On this page
Define the loss random variable
L is loss over one common horizon, with positive numbers representing loss and negative numbers gain. The cumulative distribution F(l) counts probability at or below l. Confidence c is a decimal probability strictly between zero and one in the mathematical definition.
Use the first cumulative crossing
This is a declared lower loss quantile, not a sample-interpolation rule.
Keep equality at the boundary
In the original finite model, F(0)=0.70 and F(10)=0.95. The lower quantile at c=0.95 is 10 because equality satisfies the rule. A level strictly above 0.95 reaches the next loss, 50.
Read the equal-weight special case
For n equally weighted observed loss records sorted ascending, rank ceil(nc) implements this rule, counting from one. Interpolated sample percentiles are a different convention. The tool supports confidence 50%–99.99% and does not infer observation weights or exposure repricing.