Hockey Nerd

Hockey Nerd: Interpreting the Numbers

It is time for another update on the numbers.  In this column we are going to dive into the decimal number that MYHockey labels as a team's rating. A team's rating is an extremely insightful number that can that can be useful in understanding much more than a team's ranking.

Background…  MYHockey ranks teams.  Rankings are simply a sorted view of ratings that are calculated for every team where we sort from largest rating to smallest rating.  In this article we will skip over how that is done (previous articles have addressed this topic) and focus more on the numbers themselves.  A team's rating is more accurately described as its Average Performance Rating.  It is the mathematical computed average performance for the team.  It's based 100% on current season game scores.  We boil all your game scores down to a single number.  Although, that single number is the sum of two numbers, AGD and SCHED.  AGD is your average goal differential, an easily computed number that is essentially how many goals, on average, you win or lose by.  SCHED is the average performance rating of all your opponents.  So, your average performance rating is the sum of how your average goal differential and your average opponent rating.  It is a numerical representation of who you have played (SCHED) and how well you've done (AGD).

So now you know more about what a rating is and what it means.  But there is one more important element that requires discussion.  Because of the simplicity of our rating/ranking algorithm, the number the rating is more than a number.  It has some relative meaning that is critical if you want to fully understand MYHockey.  A difference of 1.00 points between two team's ratings is equal to a difference of 1.00 goals.  That can be explained in a different way.  If you team scored one additional goal in every game it played this season and gave up the exact same number of goals, your team's rating would be exactly 1.00 points higher.  Please note that the maximum goal differential actually makes this statement slightly less than true because some games that are decided by a goal difference of greater than 7 would not change if a team scored one additional goal.  So, to repeat, if you scored one additional goal per game, your team's rating would go up by 1.00 points.  

Another fundamental way to understand what a team's ratings means is from a comparison perspective.  If Team A has a rating of 90.0 and Team B has a rating of 89.0, it is fair to say that the average performance of Team A is one point or one goal better than Team B.  I have been known to say that MYHockey then predicts that Team A would beat Team B by an average of 1.00 goals if they played 100 games.  This statement is kind of true and technicall false, let me explain.  We are definitely saying the average performance of Team A is 1.00 goals better than the average performance of Team B.  My analysis of "predictive" systems requires that I now say that MYHockey is NOT a predictive system.  Technically it is most accurately characterized as "Descriptive Analytics".  Descriptive Analytics is the analysis of past events.  That is exactly what MYHockey does is numerically simplify and describe a series of past events.  We are a descriptive system.  "Predictive Analytics" uses descriptive analytical outputs but combines them in unique ways to predict the future.  For example, if we were truly trying to predict the outcome of a game between Team A and Team B to be played tomorrow morning at 7:00am at Team A's home rink, we would probably need to know Team A's average performance rating for games played in that arena.  We might incorporate analytics on their performance for early morning games.  We might want to understand who they are starting in goal, because that might help us better understand the team's average performance rating with that goalie versus their other goalie.  Understanding any lineup changes (injuries, illnesses, suspensions, etc) would also be necessary if we were truly trying to be accurate in our prediction of the results of that game.  We would need to understand these facts or conditions for both teams and build a model that will "Predict" the outcome of the game tomorrow.  We have NOT built this predictive analytical engine into MYHockey.

So, if you are a true numbers geek, you now know that MYHockey is not a predictive system.  But for everyone else, you can use MYHockey ratings to give you a general understanding of the playing level of an opponent that you've never seen play by comparing your team's rating to their team rating.  All things being equal, if your team rating is one point higher than their rating, we'd expect you to win by an average of 1.00 goals if you played them 100 times.  That doesn't mean you will win all 100 games (historical data shows us that you would likely lose 19 of those 100 games), but it's a mathematically computed average.  And we know that smart coaches use that information to plan a weekend or even weeks of upcoming games.

MYHockey is often asked, how much difference would it have made if we had not pulled our goalie and not given up one additional goal.  In the course of a fifty game season, scoring one extra goal or not giving up one goal will impact your rating by 0.02 points.  Our quick, anecdotal research into the issue says the average team could increase their rating by 0.10 to 0.20 if they never pulled their goalie in a season of fifty games.   Depending on the density of teams in your rating, that difference could leave your team ranked exactly where it is today or it might move them up a few spots.  Either way, the movement of a couple spots in the rankings is unlikely to change anyone's assessment on the success of your season. On the other hand, pulling your goalie and making a dramatic comeback to win a championship would be remembered for a lifetime.

In writing this story, MYHockey did a little research into how "predictive" is our system.  Below are some interesting historical numbers that we saw with amazing consistency over the past decade.

  • 34% of games have a final score within 1.00 points (goals) of what is the expected winning margin
  • 28% of games have a final score from 1.01 to 2.00 points (goals) of what is the expected winning margin
  • 19% of games have a final score from 2.01 to 3.00 points (goals) of what is the expected winning margin
  • 11% of games have a final score from 3.01 to 4.00 points (goals) of what is the expected winning margin
  • 8% of games have a final score that is greater than 4.00 points (goals) of what is the expected winning margin

Surprisingly, we didn't find much difference between squirts and midgets.  We did, however, find that lower scoring girls games were more likely to have winning margin that was close to the expected outcome than games involving boys teams.

We hope you found this informative and now have an even greater understanding of the numbers.

More articles like this...
Other articles of this type...