17

L Is For Las Vegas

When it comes to winning in Las Vegas, the only real rule to follow is that you need to quit while you’re ahead.

Unfortunately for the Golden Knights, that option isn’t available when you’re playing in a professional sports league. So while they’ve certainly had a fairy tale inaugural season to date, there’s still a long way to go before it’s over, and a lot can happen between now and the start of the NHL playoffs. Cashing out is not an option.

But whatever happens between now and the end of the season (and beyond) is almost irrelevant. What the Golden Knights have done so far this season is unprecedented for an expansion team, since perhaps the St. Louis Blues entered the league in 1967 and went to three consecutive Stanley Cup Finals. But that was in a 12 team league.

The question is how has a team of cast-offs that many, including me, panned over the summer managed to pull this off?

To find out, we’re going to have to go deep.

But I don’t mean deep as in lots of math; I mean deep as in roster depth. Because what the Vegas Golden Knights are proving is that in today’s NHL, the deeper your roster, the less likely you are to finding yourself under water:

What the endless drive toward parity in the NHL has done is turned the regular season into a grind that dulls the impact of high-end players on overall results. This has been largely accomplished through the salary cap, which has been somewhat effective at forcing some redistribution of talent across the league.

The topic of talent distribution and impact on results has been looked at in the past. Most recently Alex Novet presented some results at the Vancouver Hockey Analytics Conference, and subsequently published an article over at Hockey Graphs showing that:

  • Hockey is a strong link game, i.e., the team with the best player usually wins
  • Therefore, teams should prioritize acquiring the very best elite talent, even at the cost of having weaker depth than opponents
  • This is important for roster construction now and has the potential to become even more important as teams get better at assessing talent and market inefficiencies become less common

And previous to that, back in 2015 our own Cam Lawrence found that:

Virtually every team who has made it to the conference finals in the last 5 years had both an over-20 GAR (goals above replacement) elite player, and at least one 15-20 GAR star to complement them.

But when you think about the Golden Knights, you don’t really get this feeling that they’re exactly stocked with elite, superstar talent. No offense to Alex Tuch, William Karlsson, and Jonathan Marchessault. It’s bad enough that there were at least three NHL GMs poor enough at their jobs to gift George McPhee with these three players, but they aren’t exactly in superstar territory.

What they do have, though, is half decent players up and down the lineup. And as well all now, when you put together an ensemble cast of underdogs and specialists and bring them to Vegas, more than likely there’s going to be a heist:

And sure enough, Gallant’s 20 the Golden Knights have been stealing two points from a lot of unsuspecting teams this season.

To try and understand why this is happening, let’s take a look at the underlying numbers. But first, a word about the data I’ll be using. All of the data in this post was pulled on January 16 from Natural Stat Trick, and is 5v5 score and venue adjusted.

The shot metrics show that at just over the half-way point of the season, the Golden Knights were 10th overall with a shot attempt ratio (or Corsi For %) of 51.6%:

But in terms of standings points, they were second only to the Lightning with a points % of 0.709. In other words, they have taken 70.9% of all the available points so far this year. That’s quite a heist for an expansion team.

We can see by this discrepancy, that the team-level shot attempt ratio doesn’t go all the way to explaining the results Vegas, and other teams, have achieved so far this season. In fact, a simple linear regression shows that CF% only explains about 31% of the points % results across the league:

Looking at either the table or the chart, you can see there are seven teams with basically the same CF% as Vegas, but with wildly differing results ranging from the Oilers down at a points % of 0.467 all the way up to the Golden Knights. Clearly team level shot metrics are not giving us the whole picture if you can have teams with such similar shot attempt ratios putting up such a wide range of results.

But what happens when you look at this in terms of how “talent”, as measured by on-ice shot ratios, is distributed across each team’s roster?

To do this, I built a multiple linear regression model using the highest, lowest, and mid-point of the shot attempt ratios (CF%) of all the regular skaters for each team to account for the range of talent, and then added standard deviation to account for how that talent is distributed around the mid-point. Using the player-based shot attempt ratios, rather than a single average number for each team, increased the accuracy of the model to 67%. As this only accounts for ability to control shot attempts, we can also add team-level shooting and save percentages to help strengthen the link between the underlying metrics and the observed results:

 

Our model now explains almost 90% of the observed results, which is extremely high. We saw above that using only team-level shot metrics explained only 31% of the results, and for comparison, if we were to use only shooting and save percentages for each team, the model would explain only 40% of the results. So looking at player-level shot metrics combined with shooting and save percentages greatly improves the performance of the model.

So now that we have a model that accurately predicts how well the Golden Knights have been doing, what does this tell us about the reasons behind that success?

Looking at the shot metrics for the Vegas forwards, you can see that they have one extremely well-performing top line, and then everyone else that can pretty much hold their own. There is no one down in fourth line territory, below the dotted line. The situation is even better among defensemen, with four of the seven defensemen to have played 20 or more games performing at top line level. Only our old friend Luca Sbisa sits down there in replacement level territory.

 

When you break the team level metrics down by the spread of performance across the entire roster like this, it’s easier to see why the Golden Knights are performing well. They don’t have any obvious talent gaps all the way down their lineup.

Almost every one of their regulars is above average in terms of on-ice shot generation, and the result is that they can come at you in waves. They are particularly effective when the score is within a goal. Not only do they tend to outscore opponents at 5v5 when tied, but they lead the league in goal differential when trailing by one, and unlike the Canucks, they seem to have that killer instinct and tend to put teams away when they get up a goal:

The Golden Knights score early, and they score often. There maybe not be any superstars on the rosters, but they have the speed and depth to continuously put on the pressure, and can counterattack quickly. In fact, they are third behind only the Islanders and Lightning on scoring chance conversion rate, and this is largely because they are often found on the rush. So much so, that if the US Army succeeds in its trademark infringement lawsuit against the Golden Knights, maybe they should consider an arrangement with a long-standing Vegas organization and just change their name to the Golden Nuggets:


RECENT GRAPHIC COMMENTS

  • Does McPhee get a reprieve for the Forsberg for Erat trade now given all the success with Vegas? Still not as bad as a first (Barzal) and 2nd rounder for Reinhart, who’s not even an NHL’er.

  • The Knights success is a key reason why Benning should be canned asap.

    While LinBenning have floundered and flip flopped like turned over turtles for FOUR YEARS with zero progress, McPhee looked at the league, identified a team identity of speed, skill and hard work with depth and parity throughout the lineup. He then hired an excellent coach and the rest is history. Incredible work with everyone elses cast-offs that has simply put Vancouver to shame and Vegas in the playoffs. Let’s also remember, none of McPhees draft picks are even playing for the Knights yet! (9-2 against us in pre-season i recall)

    Guys, Colorado and New Jersey and are also great examples of non playoff teams who have turned their franchises around in half a season with great moves and management decisions… but hey, the blowhards here are talking about a Canucks contender by 2022, so yeah, keep drinking the koolaid suckers. LOL

    Viva Las Vegas

    • You keep saying this, but they have nothing to do with each other. Managing a team that’s been around for decades, that has a farm system, ELC players, RFA players, UFA players, players on multi-year contracts, some of them with NTCs, and so on, is a completely different ballgame from building a team from scratch through the expansion draft. They are not comparable in any way. By this logic, every GM in the league except Yzerman should get the ax because Vegas is playing better.

      You want to criticize Benning, compare his work to other GM’s who’ve been in similar situations. Compare him to the job Calgary’s management is doing, or Toronto’s management is doing, or Carolina’s management is doing. Or compare him to how badly the Oilers have been bungled.

      I’m of the opinion that Benning’s time as Canucks GM has been very much a mixed bag, and that there’s much to be criticized. It’s *easy* to find legit things to criticize, you don’t have to BS and come up with complete nonsense.

      Alternately: trolololo lololo lololo, hahahahha.

    • You mention Vegas and that is an example of a team with some high end talent being complemeted by quality depth players but you dont seem to realize that the nhl has a cap for every team. You seem to want a team that has a Boeser on every position; probably another run at the xbox stanley cup will satisfy you. I know I should not be feeding these trolls but….. Pls indulge us with your expertise this coming draft without the benefit of hindsight.

  • Couple questions on your models:

    I’m your first regression model an Rsquared value in the 30% is statistically irrelevant and basically has no correlation… So the model is junk from a predictive and statistical stand point.

    In your multiple regression model, shouldn’t you have used the adjusted Rsquared? And maybe touch on the Pvalue… to explain the significance of each factor?

    Love the analysis but that statistician in me is cringing…

    Finally, is there any multiple regression model that has 95% or better R2 for predicting points or wins? I feel like with all that data you guys got the models should be more accurate…

    • 1. The relationship between CF% and Points % may not be strong, but the regression *is* statistically significant at the 95% confidence level.
      2. Sure, I could have gone into way more statistical detail, but this blog isn’t the place for that. I’ll probably be writing more on the general model over at Hockey Graphs soon…
      3. Yes, I’m sure you can overfit a model to give you a very high correlation for descriptive purposes. Just base it on goals. But that’s not going to be very helpful for out-of-sample predictions because goals are relatively rare events compared to shot-based metrics.

      • Oh Yeah, I guess that totally makes sense on point 2 and 3. In regards to 3, based on maybe shot metrics, players above replacement, special teams, etc… not using goals and box stats, do you think a more predictive model could be made?

        As for 2, would you ever post the excel output of your regression model just in table form? As the graph doesn’t tell the whole story for multiple regression (as I’m sure you know) plus for some of us, it could be pretty interesting!

        Finally, I’m not sure I agree with you on point 1 but I didn’t make the model nor look at the data so I’ll take your world for it.

        Thanks for post!

    • Asking about statistical models on an analytics blog is trolling and gibberish? Just cause you don’t understand it doesn’t mean it’s gibberish pal. Go read a book or focus on box stats if you don’t like it.

    • I just realized that you might of actually been taking about wise canucks and not “a” wise canuck… in which case sorry lol. If you were referring to me.. . Well then, see above hahah.

  • You can’t do this analysis without pointing out the structural conditions that created this particular team. The NHL set the Golden Knights up to be competitive but they they rigged the deck more than even they anticipated. You can say that CBJ, FLA and MIN were mismanaged, but in truth the structure of the expansion draft rules meant that teams with good young players at a certain age rage were much more vulnerable than a team like the Canucks who lack good players in their mid-20s. And it led to the redistribution of a lot of talent not only to LV but throughout the league. Las Vegas is a terrible example of any expansion team in any league. My question would be whether Seattle is gifted the same conditions.

    • Seattle will pay the owners more and bring in better TV revenue so they will get at least the same Vegas deal or probably better. Maybe an extra draft pick for example like pick 32 and 35.

      Depth with controllable contracts is key but look at what Vegas has to do now with bargain contracts they have to pay their top players big money. It adds up to where depth is depleted or windows shorten depending on how many prospects pan out to compete on their ELC.

      What may be needed to not punish teams for paying their star players is a luxury tax type cap circumvention that they have in the NBA. The old long term 12-15 year contfact to spread cap dollars out loophole was closed. But in the next CBA they should just be open about it and exempt one or two star players. That team pays $ as in tax to the rest of the owners or NHL operations budget and not player shared revenue. Quid pro quo benefits the star players who will always get paid don’t hamper the team cap to pay some other players for depth and better quality teams. Owners make more money from tax and small market teams can use the extra cash on their players if they wish as well.