Tail Gini
Tail Gini measures the loss tail like CVaR does, but it does not weigh the tail evenly. It mixes CVaRs at tail levels from 5% down to 0.01%, so the most extreme days count more than in CVaR (95%).
Mathematical Formulation
As an ordered weighted average
The weights c_k are riskfolio-lib's: they approximate the tail Gini integral with 100 levels. Every CVaR here uses the same fractional-tail estimator as CVaR (95%).
Reading it
Tail Gini is never below CVaR (95%) on the same returns, and the gap widens as the extreme tail gets heavier than the rest of the tail.
It is a daily figure, reported as a positive loss.
Like every metric on FolioLab, this value describes the run window. It is a measurement of past returns, not a forecast of future ones.
Where FolioLab reports it
results.<method>.performance.tail_gini_95: daily, over the run window.
Each value appears on the result page under Further Risk Measures, in the PDF and Excel reports, in the compare view and through the MCP explain_metric tool. A run stored before this measure existed does not carry it.
Only riskfolio-lib has this measure. FolioLab lifts its owa_tg weights under the BSD-3 licence and matches riskfolio-lib 7.0.1's TG to rounding on frozen test series.
References
- Furman, E., Wang, R., & Zitikis, R. (2017). "Gini-type Measures of Risk and Variability: Gini Shortfall, Capital Allocations, and Heavy-tailed Risks." Journal of Banking & Finance, 83, 70-84.
- Cajas, D. (2021). "OWA Portfolio Optimization: A Disciplined Convex Programming Framework." SSRN Working Paper.
Not investment advice. Past performance is not indicative of future results.