As the beginning of every NBA season approaches, one of the primary things every fan wonders is: how many games will my team win this year? Most of the predictions one sees out there are based on some combination of guessing, opinion and "gut feeling". However, if you are more inclined to go the analytical/mathematical route, there is a metric that will shed some light on the expectations one should have for a given NBA club.
Expected Wins (also referred to as Pythagorean Record) is a metric that uses mathematical principals to predict how many games a team should win, based on how many points they score, and how many points they give up. This sounds like a simple concept, but it tends to yield accurate data in terms of predicting a team's win-loss record.
Expected Wins started out as a baseball metric, developed by statistician Bill James. Several basketball statisticians, including Daryl Morey, Dean Oliver and John Hollinger have adapted James' concept to a form that can be used for NBA analysis.
How Expected Wins Are Calculated
The calculation for Expected Wins is not complicated, although the thought behind it is a bit deeper than it first appears. The formula is a modified version of the geometric principal called the Pythagorean Theorem (a² + b² = c²). It turns out that by selecting the proper exponent, this variation of the Pythagorean Theorem can yield an expected winning percentage for a team at any point in the season (although as with any statistical estimate, the more data you have, the better).
The exponent used in the formula needs to be at or near the average margin of victory for an NBA team. There is some debate in the NBA analytics community over just what the exponent value should be, but the exponents commonly used are all between 13 and 17.
How To Interpret Expected Wins
Basically, the team record predicted by the Expected Wins formula is the record one can expect if there is no particular good or bad luck, or any other outside factors that might change expectations for the success of a team. When you look at how many points a team is scoring versus how many points their opponents are scoring, there is a sound mathematical expectation of how many wins and losses a team with those statistics should have.
If a team's actual performance varies much from the Expected Wins value, many analysts chalk those differences up to luck. However, we believe that good teams come through in the clutch on a pretty consistent basis, and poor teams, even if they perform well throughout most of a game, tend to fail in the clutch. Further research can certainly be done on this, but we suspect that crunch time performance may play a part in how successful teams are at maximizing their win-loss record based on a given scoring differential. Coaching is another likely factor in how a team performs compared to its Expected Wins value.
Benefits Of The Expected Wins Statistic
Having a way to assert how many games a team should win is a very useful tool. If a team is right at or very near their Expected Wins, one can conclude they are performing as they should. However, if there is a discrepancy between expected and actual wins, the team in question would seem to be underperforming or overachieving, and from that point, further scrutiny can lead to the discovery of why that is the case. There may be concrete reasons for this deviation, or in some cases, it might just be blind luck!
Negatives Of The Expected Wins Statistic
Although the principals behind Expected Wins are solid, it is, after all, a metric based on total points scored and total points allowed. If you are looking for in-depth information about a team's performance and the reasons for their performance level, this calculation is only a starting point.
Expected Wins - Behind The Number
The formula for Expected Wins is not complex, but as stated earlier, choosing which exponent to use is the tricky part. The excellent basketball statistics website Basketball-Reference.com uses 14 as their exponent value, so for the purposes of this discussion, we will use their version of the formula, which is:
This calculation will give us the number of team victories to "expect", and from this, we can easily determine the expected overall record and expected winning percentage for a given team.
For example, if a team has played 10 games so far in the season and scored 923 points and had 873 points scored against their Expected Win total over the 10 games would be 6.85 (or 7 if you want to round up). So if the team won more or less than the expected win totals, they would be either outperforming or underperforming expectations.
Everyone involved with NBA basketball, whether you are a fan, member of the media or team executive, wants to know what performance expectations are reasonable for their teams. Guesswork can only get you so far, so it is of great benefit that one can actually calculate how many games a team "should" win, based on offensive and defensive scoring totals. The Expected Wins metric gives us this type of meaningful information, and does so with a minimal amount of number-crunching.