Every sport has a few teams that can’t seem to get out of their own way. In hockey, the Ottawa Senators wear this crown, but the Buffalo Sabres are giving them a run for their money. All the negative press around the Sabres recently got me looking at their painful draft record and wondering, how much different might the Buffalo Sabres look today if they had drafted by something simple like NHLe instead of picking the players they did?
NHLe vs. the Buffalo Sabres Drafting
What is NHLe?
NHLe is a simple yet powerful concept that means NHL equivalency. The general idea is simple, how much is a point in any league worth compared to a point in the NHL. It can also be called a league’s “translation factor” because it roughly translates scoring in any given league to NHL scoring. For example, the KHL has an NHLe of 0.63. This tells us 10 points in the KHL are worth about 6.3 points in the NHL. Exact estimates vary, but this post will use CJ Turtoro’s “Network NHLe” translation factors.
This becomes a useful tool for evaluating prospects. A constant challenge for scouts is adjusting for the quality of various leagues. For example, it can be difficult to intuitively adjust for the difference between the OHL and KHL. With NHLe, those scouts can get a reasonable comparison of two prospects scoring rates, no matter what leagues they play in.
The NHLe Method
With that being said I wanted to test, would an “NHLe method” have out drafted the Buffalo Sabres historically? The idea is simple. For each Buffalo Sabres pick, simply select the player with the highest NHLe per game played among the available drafted prospects. If the Sabres did select the highest scoring player available, don’t change the selection. Then compare the NHL outcomes of my method’s selections to the Buffalo Sabres picks from 2007-2013.
For this post, our selection method will have many drawbacks and one advantage. The first drawback is that I have not accounted for age. Adjusting for age can be crucial for prospects because a year or two represents a large chunk of their lives. So, in 2007 for example, Patrik Zackrisson will be the most desirable prospect, not Patrick Kane. Kane’s NHLe was slightly lower than Zackrisson, but Zackrisson was a year and a half older. As a result, nobody ranked Zackrisson nearly as high as Kane. Our NHLe method will not be able to account for this and would pick Zackrisson first overall in 2007 if given the chance.
The next drawback is the NHLe method will have no contextual information. Points per game are largely a function of five things,
- Player talent
- Quality of Teammates
- Total Time on Ice
- Total Time on the Powerplay
Four of which I consider contextual. Scouts can try to discern what % of prospects production is from each category, this method cannot. It also misses out on any physical information about the prospects.
The third drawback of this method will be no positional adjustment. This method will never draft a goalie and will have a 0 percent chance of selecting high upside picks like John Gibson. Additionally, it will be unable to discern the difference between defencemen and forwards. So take Ryan Ellis for example. He scored at a historic rate for a draft-eligible defenseman in 2009 OHL season. This gave him the statistical profile of an elite defensive prospect. Despite this, the NHLe method would still take a 20-year-old forward in his draft plus 2 season over Ellis if the forward outscored him, even just slightly.
The final drawback is incomplete information. The NHLe method only gets to use a player’s scoring rate from their primary league. So, imagine a player split time between the USDP and USHL. The NHLe method will only get to use a player’s statistics from the league they played in most. The lone exception being (usually with KHL defenders) where prospects played in 2 leagues but produced 0 points in one league. Here I used the player’s output in the league where they scored.
The only advantage our NHLe method will have is knowing which players got drafted. Rather than sorting by all available prospects like teams had to do, the NHLe method will only select from players who were drafted that year. This means the method cannot select the Artemi Panarin‘s of the world but probably helps lower the false positive rate of high scoring players. All in all, it should be the single advantage for the NHLe method. This is an unrealistic advantage but it should absolutely not allow a “sort by points per game” method to beat an actual NHL team’s drafting record, but let’s see what the differences are.
NHLe vs. the Buffalo Sabres Drafting: 2007
The first draft is the Buffalo Sabres forgettable 2007 draft.
(Note the statistics represent player’s cumulative statistics in their seven years after the draft from Evolving Hockey. “ADJ value” uses a combination of Evolving Hockey Standing Points Above Replacement and Expected Standing Points Above Replacement. From there, the metric’s values pro-rated to per 82 games and regressed towards replacement level for players with fewer than 100 NHL GP. It is the same method used in my over/undervalued leagues post to define player output).
The Buffalo Sabres did not have any high draft picks in 2007. Instead, they picked late and often. This yielded them an AHL superstar in T.J. Brennan and solid NHLer Paul Byron. Altogether the Sabres got more or less what you would expect given their picks in 2007.
On the contrary, our NHLe method starts miserably. It starts something you’ll notice as a common trend, aggressively reaching for old and obscure prospects with its early picks. Turns out, not selecting anyone drafted in the top 150 is a risky bet. Our NHLe method only lands Dustin Jeffery as a real NHLer. Jeffery performed reasonably well in a limited role, but his actual results never caught up to expectations. Advantage Sabres.
Next up is the 2008 NHL draft where we get to see some high picks.
The Buffalo Sabres had two picks in the first round, and they selected solid NHL players each time. First, they selected Tyler Myers. Many consider Myers overrated, but he is far from a failure in the NHL. Then the Sabres selected Tyler Ennis. Using my value method Ennis was (just barely) the most valuable player selected by the Buffalo Sabres in the entire sample. It is easy to forget how well Ennis played to start his career. Unforatley for the Sabres, outside the first round they whiffed on every pick (sounds familiar).
Then NHLe method begins to show some promise. Unfortunately, it first reaches dramatically for two picks who never made the NHL. After the first round, it finally gives us a player NHL teams considered worthy of a top 150 pick, and our first real NHL player, Jori Lehtera. In the later rounds, the NHLe method adds Ben Smith and nails its first and only successful defensive prospect, Jason Demers. The inability to draft defenders has killed the Sabres over the years, and picking Demers late may have been huge. Neither draft was amazing, but the NHLe method has a distinct advantage using catch-all metrics instead of games played.
In 2009 the NHLe method finally does something the Buffalo Sabres could. Find a legit first-line forward.
Yes, the Sabres did select some sneaky good players like RAPM superstar Marcus Foligno, but Mike Hoffman is in a tier above any Sabres pick in this sample. While not exactly a superstar, Hoffman has been a top-tier NHL forward multiple times. Outside of Hoffman, the NHLe team adds some depth with Jordan Schroeder, and busts on the rest of its picks. That doesn’t matter though, because finding one first-line winger is more than the Sabres were able to do in seven years. This gives the NHLe method the edge again.
2010 was a year to forget for both the Buffalo Sabres and our method, with one key exception.
The Sabres had nine picks and selected one NHL player in 2010. While Mark Pysyk was a good pick at 22nd overall, the NHLe method found a diamond in the rough. After adding some depth with Chris Wagner and Tye McGinn, the NHLe method then found another first-line forward. His size crushed his draft stock, but the NHLe method doesn’t care so it selects Brendan Gallhager 143rd overall. By the Gallhager pick, the NHLe method has already found about as much value as the Sabres picked in total. Which is hilarious because there are still three drafts left.
In 2011, the NHLe method pummeled the Buffalo Sabres by finding a first-line forward for a third straight draft.
In the seventh round, our NHLe method selects late-round steal Ryan Dzingel. Dzingel produced more than the Sabres 2011 draft class by himself and wasn’t even the best player picked by the NHLe method. Ondrej Palat is the real steal. We select him 101 spots higher than his real draft position and he provides more value than any pick in this exercise. Advantage NHLe, again.
After getting stomped by a simple method for four straight years, the Buffalo Sabres finally add some value beyond an incomplete spreadsheet in 2012.
With a handful of early picks in 2012, the Buffalo Sabres didn’t do well, they just did better than my spreadsheet. Nothing against Mikhail Grigorenko and Zemgus Girgensons, but they leave plenty to be desired for mid-first-round picks. That being said they still provided more value than Jimmy Vesey and Thomas Spelling. With the depth picks neither side did much of anything. Although the NHLe method did select Nikita Gusev, a what-if with some upside. Altogether, the Sabres finally get another win.
These drafts mercifully come to an end as both methods fail to find much talent despite picking early and often in 2013.
This draft is a great example of how evaluating draft picks with games played can lead you astray. Rasmus Ristolainen and Nikita Zadorov have both played plenty of NHL games and may be considered “successful” picks as a result. Unfortunately for the Sabres, hockey is about scoring more goals than you allow not showing up to the rink. Because of their terrible isolated on ice statistics, this system considers both busts even though they played plenty of games.
The NHLe method runs into a similar problem here. Its 2013 picks played more NHL games within seven years of their draft than any of the NHLe method’s other draft classes. As for actual value, it ranks either dead last or second last among its classes depending on how you calculate “value”.
In sum, the anecdotal evidence sure favours the NHLe method. Due to the scarcity of roster spots and time on ice, Ondrej Palat, Brendan Gallagher and Mike Hoffman are probably a more valuable trio than the Sabres 56 draft picks combined. As for the cumulative statistics, we see how badly a simple method trashed a real NHL team’s draft record.
(Note the cumulative statistics treat below replacement level selections as 0 value, not negative value)
When I started this exercise, I had a hunch this simple method would beat the Buffalo Sabres. Then I realized the NHLe method involves throwing away every top-40 pick. Usually wasting them on some prospect who was one-to-two years older than everyone else, so I figured the Sabres must have been better. Wow was I wrong. Using my regressed SPAR Index introduced above, the Sabres got roughly half as much value as expected, and half as much value as picks based on points per game. Using raw (Evolving Hockey) WAR totals instead, we see the same story.
The differences were mainly driven by the NHLe method selecting high-risk high reward forwards, and it paying off multiple times.
Of course, much of what is important in hockey cannot be quantified. There’s context, psychological makeup, physical tools, chemistry and much more that scouts and analysts can discern. Because of all these factors, teams certainly know far more than the public on many occasions. And yet the simple heuristic, “draft the highest scoring players among all drafted prospects” would have absolutely crushed the real-life Buffalo Sabres draft record. The NHLe method found more NHL players (29 to 23) and more good NHL players (three definite first liners to zero) over a seven-year period than a hundred plus million dollar organization. I ignore things like development and give my terrible selection method the advantage of knowing who got drafted, but relying solely on (incomplete) points per game should not have found more talent than an NHL team over seven years and 50-plus picks. Score one for spreadsheet hockey.
For those curious, here is the full list of each player selected by either method with at least one Game Played. Statistics listed are within seven years of the players being selected.