Same VaR with different tail severity

Change one outcome while retaining probabilities

Both original models use probabilities 30%, 40%, 25% and 5%. Their first three losses are −10, 0 and 10. Model A’s final loss is 50; Model B’s final loss is 500. These are assumptions, not observed market alternatives.

Keep the same cumulative boundary

In both models F(10)=95%, so the declared lower 95% quantile is 10. The threshold calculation alone does not report how severe the final 5%-probability outcome is.

Inspect the worst 5% average

Use a summary suited to the question

ES distinguishes these tail severities because the selected tail consists of the final outcome. VaR remains useful as a percentile threshold, but equality of that summary does not establish equality of the full loss distributions.