Photo Credit: John E. Sokolowski / USA TODAY Sports

Nikita Zaitsev: Actually Good

If you haven’t already read it, Draglikepull wrote an excellent article where he statistically breaks down Nikita Zaitsev’s game and explains his concerns with the contract extension. This deal been a pretty controversial topic in Leafs land lately. There’s been some great debate in the blogosphere outlining the pros and cons of locking up Zaitsev long term. Personally, I’m not a huge fan of the 7-year term, but I wanted to take this opportunity to explain why I think Zaitsev is a very effective NHL player. Consider this my “rebuttal” to Draglikepull’s piece.

Before I get into some of my key points, I want to acknowledge that everything Draglikepull has written about Zaitsev is objectively true. His zone exit & entry numbers are less than ideal; he hasn’t performed as well against top competition as Gardiner; he hasn’t generated too many shot assists this season, and he’s piled up a lot of Secondary Assists. These are facts, and I’m not going to dismiss them as #FakeNews.

Quality of Competition

The crux of my argument is based on the greyest of grey areas in hockey analytics these days: Quality of Competition (QoC). Frankly, this area of research is a bit of a mess, which I’ve written about before. The biggest problem is that most QoC measures are significantly flawed. For example, you can’t just sort players by CF% and say it’s a list of the best players of the league. This logic holds true for any stat, including xGF%, GF%, and TOI (although Dom Luszczyszyn used Points in an article last year, which I found interesting).

My personal favourite measures of QoC are Dellow’s star percentage and WoodMoney. These metrics try to determine how often players face “elite” competition. Stats nerds like myself will point out binning issues when you break things down into smaller samples like this (which I agree with), but I think that these measures do a great job of showing what matchups coaches are trying to give their players. Consider them great proxies for the great mystery that is QoC.

So let’s take a look at how Zaitsev was used this season:

Clearly, Zaitsev had some tough minutes this year, and he played most of them with Rielly, who’s notorious for his defensive struggles. Although Zaitsev was on the ice for a lot of scoring opportunities against (xGA/60), it’s difficult to know how much to blame him considering he faced extremely tough competition, as well as the fact that he played the majority of his minutes with a defenseman who consistently gives up a ton of chances.

Goals Above Replacement (GAR)

This is where DTM About Heart’s GAR data can help us. I’ve written about this metric before, but if you want the short version here it is. Using fancy machine learning techniques and regressions (which you can read about here), his model recognizes which 10 skaters are on the ice, and how much of an impact they have on play based on their prior results. I personally believe that this is currently the best way of accounting for both competition and linemate quality. It’s tricky when you have players with small samples (ie. Zaitsev, who’s only played one NHL season), but I’d still argue that it’s the best tool we have right now for making these adjustments.

So what do Goals Above Replacement say about Zaitsev?

Based on the GAR data, Zaitsev was the Leafs’ second most effective defenseman this year. After Gardiner, he was the best at driving play 5v5, his penalty differential was solid, and although he wasn’t dominant at 5v4, he was a helpful piece on one of the league’s best PP units. All in all, Zaitsev had the 62nd highest GAR among NHL defensemen, making him a high-end #3D according to the metric.

Now, I think it’s important to take this data with a grain of salt, especially considering that we don’t have any prior data on Zaitsev at the NHL level. It’s possible that the regression model may be giving him too much credit for his offensive play-driving ability, and not assigning enough credit for his defensive struggles. However, it’s also possible that the model is accurately accounting for his extreme usage and giving him a huge bump due to his insane quality of competition. Personally, I feel that the GAR data is slightly overrating him. Based on my observations this season, I would have him ranked as the Leafs’ third best defenseman, falling just behind Rielly (I’d argue that Gardiner was Toronto’s best defenseman by far, but that’s another conversation for another day).

Point Production

Although there’s been some concern about his point totals, I think that Zaitsev’s offensive production is actually repeatable to a certain extent. One of Draglikepull’s biggest criticisms of Zaitsev was that his offence was largely a result of his Secondary Assists, particularly the ones that he put up on the power play. Secondary Assists are weird because we know that they don’t have much repeatability for forwards at even strength (which is, in part, due to scorekeeping issues). When it comes to defensemen though, I would argue that they’re much more repeatable. Using machine learning techniques, here’s a breakdown of how important DTM About Heart found each box score stat is at impacting future goals:

If you find these graphs confusing, don’t worry I did too (it took me a long time to finally get any meaning out of them). What I want you to focus on are Primary Assists (A1/60) and Secondary Assists (A2/60) at both even-strength and on the powerplay. While Secondary Assists aren’t that valuable for forwards at even strength, it appears that they have significantly more value for defensemen. What’s even more intriguing is that the model actually sees Secondary Assists as more important for a defenseman than Primary Assists, which I found very surprising.

With this in mind, I would make the argument that Zaitsev’s Secondary Assists aren’t necessarily a product of luck, but reflect some semblance of skill. On a power play unit quarterbacked by Marner (or Nylander), I think it’s reasonable to predict that he can produce around the same rate that he did this season (3.41 Points per 60 at 5v4, which was 46th among NHL defensemen). When it comes to his even strength play, I’m not saying that I expect him to have more Secondary Assists than Rielly (which he did this year), but I think it’s perfectly reasonable to project him as the Leafs’ third best playmaking defenseman behind Gardiner and Rielly.

My reasoning is based on his numbers in Ryan Stimson’s passing data. Using this data (shot assists, passes into the danger zone, passes that go from one side of the ice to the other), Ryan’s come up with an “Expected Primary Assists per 60” stat that’s actually a better predictor of future Primary Assists than previous Primary Assists. You can read all about it here, which I highly encourage.

Now let’s take a look at the Leafs defensemen’s Expected Primary Assists per 60 (xA1/60) at even strength this season:

Much like the rest of my arguments, I don’t think Zaitsev is anywhere near as great as Gardiner, and not quite as good Rielly, but his passing numbers indicate that he’s creating offense at the level of a #3 defenseman in the NHL.

Another fun fact about Zaitsev is that the team was actually quite snakebitten when he was on the ice. Based on their shot locations, we would expect the Leafs to score on 6.22% of their Unblocked Shots (Fenwick), but when Zaitsev was on the ice, they only scored on 5.44% of them. This 0.78% difference may seem small, but when you do the math, you realize that they should’ve scored about 8 more goals when he was on the ice. How many of those would he have been involved with? Well, he was involved with about 33% of Toronto’s goals at even strength, so that’s about 2 or 3 points that he should’ve added to this season’s total.

I’d consider that expected boost a wash with the expected decline in his 5v5 Secondary Assists (he had 8 this season, while Riely had 5 & Gardiner had 9). What I mean by this is, although we wouldn’t expect him to have as many Secondary Assists as he did at 5v5 last year, I think the team’s shooting luck regressing back to the mean when he’s on the ice will help compensate for it. We also know based on DTM About Heart’s Expected Goals data that he should’ve scored 1 more even strength goal this year based on his shot locations.

This isn’t to say that Zaitsev’s an offensive juggernaut who was snakebitten this year. I’m just trying to make the argument that his offensive production wasn’t necessarily a fluke this year, and if given the same PP time next season (on a unit with either Marner or Nylander), I would expect him to produce points at a similar rate based on the numbers we’ve touched on. I’d also imagine he’ll score more than 0 PP goals next year considering the bomb of a shot we saw in his KHL highlights. Will he hit the 40-point mark next year? Probably not, but I’d argue that he’s very capable of repeating a 35+ point season if he stays healthy.

The Contract

Zaitsev’s a unique case, which makes him so tricky to evaluate. On the one hand, he’s playing his first year on North American ice and is still adapting to the NHL after spending his entire life playing hockey in Russia. We see so much natural talent when we watch him play, so many get the feeling that he’s bound to improve next season. Although defensemen tend to peak around age 24, it’s worth noting that this is a trend, and I would argue that he’s more likely to peak closer to age 25 or 26 considering his special circumstances. On the other hand, we have evidence of him struggling to break out of his zone and defend against opposing zone entries at the NHL level, which are very important aspects of driving results as a defenseman. Are there QoC factors that play into this? Absolutely, and considering how insane his matchups were this year, it’s difficult to know how much of an impact his QoC has on these results.

You also have the rare contract situation of a player one year away from UFA who can use KHL offers as leverage in contract negotiations. For example, if Zaitsev didn’t get an offer that he liked from Toronto this offseason, he could easily go to the KHL for a year or two and come back to the NHL as a UFA. So despite only playing one NHL season, this isn’t your typical “rookie”, and it definitely isn’t your typical RFA. So how much do you pay that player? Well since Zaitsev’s situation makes him a “pseudo-UFA”, we can use Matt Cane’s UFA projection model to predict how much NHL teams would have been willing to pay for him on the open market:

Based on his age, point production, and PK usage, the model projects him to be worth an AAV of about $4.7 million on the open market. Considering the Leafs have him locked up for $4.5 million moving forward, that’s not a bad number at all based on recent comparables. I used to make the argument that the Leafs didn’t have to re-sign Zaitsev and could have called his KHL bluff, but then I started to look into the alternatives. The list of NHL calibre RHD under the age of 30 available as UFAs this summer is pretty bleak:

In a sense, Toronto almost had to re-sign Nikita Zaitsev, so it’s understandable how his agent was able to use this leverage to negotiate a contract that most of us can agree is a pretty favourable one for his client. Although the Leafs can explore trade options this offseason (with names like Manson, Tanev, Dumba, and Pysyk potentially available), it’s tough to imagine them improving their defense on the right side next season by letting Zaitsev walk. He’s not easily replaced, and you’d likely have to give up significant assets to acquire a player of similar value. For these reasons, I’m beginning to come around on the contract, especially the AAV.

Now, the biggest concern most people have with the contract is the term. I genuinely like Zaitsev as a player (if you couldn’t tell from the previous 2,000 words), but I always worry about signing players who aren’t a part of your “core” to long-term deals that take them into their 30s. I understand that Zaitsev’s in a unique situation that may allow him to peak as a 26-year-old (maybe even 27 or 28), but based on the research, I’m not sure how likely that is. Realistically, I would wager that this upcoming season will probably be his best year (after fully adapting to the NHL), with his decline beginning shortly afterwards. With that being said, sometimes nerds like me tend to exaggerate the extent of a player’s decline into their late 20s & early 30s:

Is he going to regress throughout the course of this deal? Based on the evidence, yes. Is he going to be completely unusable in years 5, 6, and 7, to the point that the Leafs will have to send him down to the AHL and pray that he hops on the next flight to Russia? No, probably not. In all likelihood, what we’re looking at is a contract that actually looks pretty good for most of its duration, and will look slightly worse towards the end.


I don’t think Zaitsev is better than Gardiner or Rielly. I don’t even think he’s a top pairing defenseman. What I do think is that he’s an effective #3 defenseman who played against ridiculous competition this year. Predictably, his shot metrics didn’t turn out too well, but when you look at his results with a measure that tries to account for his absurd competition (GAR), he comes out looking like a very solid #3D. His point production was great this season, and I would argue that he’s capable of repeating it next season given his passing metrics, PP ability, and natural scoring talent.

Unfortunately, points are always something that you have to pay for on the open market, so he’s not exactly the bargain that Gardiner was on his last contract. Considering the KHL leverage situation that we touched on earlier, I think it’s fairer to consider Zaitsev a UFA (or “pseudo-UFA”) when evaluating this deal. Like any UFA contract, you were never going to get an absolute steal in terms of value (Anton Stralman notwithstanding). What the Leafs ended up with is a contract that’s going to pay Zaitsev market value for the next 4-6 years, and after that we’ll have to wait and see. Personally, I’m still not the biggest fan of the risk associated with a 7-year contract, but this deal isn’t anywhere near the catastrophe that I thought it was when I emotionally reacted to it on twitter a few days ago.

I won’t go as far as to say that I like the contract, but after digging deeper into things, I think it’s a perfectly reasonable deal considering the circumstances. What I will say is that I really like Zaitsev as a player. I would make the argument that he’s a legitimate #3 defenseman in the NHL based on his point production and play-driving ability (after accounting for his extreme QoC). He’s a player who can put up points, provide value on the PP & PK, drive penalty differential, and drive goal differential when he’s not given absurd usage. If the Leafs can acquire a top pairing RHD this offseason, it will allow them to move Zaitsev into a more suitable 2nd pairing role, where I believe he’ll be able to thrive moving forward.

This site uses Akismet to reduce spam. Learn how your comment data is processed.

  • MartinPolak

    One flaw I find many of these online analytic bloggers (though to the credit of author not here) make is a misleading assertion that Zasitev is the 53rd highest paid defenceman but he is not the 53rd best dman. I won’t address if he is a top 50 dmen, but these so called analytics practitioners are moving the goal post and are passing misleading data as factual analysis.

    A fairer way would be to comparing zaistev’s salary next year when cap is 76M to the salary of players who were signed with a much lower cap by adjusting for cap inflation.
    Zaistev at 4.5M is 5.9% of cap and shrinking each year going forward.
    Rielly is 6.9% of cap (5M) the year he signed.
    Josi/Gardiner were 6.2% of cap they year they signed.
    Klingberg 6% of cap the year he signed

    The reason this is important is it shows the cap “inflation” adjusted salary rather the nominal version that always presents the newly signed player in the worst light. My point for those new to analytics is you need to be careful when reading these stats as sometimes folks present “alternative facts” to support their biases. I wish there was some sort of quality control for analytics because of these twitter folks like the article I read yesterday on zaistev resemble the climate change deniers with how they craft and skew their analysis.

    Basically, a group of these twitter analysts have become people with strong dogmatic opinions who think their opinions are “better” or “true” because they have data or analysis. You folks of the analytic community need to do better with these renegades before they all the good work done up to now.

      • Haha I like that quote. It’s so easy to cherry pick the right stats and say “see Player X is good” (guys like Mike Kelly, David Johnson, and David Staples do it all the time). What’s difficult is trying to take all of the data into account and come to an objective conclusion. It requires a lot of time & effort to fully assess a player, since you have to take in all the available info (including the infamous “eye test).

        For what it’s worth, I thought that Draglikepull did an excellent job using on a bunch of different metrics to form his argument. He pulled zone entry & exit data, passing data, shot rates, point production, with-or-without you’s (WOWY’s). He definitely wasn’t trying to cherry pick, that was a comprehensive piece. The tricky part with a player like Zaitsev is that he’s very unique (good at some things, poor at others, and when you mix in the extreme Quality of Competition, it muddies the waters quite a bit).

        Sometimes even the nerd world is going to disagree on a player, and Zaitsev appears to be the right mix of small sample, extreme Quality of Competition, and peculiar zone exit/zone entry/passing data that creates this kind of divide. It don’t think either side of the debate wrong (I agreed with almost everything DLP wrote) – if anything I think a player like this is great for sparking discussion and helping us all learn more about player evaluation. I learned a lot by reading his piece, and I would hope that my rebuttal gave readers something to think about as well. #KnowledgeIsPower 🤓

        • MartinPolak

          I enjoyed your piece and maybe I’m being too hard on others in the analytic community, but too many use similar methods to what the climate change deniers use (shoddy peer review inside their echo chamber, cherry picked data, etc). I’m probably annoyed at the lack of accountability, by those in analytics, after the nonsense that has been published on Bozak, Corrado, Martin, Polak and Hunwick that is now debunked after this season. And I understand what you say about being open minded to different opinions, and I try to myself. But I am probably in the minority here, because I feel some folks, like the climate change deniers do not really deserve to have a voice.

          • lukewarmwater

            Yeah Martin these climate deniers really, really are a pain in the ass. Now take a decade ago or so, the state of Florida had three huge hurricanes in the month of June alone. Well next season the two leading experts on climate change, namely Canada’s David Suzuki and Al Gore who is a hockey fan as he used a hockey stick curve in his infamous sermons, preached from the pulpit that they guaranteed that poor Florida would have even more hurricanes the next season. Yep people believed them, numerous conventions were cancelled, millions refused to go to Florida the following year and Florida tourism lost 10 billion. Oh you are asking how many hurricanes did Florida get the next year. NADA, ZILCH, ZERO, Kind of reminds me of some of the analytic geeks records of predictability. Just saying.

  • JB#1

    A well written piece, thanks for this Ian.

    I’ll be interested to read what the other “mature” readers, old foggies(?), like myself think of this piece but I was finally able to make some head and tails of these new-fangled advanced stats. Though I still rely heavily on the eye test to form my opinion of players.

    Last year around this time, after having devoured what video and articles I could find about Zaitsev, I put out my opinion that he would be a solid 4th D-man, with hope that he might out-perform and be a satisfactory 2nd D-man. Well, let me tell you, I was ridiculed for that opinion.

    After his 1st season, it is good to read that a lot of these advanced stats back up what my limited eye-test could determine and that it seems I was proven right about Zaitsev. For all my detractors out there – stick it! lol

    For the first 5 years of this contract, I see Zaitsev ably patrolling the right-side of the 2nd pairing and, if need be, moved down to 3rd pairing for years 6 and 7 of his contract. Time will tell…

    • The Russian Rocket

      The thing to remember about year 6 and 7, with a player like Zaitsev, is that when his above average foot speed begins to decrease, he’ll still have his hockey intelligence and his stick skills. A crash and bang D-man is going to look a lot worse late in their contract. For Zaitsev, I expect it won’t be as noticeable.

  • Harte of a Lion

    Thanks Ian for the well written article. I was excited when rumours of Zaitsev signing with the Leafs surfaced and after watching every game this year, I am glad the team locked him up.

  • lukewarmwater

    I wish we could have a statistical analysis of how many other N.H.L. teams would have liked to have signed this talented defenceman who just played his first year in the N.H.L. against usually the top two lines and with two partners not known for their defensive game, namely Gardiner and Rielly. The Russian comrade did fine for his rookie year in the N.H.L. and was rightfully rewarded. Again a REAL IMPORTANT STATISTICAL ANALYSIS. Leaf management have our top three defencemen for a few years at under $14 million. Put that in your analytical cup and drink it.

    • DukesRocks

      Luke, as usual, you bring up a good point. What if Zaitsev was playing for a solid veteran team like Chicago. How much different would his metrics look? The Leafs should be thanking their lucky stars he was playing on a team with a bunch of rookies or they would be paying him 6 to 7 million a year.

      I love JB#1 line, “There are three kinds of lies: lies, damned lies, and Statisitics”.

  • Ty

    Great article to look deeper into the stats. People are fixated on corsi but it only has an r^2 of 0.11 with respect to team winning percentage so it only explains 11% of the variation in winning. Regression to the mean is also often quoted, but some individuals are aften above or below the mean consistently and with statistical significance based on their ability. There is always a distribution about the mean. The top players listed in terms of corsi don’t match with the top players in the league. You have to build better more predictive models. For possession count the actual minutes of possession in each zone instead of using a proxy line shots. For goal scoring use a direct measurment like goal scoring differential and then break it down using regression/machine learning.

  • A comment I made on DragLikePull’s piece that I think still applies given this article didn’t fully address it either:

    “All we heard this season is that Matthews was good but, if he had one flaw, it was in the faceoff dot. He needed to get accustomed to the league and his opposition’s tendencies. He already had quick hands and won his relative fair share but a minuscule improvement in hand-eye should get him there shortly. You hear the same things when a players jumps from the AHL. Marginal upgrades in speed and mental processing.

    Zaitsev needs to learn his opposition’s tendencies. Who are the preferred zone entry leaders on each team and how do you defend them? Additionally, his rookie status means he likely defers on zone exits to someone like Gardiner or Rielly. I know their ages and expectations are different, but my expectations of Zaitsev last season aren’t considerably higher than Carrick as opposed to this article moreso comparing him to Rielly or Gardiner. Sure, his expectations are higher now given the contract, but I find it unfair to judge his performance last season so harshly when his contact only takes effect next year.

    On a final note, I wonder what the effects of the opposing wingers have on these stats. What’s the chances that he faced a higher quality of competition on the opposing left wing compared to what Rielly or Gardiner faced on the right wing? I’ve done some research before for fantasy hockey reasons and one thing I noticed is that while there is lots of offensive depth on the right wing, some of the top winger producers play on the left side (Ovechkin, Benn, Gaudreau, Pacioretty, Hall, Forsberg, Panarin, etc.). I’m sure the balance changes every few seasons but I’ve never seen this considered, especially given Zaitsev was playing directly against these top line players in his first season.”

    • Ty

      good point direct measurements instead of indirectly using averages.

      have to do a sanity check on some of stats. also all stats should have a importance hierarchy like winning percentage first. that’s probably why mangment keyed in on Andersen. babcock always focuses on 6 pts in 5 gms first thats a 5 game win percent of 0.600. Then he looks at goal differential and its lower level stat chances created vs chances given up and getting to the scoring areas to increase the probability of finishing. Then he looks at “managing the puck” which they probably have different stats for like possession time, giveaways in a particular zone and time (know where you are, know the score and know the clock), good passes and breakouts. stats for being “strong on the puck” like takeaways, entries, box outs, gap control, finishing checks that lead to turnovers. Uses matchups so probably has matchup stats and plays five man units a lot so has a large sample of stats for each unit. probably has a stat quantifying how much they get slapped around. Has all of the player health and physical stats so they know what they need to work on in the summer.

      they have a whole department whose job it is to do that and probably outsource some of it too, they use biometric meters and probably do video recognition by now.

      • Ty

        don’t forget to consider the entertainment factor/variable. These guys are in the middle of an arean beting each other up like gladiators. “Are you not entertained? Is this not why you are here?” Maximus, Gladiator

        • lukewarmwater

          They don’t come any tougher than Russel Crowe who throws the meanest phone at a hotel staff member than anyone. Google his infamous melt down well down under when he didn’t get the service he deserved in a hotel one evening. In a bizarre way Crowe was entertaining.

  • jimithy

    Good point, but I didn’t know what you were talking about. Do you realize I was drunk when I wrote this. My punctuation cannot be accountable. Rememberv Qe neer said goodbe – GOODBYE …