Read on Twitter
A brief thread on my look into some simple hockey "analytics" 
:
Using Rob Vollman's @hockeyabstract "Stat Shot" from a couple years ago, I wanted to take a look at how translation factors could work at the NCAA level (i.e. if you could measure [junior league] -> NCAA)
: Using Rob Vollman's @hockeyabstract "Stat Shot" from a couple years ago, I wanted to take a look at how translation factors could work at the NCAA level (i.e. if you could measure [junior league] -> NCAA)
BACKGROUND: Translation factors (NHLe) are a way to compare a player’s stats between relevant leagues. It's a projection of one player's point production in one league, and how that player may then perform in the NHL in the following year(s)...
How NHLe works: A player's points per game (PPG) in one league is multiplied by that league's NHLe to determine their projected PPG in the NHL. Below is a few examples based on prospects from the KHL, OHL, and Hockey East (it doesn't always work this accurately, obviously)
Based on that idea, I wanted to look into how this could work with player translations into the NCAA. In particular, I wanted to see how you can "project" one player's performance in a league prior to the NCAA and into their freshman season. So, I tried to create "Big10e."
Nothing too complicated here. I just wanted to see how (Player A)'s final season prior to college translates into (Player A)'s freshman year (FY) in the NCAA.
So, for raw data, I used the most popular recruitment league (and highest sample size), the USHL, and subsequent freshman year performance in the NCAA. I'm not a statistician (major emphasis on this), but this math seemed simple enough to dig up.
Using @eliteprospects, I collected data from a player's final USHL season, and using Vollman's NHLe calculation, estimated Big10e. I chose the Big10 because (1) I'm a biased penn stater and (2) the league is new enough to take ALL available info since the league started...
RESULTS: In all, I collected 197 players' stats that played in the USHL prior to their FY season (ND data limited to when they joined the Big10). Also should note that data in the player's FY season is limited to Big10 conference games (to avoid easier OOC schedules).
Per Vollman's NHLe, each player's performance in the USHL/Big10 has to be "normalized" to that years' scoring rate in each respective league...
So here's what we get: Big10e is set at 0.55. On its face, a reasonable translation from USHL -> Big10. But, I also wanted to increase accuracy of the number by separating age (and eventually position). Age here represents the player's age during their USHL season:
As seen above, younger USHL players that commit as true freshman in the NCAA tend to be future NHL draft picks, and thus are likely to perform better at an NCAA level. Not exactly groundbreaking, but notable nonetheless.
For personal interest, I also broke down how each team has recruited USHL players and their respective translation factors. This may be more arbitrary, but it's a fun visual, particularly because PSU >>>>>.
How does this translate to this year? Since we are about half-way through the Big10 season, these are the estimated point productions (xPts) for Big10 FY players that played in the USHL last year, and what their full season projections are on pace for:
For these translations, instead of the blanket 0.55 number, I did go back and calculate each player's projection based on their age and position, broken down below:
For a team example, here's Michigan. Each player's USHL PPG is multiplied by that player's relevant Big10e factor depending on their position and age. That result is their projected output over 20 games as a FY player. Add 'em up, and you get what the class could produce.
In the end, this turned more into how a team's overall recruiting class may be translated, rather than an individual's performance, but the metric can be used for either. I saw it as a way to balance SR/JR lost to what may be incoming into the lineup (see: PSU's roster this year)
We'll see how the rest of the season (hopefully) plays out. Obviously this isn't a "normal" season, but for what its worth, I do look forward to keeping track/improving this metric as the years go on.
Finally, shoutout to some of my favorite follows on this analytics stuff that I don't fully understand but pretend to anyway: @IneffectiveMath @ChartingHockey @domluszczyszyn @JFreshHockey and @scootszn99 in particular for all his NCAA hockey work #HockeyTwitter 
Anyways, if you've made it this far, hope this makes some sense. I'm not sure if this particular analysis has been done already by more qualified people, but it was a fun little project to avoid law school responsibilities 
