Patterns · Classes
Does the class separate anybody?
Each class spreads its entities differently. The comparison runs against FORMULA 100, the largest one.
This ORDERS, it does not separate, and it does not say a class is harder: the class is CHOSEN at entry and the source does not publish its rules, so any difference mixes what the rule imposes with who decides to sign up for each.
| Class | Entrants | Q1 | Median | Q3 | Difference of medians | Shuffles that reach it | Above the median of FORMULA 100 |
|---|---|---|---|---|---|---|---|
| FORMULA 100 | 777 | 55 | 95 | 135 | — | — | — |
| UNLIMITED | 511 | 91 | 135 | 177 | 40 | 0 / 2,000 | 72.2% |
| LIMITED | 182 | 69 | 116 | 153 | 21 | 4 / 2,000 | 58.2% |
| FORMULA UNLIMITED | 77 | 80 | 125 | 168 | 30 | 1 / 2,000 | 64.9% |
| QRP UNLIMITED | 55 | 26 | 54 | 91 | 41 | 0 / 2,000 | 23.6% |
| QRP LIMITED | 2 | 17 | 123 | 17 | — | — | — |
The comparison shuffles the class labels 2,000 times and counts how often chance reaches the observed difference. A permutation rather than a table test, because the spreads are neither symmetric nor alike.
The shuffle uses a fixed seed: a number that changed on every reload could not be quoted or checked.
FORMULA 100 carries no overlap column: comparing it with itself gives 50 % by construction, and that 50 % would read as a result.
With fewer than 20 entrants no comparison is made: at that size chance reaches almost any difference.