Adjusted Stats

These posts make use of the concept of “adjusted” goals, assists and points as a (pretty inaccurate) way of comparing one era to another. The rationale is pretty simple: without attempting to adjust stats from one era to another, we are comparing numbers straight up, which means that we are saying that the 92 goals Gretzky scored in 1981-82 – one of the highest scoring years ever – are exactly 43 goals better than the 49 Gordie Howe scored in 70 games 1952-53 or exactly 27 goals better than the 65 Ovechkin scored in 82 games in 2007-08. And we’re saying that Ovechkin’s .79 GPG in ’08 is exactly .09 GPG better than Howe’s .7 GPG in ’53 and exactly .36 GPG worse than Gretzkky’s ’82 season. Why are these things not actually correct? Well, context is everything. Gretzky may or may not be the greatest hockey player (skater) of all time – we will get to that question in time – but he played most of his career – and amassed most of his records – in an era where scoring was totally wide open: nobody played (very effective) defense, goalies were proud of GAAs over 3 and numerous single-season scoring records were set for individuals, franchises and the league as a whole. It’s totally unfair to compare Gretzky to Howe to Ovechkin – or any other players across eras – without taking into account the differences from one era to another.

To do that, Hockey Reference created ‘adjusted stats’, which let us compare players across eras. Here is the full adjusted stats explanation. But the formula is essentially this:

  • Goals: actual goals multiplied by the “schedule adjustment” multiplied by the “roster adjustment” multiplied by the “era adjustment” = adjusted goals;
  • Assists: the only difference with assists is that the “era adjustment” is different;
  • Points: adjusted points are the same as regular points, merely the total of adjusted goals and assists.

The schedule adjustment is merely 82 divided by the contemporary schedule. The roster adjustment is 18 divided by the roster of the day. The era adjustment for goals is 6 – a standard number of goals per game – divided by the number of goals scored minus the goals scored by the player. The era adjustment for assists is 10 divided by the number of assists per game not including those by the player. This is a reasonable guess as to what the player would have scored if he played in a normal 82 game season.

There are a few major problems with these numbers we will encounter constantly:

  • No matter how much number crunching we do, there is no accounting for chemistry and circumstance: It might be a neat thought experiment to see calculate how many goals 2007-08 Ovechkin would have scored in 1981-82 or 1952-53, but nobody will ever know the reality because Ovechkin was born when he was born. He grew up with the family he had, watching the players he watched, coached by the coaches he had, at a time when sports medicine had grown in leaps and bounds even from the early ’80s, and he had the teammates he had. And the same goes for every other player ever.
  • Sample size: The NHL has had different schedule lengths over the course of its history, from around 20 games at the beginning to the current 82 game standard now. The players who played in the first half of the 20th century played far fewer games. As a result, there adjusted numbers are highly skewed. Of the top 30 NHL players in terms of adjusted PPG, only one – Wayne Gretzky – played since [what year?]. So early numbers must be taken with a massive grain of salt.
  • The NHL has never really done a good job of keeping track of information: The NHL started tracking minutes-played in the late ’90s; though we can adjust stats based on average scoring and the like we will only ever have offensive numbers to go by. This means that we really can’t evaluate defensemen, except by their scoring (and their personal and team’s goal differential). We have the same problem with goalies, since shots have only been tracked since [when?]. And we have no idea how much anyone played, of who they played with (to any consistent standard) which of course means we can’t really evaluate the quality of the offense. Another interesting problem: at one time there was only 1 assist per goal. And these are just some of the issues that revolve around the lack of data provided by NHL records. The short answer is that adjusted numbers aren’t really very useful. (But that won’t stop me.)