a typo. And while, subjectively at least, I’m not sure it’ll happen, hear me out on why it could First, the Leafs have shed a lot of dead weight: Dave Bolland, Tim Gleason, Colton Orr, and Frazer McLaren will all start the coming season with different teams (although admittedly, the last two could find their way back into Randy Carlyle’s lineup). Second, the players who’ve replaced them are generally good possession players: Robidas, Komarov, Booth, and Santorelli are all players who have posted a higher FF% than their teammates over the past 3 years. And lastly, they have incredibly strong goaltending with the pair of Jonathan Bernier and the reluctant James Reimer.

Now none of this is to say that this is a sure thing, as the Leafs may be one team that highlights a particular flaw in the model, being that it tends to correlate somewhat strongly with Corsi-Rel. While the Leafs roster this year is full of players who have generally outperformed their teammates, when your teammates were as bad possession-wise as the Leafs were the last few years that’s not really saying much. The other factor that’s not being accounted for here is Randy Carlyle and what affect, if any, he’s having on this team. In general, it’s very difficult to quantify a coach’s influence over a team’s performance, but this is one case where a lot of the non-analytical evidence seems to suggest that perhaps the style of play he’s advocating for is holding the team back. If that’s the case, then any additions the new management team has made this offseason may be all-for-naught. Of course, if they discover it early enough, who knows where the Leafs might land with a new captain steering the ship.

• On the less controversial side of things, Boston, you’re my pick to win the Cup this year. Going out on a limb, I know.
• Ottawa is not in as low a position as I expected they might be. While the model still views them as being heavily dependent on Erik Karlsson, if they can get decent play out of the rest of their defensive core (or at least better than the model is predicting), they could compete for the last playoff spot in the East.
• Tampa may be a bit low here, and that will be particularly true if Jonathan Drouin plays up to his potential. With that being said, take a look through their roster: swapping out Marty St. Louis for Ryan Callahan hurts. So does adding Brian Boyle and while Anton Stralman was obviously a big pick-up, this is still a team that missed the playoffs 2 of the past 3 seasons even with Steven Stamkos in the lineup.
• Oddly enough, while the Sabres appear to be a near lock for the basement of the Atlantic division, they’re not favoured to win the Connor McDavid sweepstakes. That honour goes to…

Metropolitan

1. Columbus Blue Jackets
2. New York Islanders
3. New Jersey Devils
4. Pittsburgh Penguins*
5. New York Rangers*
6. Washington Capitals
7. Carolina Hurricanes