Each statistical category that is evaluated at FantasyRef.ca is done using the same algorithm.
1st – The number of projections being compared are selected.
- We generate rankings for sets of 300, 200, and 100 projections.
2nd – The absolute error of the forecasters projections are determined for each player.
- The absolute error is calculated by taking the absolute value of the difference between what the forecaster projected and what the skater achieved.
3rd – The absolute error is summed in groups (25 for skaters, 10 for goalies).
Below is an example of the calculation to determine the absolute error in Scott Cullen's Pts Total projections for this Top 25 Players.
4th – The relative absolute error for each group of projections are calculated.
- For each group of skaters the forecaster with the minimum error is identified.
- The absolute relative error error is calculated by the following equation
5th – The complement of the relative absolute error is calculated.
6th – The points for each forecaster are determined.
- The number of points that each group of projections is allocated is determined by the percentage of Pts ( or what ever statistical category we are comparing) scored by that group of players. For example at mid season the Top 25 skaters had 1113 total points. This represented 15.13 % of the total Pts of the top 300 skaters, thus the Top 25 projections are weighted to 15.13 points. The group of skaters from 26 to 50 would get 12.17 points, and so on.
- The maximum number of points allotted to each group of skaters is then multiplied by the forecasters complement relative absolute error to determine how many points the forecaster receives. See calculation below.
- Below is a table showing a summary of the three calculations described above.
7th - The number of points accumulated by each forecaster for all skater groupings are then totaled.
8th - All forecasters total points are compared and ranked from greatest to least.
- In the example below 2 forecasters did not have 300 Pts Total projection so they were excluded from the ranking comparison.