LastWordArmy
Registered User
s seen by the Buffalo Sabres never-ending rebuild, the ability to find future talent at the NHL Draft can affect a franchise for years. This is why teams should always be looking for future market inefficiencies. They should look to correct for past biases at the NHL Draft. Various kinds of biases have been shown to causes groups of prospects to be undervalued. For example, I previously looked at age and showed forwards with later birthdates tended to be undervalued. In a similar vein, today we will be looking at how the league which these prospects have come from affected their value. More concisely, we will see, have players from certain leagues have been undervalued at the draft table?
NHL Draft: Undervalued Leagues
Evaluating Drafted Players
First and foremost, we must assign an “expected value” to each draft pick. If we don’t, we may conclude certain leagues are “undervalued” when in reality they just produce higher draft picks.
So, how will we define the expected value of draft picks? Well, let’s start with “value”. We will need to define the NHL player’s contributions to their teams in some way. Then we can look at how that value changes with draft position, on average. We will be using what I call “SPAR Index” to evaluate player output. This simply takes each player’s Standing Points Above Replacement (SPAR, found at EvolvingHockey.com), and their Expected Standing Points Above Replacement, then average the two of them.
From there the SPAR Index has been prorated over a full season. This means the values are per 82 games for skaters and per 60 games for goalies. This was to account for the uneven playing field in the data. I am using the value players produced in the seven seasons after the players draft as their output. This is because teams only retain a player’s rights for seven seasons after they drafted them.
A note on sample sizes
Ideally, we would have the data far enough back that I could use the player’s first seven seasons in the NHL. After all, this is when the team’s rights to a player expire, but that would greatly reduce the sample size for this analysis. So to increase the sample I have prorated the output to a per-season basis. Prorating data can give players who performed well in small samples extreme values, so the SPAR Index for each player has been regressed towards replacement level up to each player’s 100th game.
The Expected Value of NHL Draft Picks
Once we have defined player output, it’s (reasonably) easy to evaluate the expected value of each draft pick. We can simply look at the trendline of how much value (SPAR Index Per Season) is generated by each draft position. For example, we can see a general trend where output is usually high for the top few picks. After those first few picks, the value falls off much more slowly towards.
In our sample of drafts from 2007-2013, the first overall pick tends to add 2-3 SPAR Index per season. (1st overall picks add 2-3 standing points above replacement per season for 7 seasons following their draft, on average). Then the values begin to fall off very quickly. By the end of the first round, the players are usually only producing a fraction of what the first few picks do. After pick 30 or so, average values decreased at a slow rate until the end of the daft.
Using this data, I fit a logarithmic regression line to the data to generate out expected draft pick values.
This way we see a classic draft pick value chart. The first picks have large differences in expected value. Then after the early first-round value begins to fall off very slowly.
The Article Continues Here
Undervalued Leagues at the NHL Draft - Last Word on Hockey
NHL Draft: Undervalued Leagues
Evaluating Drafted Players
First and foremost, we must assign an “expected value” to each draft pick. If we don’t, we may conclude certain leagues are “undervalued” when in reality they just produce higher draft picks.
So, how will we define the expected value of draft picks? Well, let’s start with “value”. We will need to define the NHL player’s contributions to their teams in some way. Then we can look at how that value changes with draft position, on average. We will be using what I call “SPAR Index” to evaluate player output. This simply takes each player’s Standing Points Above Replacement (SPAR, found at EvolvingHockey.com), and their Expected Standing Points Above Replacement, then average the two of them.
From there the SPAR Index has been prorated over a full season. This means the values are per 82 games for skaters and per 60 games for goalies. This was to account for the uneven playing field in the data. I am using the value players produced in the seven seasons after the players draft as their output. This is because teams only retain a player’s rights for seven seasons after they drafted them.
A note on sample sizes
Ideally, we would have the data far enough back that I could use the player’s first seven seasons in the NHL. After all, this is when the team’s rights to a player expire, but that would greatly reduce the sample size for this analysis. So to increase the sample I have prorated the output to a per-season basis. Prorating data can give players who performed well in small samples extreme values, so the SPAR Index for each player has been regressed towards replacement level up to each player’s 100th game.
The Expected Value of NHL Draft Picks
Once we have defined player output, it’s (reasonably) easy to evaluate the expected value of each draft pick. We can simply look at the trendline of how much value (SPAR Index Per Season) is generated by each draft position. For example, we can see a general trend where output is usually high for the top few picks. After those first few picks, the value falls off much more slowly towards.
In our sample of drafts from 2007-2013, the first overall pick tends to add 2-3 SPAR Index per season. (1st overall picks add 2-3 standing points above replacement per season for 7 seasons following their draft, on average). Then the values begin to fall off very quickly. By the end of the first round, the players are usually only producing a fraction of what the first few picks do. After pick 30 or so, average values decreased at a slow rate until the end of the daft.
Using this data, I fit a logarithmic regression line to the data to generate out expected draft pick values.
This way we see a classic draft pick value chart. The first picks have large differences in expected value. Then after the early first-round value begins to fall off very slowly.
The Article Continues Here
Undervalued Leagues at the NHL Draft - Last Word on Hockey