Toronto Maple Leafs: Kyle Dubas and the Use of Analytics
![TORONTO, ON-Toronto-DUBAS.The Maple Leafs announced today the promotion of Kyle Dubas to General Manager. Brendan Shanahan was on hand for the announcement..October 30, 2012. (Rene Johnston/Toronto Star via Getty Images) TORONTO, ON-Toronto-DUBAS.The Maple Leafs announced today the promotion of Kyle Dubas to General Manager. Brendan Shanahan was on hand for the announcement..October 30, 2012. (Rene Johnston/Toronto Star via Getty Images)](https://images2.minutemediacdn.com/image/upload/c_fill,w_720,ar_16:9,f_auto,q_auto,g_auto/shape/cover/sport/ba991a46fc380255eb4c50fb36fe88a317933dc631ee895d6b2ea304c82f5d69.jpg)
As one of the younger general managers in the NHL, many have labeled Toronto Maple Leafs GM, Kyle Dubas as the so-called “analytics GM,” but is this entirely true?
Over the past year or so, the Toronto Maple Leafs G.M’s has made several moves that do not seem aligned with the general consensus of the ‘hockey analytics community.’
Two moves that Dubas made seem almost antithetical of what analytics supposedly tell us. These were of course the acquisitions of Tyson Barrie and Cody Ceci. Both players are defensemen that play plenty of minutes without driving play at an effective level.
Furthermore, in Barrie, the Leafs acquired a high scoring, low defense defenseman, which tends to be something NHL teams overpay for unnecessarily, which is a market inefficiency. One of the main purposes of hockey analytics is to identify and avoid these, something Dubas seems not to have done in this situation.
Ceci and Barrie both are players who look bad by the numbers, yet Dubas still paid a hefty price in both cap hit and by trade to acquire these players.
Generally, people using analytics would be finding defensemen who are underrated by both points and TOI, not the opposite. Dubas decided to get two defensemen who represent the opposite of undervalued in the analytics circle.
For the most part, teams pay more on the cap for offense than they do for defense. This is where I find it really inefficient of Dubas to go all-out of offense and not prioritize players who have strong defensive impacts.
When Dubas started off with the roster, it was obvious the team was offense heavy, but the issue I find is that Dubas has catered to guys like Barrie who are empty-calorie, offensive players.
John Tavares is obviously a spectacular offensive player, but even he has really poor defensive metrics, and paying a player who is bad defensively into his mid-thirties 11M seems counterproductive.
The use of the 11M cap hit on Tavares could probably have been allocated to a player who isn’t a drag defensively and not going into his thirties.
The acquisition of Tavares seems to lead the Leafs more onto the idea of paying for players who produce offense at a higher cap hit over players who are sound defensively at a lower cost.
Obviously the Toronto Maple Leafs were never bound to get the perfect hockey player in Mark Stone but the Leafs seemed to lack concern over Tavares’ poor defensive game and his age when they signed him.
This is all not to say that Tavares isn’t a fantastic player, but the addition of Tavares is not a clear win in the whole scheme of things.
Another important point is, many believe that the Leafs have some kind of internal analytics on some of these players that show they are effective, unlike the public models, but this is again problematic.
Thomas Drance explains here why John Chayka wasn’t an analytics GM, which the point can stand the same for Dubas: “Unlike Dellow, Tulsky, Metcalf, Parnass, and co. Chayka didn’t publish publicly. His statistical work wasn’t reviewed in academic circles or by the online community. Hockey’s statistical analyst community has tended to be deeply skeptical of black box work”
I will go as far as to say that any sort of statistical work used in any industry should be peer-reviewed just like any other research article.
Some people who work with data have said that if private data completely changes the views on a lot of players, it may be not be very good. So if Dubas is telling us that Ceci’s metrics were great, but he has the only data that supports this, it makes you skeptical.
Moving on from the Ceci and Barrie ideas to even a more basic concept, let’s discuss the Mitch Marner situation. Dubas signed Marner for far too much money and everyone is aware of this.
In his contract year Marner had an 11.37 on-ice shooting percentage which is extremely high and unsustainable and the analytics community has identified that players who shoot at abnormally high rates are bound to get overpaid relative to their value. (naturalstattrick.com).
It seems as though Dubas didn’t take enough into consideration when evaluating Marner’s value to the team because, at the current cap hit, Marner is being extremely overpaid.
It would have been better for Dubas to negotiate more effectively of course, but if Dubas was truly a big believer in analytics he would have tried to push for a bridge deal which didn’t seem to be the case from an outsider’s point of view.
Moving to the opposing side of the spectrum, there are clearly many moves Dubas has made that coincide with analytics. Such as in trading for Kyle Clifford who has fantastic analytics.
As a whole, the point I’m trying to make here is that Dubas shouldn’t be labeled as an “analytics GM” because a lot of his moves don’t make sense if viewed in that light. We tend to label people with superficially representative labels, when in fact things are far more complicated.
Kyle Dubas obviously uses analytics, but his critics seem to view him as nothing but a math nerd, when, in reality, a lot of his moves are criticized heavily by the ‘analytics community’ and do not seem to correspond with much of their thinking.
Next. Don't Romanticize the Failures of Lou Lamoriello. dark
Over Dubas’ tenure, It is truly hard to say how much emphasis Dubas puts on analytics, especially given he has only been in charge for a few years. This offseason will likely answer plenty of questions regarding Dubas’ use of analytics as a whole.