Sortino vs Calmar: Target Shortfalls and Ordered Drawdown

Sortino and Calmar answer different historical measurement questions. A specified Sortino calculation uses return shortfalls relative to a target. A CAGR-based Calmar calculation uses the largest observed peak-relative drawdown in an ordered path. They need their own units, conventions and observation window; neither is a universal winner or proof of future safety.

Define Both Calculations before Comparing Them

The CFA Institute’s Sortino discussion explains why target choice and calculation variations matter. The published Empyrical implementation provides a documented CAGR/drawdown convention. These references do not establish that every dashboard uses the same inputs.

Downside Deviation Is Not the Standard Deviation of Losing Trades

d_i = min(periodic return_i − same-period target, 0)
downside deviation = sqrt(sum(d_i²) / N)
periodic Sortino = (mean periodic return − target) / downside deviation

This explicitly chosen discrete convention divides by the full included observation count. It does not centre only the negative returns around their own mean, and it does not silently replace N with the number of shortfalls. Other conventions need separate labels before comparison.

A shortfall is relative to the target, not necessarily below zero. For a 1% period target, a positive 0.5% return still has a −0.5% target shortfall. Use the same-period target in the numerator and denominator; an annual target must not be inserted unchanged into a monthly series.

No observed shortfalls gives a zero denominator under this convention, so there is no finite measured ratio. It does not prove that future shortfalls are impossible. Historical samples can miss outcomes that did not occur within the chosen window.

A Fully Specified Example: Same Returns, Different Order

Consider four hypothetical monthly simple returns, with no external cash flows, on an already after-cost basis. Use a constant monthly target of zero. Start a wealth index at 1.

Both sequences contain the same return values and the same constant target, so the mean, downside deviation and periodic Sortino are unchanged. Clustering the losses changes the ordered path’s drawdown. The final compounded gain is 9.2025% for both; annualising four monthly observations uses (1.092025)^3 −1. No later-year return is observed or forecast.

Keep Numerator and Time-Scaling Conventions Consistent

The example intentionally labels Sortino as periodic and Calmar as CAGR-based annualised. Their numeric sizes must not be compared as though 6.0452 were “better than” 0.7071. They are different statistics.

Do not substitute CAGR for an arithmetic mean excess-return numerator without stating that altered definition. Do not assume that a square-root time-scaling rule reproduces an exactly aggregated annual downside distribution: return dependence, target conversion and the selected convention matter. Preserve the original frequency and any stated annualisation assumptions.

Changing the window or observing only daily closes can also change measured drawdown and shortfalls. Use the same complete available period for comparisons and distinguish account balance from floating equity. Deposits and withdrawals must not become strategy returns simply because they change balance.

What These Ratios Can and Cannot Establish

A sampled maximum drawdown is an observed path decline, not the worst loss that could ever occur. Downside deviation is not a maximum-loss guarantee either. Neither statistic establishes an acceptable position size, a prop-firm rule or eligibility for funding.

There is no universal “good above 1”, “excellent above 2” or risk-free ranking. Negative numerators and different denominators can make simple higher-is-better comparisons misleading; the Calmar calculation guide gives an explicit example. Inspect the numerator, path, record count, missing data and cost basis rather than treating a score as a complete risk judgement.

Reconcile MyVeridex’s Actual Product Fields

In the reviewed recomputation paths, MyVeridex groups reported closed-trade money P/L by date with closed records. Its Sortino calculation uses a zero threshold on that money series and a 252-based scaling convention. It does not expose the arbitrary percentage-return target used in the example above. Dates without closed records are absent rather than a complete regularly sampled return series.

The current Calmar field annualises reported profit per date with closed trades and divides by reconstructed monetary drawdown. It is not automatically the CAGR/peak-relative-percentage Calmar calculation. Separately stored costs are not selected in that recomputation, and starting balance is inferred from current balance minus the included reported profit total. Reconcile source semantics and external cash flows before treating the fields as matching a complete after-cost return report.

Direct product routes currently support MT4 and MT5, with supported MetaTrader statement structures. Other named direct platform routes are unavailable. A broker-directory entry does not prove tested compatibility, reconstructed closed-trade balance does not recover all floating-equity history, and a product badge is not an independent audit or provider acceptance. Source scope was checked on 16 October 2026; no account-calculation repair or customer/broker test is claimed.

A Reproducible Comparison Checklist

1. Save the original record, currency, observation dates and treatment of external cash flows and charges.
2. Specify whether the series contains periodic percentage returns, cash P/L, balances or equity observations.
3. For Sortino, record target frequency, denominator convention and missing/zero-shortfall cases.
4. For Calmar, record compounded return, year basis, peak-relative drawdown and observation frequency.
5. Label product proxies separately; check provider objectives from their actual terms and keep later observations separate from historical selection.