Luck doesn't factor into a team's rating. It's a metric that compares a team's record to what they deserved based on their game-by-game efficiency.

If a team is involved in a lot of close games, it shouldn't win or lose all of them. If a team wins all of the close games, they're viewed as a lucky. An unlucky team would lose all of their close games.

Luck is measured using Dean Oliver's correlated Gaussian method.

If this sounds complicated, it definitely is.

Oliver explains it as something similar to a bell curve.

`Luck = NORM (AdjOE - AdjDE) / SD(Rating Difference)`

| (Rtg-Opp.Rtg) | Win% = NORM |-----------------------------------| | SD(Rating Difference) |

The components:

- Rtg: points scored per 100 possessions (offensive rating)
- Opp.Rtg: points allowed per 100 possessions (defensive rating)
- SD(): statistical standard deviation of quantity in parentheses ()
- Var(): statistical variance of quantity in parentheses ()
- Cov(): statistical covariance of quantities in parentheses ()

Luck tells you the difference between expected winning percentage and its actual winning percentage.

Luck **doesn't** use Pythagorean Winning Percentage (Pyth) as the expected winning percentage. Pyth is calculated by a team's offensive and defensive efficiencies.

The correlated Gaussian method uses the distribution of a team's game efficiencies to determine the expected winning percentage. It includes both the **average margin of victory** and the **variation in a team's margin of victory**.

This method takes into account that the majority or teams play to the level of their competition.

In 2015-2016, Hampton was viewed as the second luckiest team in Division-I.

Hampton was 21-11. It's actual winning percentage was .656.

Using the correlated Gaussian method, Hampton's expect winning percentage was .499.

Hampton's actual winning percentage is .157 points higher than its expected. This translates into roughly **5 wins** that are attributed to luck.

If you take a glance at Hampton's results, it played 3 overtime games and one double-overtime game. It won all 4 of these contests, which can be viewed as *lucky*.

In comparison, the 2015-2016 Clemson Tigers finished 17-14. Clemson was ranked 339th out of 351 teams in luck.

Clemson's actual winning percentage was .548.

It's expected winning percentage using Oliver's method was .643.

This is -0.095 points lower than expected. This means almost **3 losses** were attributed to luck.

Taking a look at Clemson's game-by-game results, you'll find 10 losses by a 7 points or less. This can be seen as *unlucky*.

Luck isn't used when rating a team. It gives you an idea on how a team performs in close game and how it plays to its competition.

A very lucky team will likely be rated lower by KenPom.

Where an extremely unlucky team could be rated higher.

Using our example above, Hampton (lucky) was 229th in the final 2015-2016 KenPom ratings.

While Clemson (unlucky) was 48th in the final 2015-2016 KenPom ratings.

Question, concern, or spot a typo? Please send me an email and let me know.