Risk & Efficiency Metrics (Risk-Adjusted Metrics)

Evaluating a portfolio based solely on absolute return is insufficient: high returns may stem from taking excessive risk. Firewire Analytics calculates a suite of institutional risk metrics to measure capital management quality.

1. Sharpe Ratio

The Sharpe ratio measures the excess return of a portfolio over the risk-free rate (the Bank of Russia key rate) per unit of total risk (standard deviation of returns):

where:

• $R_p$ — average portfolio return per interval between value snapshots (adjusted for deposits and withdrawals);
• $R_f$ — risk-free return for the same interval: the Bank of Russia key rate scaled to the interval length;
• $\sigma_p$ — standard deviation (volatility) of portfolio returns across intervals;
• $N$ — annualization factor: 252 for daily snapshots, otherwise $365 / \bar{d}$, where $\bar{d}$ is the average interval length in days.

Interpretation:

• Sharpe < 0.5 — Weak risk-adjusted performance;
• 0.5 – 1.0 — Good balanced portfolio performance;
• Sharpe > 1.0 — Excellent risk management;
• Sharpe > 2.0 — Outstanding institutional grade.

2. Sortino Ratio

Unlike the Sharpe ratio, the Sortino ratio penalizes only downside volatility, avoiding penalizing positive upside surges:

where $N$ is the same annualization factor as for Sharpe, and $\sigma_{\text{down}}$ is the downside deviation: only intervals with return $r_i$ below the risk-free $R_f$ count, while the sum is divided by the total number of intervals $n$:

3. Alpha ($\alpha$) and Beta ($\beta$)

Compare the portfolio against the benchmark — the Moscow Exchange total return index MCFTR (with reinvested dividends; the price index IMOEX is not suitable here):

• Beta ($\beta$) — market sensitivity. $\beta = 1$ moves in line with the benchmark, $\beta > 1$ is more volatile, $\beta < 1$ is defensive. Firewire computes beta by regression: the same flow-adjusted sub-period portfolio returns $r_p$ used for Sharpe against the MCFTR index returns $r_m$ over the same dates (MOEX ISS data): $\beta = \dfrac{\operatorname{cov}(r_p, r_m)}{\operatorname{var}(r_m)}$. The method is exposed in the metric's methodology field (2.0-regression-mcftr). If fewer than 20 overlapping points exist (about a month of trading) or the index history is unavailable, the reference estimate is kept: the value-weighted beta of positions $\beta_p = \sum_i w_i \beta_i$ (bonds 0.15, cash 0, stocks 1.05 or an individual beta for the largest issuers; methodology 1.0-weighted).
• Alpha ($\alpha$) — excess return beyond what market risk explains (Jensen's alpha):

where $r_p$ and $r_m$ are the sub-period returns of the portfolio and MCFTR, $r_f$ is the Bank of Russia key rate scaled to the sub-period length, the bar is the series mean, and $N$ is the number of sub-periods per year (252 for daily snapshots, otherwise $365/\text{interval}$ — as in Sharpe). Alpha is annualized, in percent. The Treynor ratio uses the same units: $\dfrac{(\bar r_p - r_f) \cdot N}{\beta}$. In the fallback mode (reference beta) alpha uses the portfolio's cumulative return annualized with compounding; it is hidden for periods shorter than 30 days.

4. Value at Risk (VaR 95%)

VaR (95%) is the loss per interval between value snapshots (per day for daily snapshots) that the portfolio's history exceeded in no more than 5% of cases. In 95% of intervals losses did not exceed VaR; it is not the maximum possible loss.

Firewire uses the historical method: interval returns are sorted ascending and VaR is the absolute return at the 5% quantile:

where $r_{(k)}$ is the return with index $k$ (zero-based) in the sorted series of $n$ values. VaR 99% is computed the same way.

5. Max Drawdown

Measures the maximum percentage drop from peak portfolio valuation to local trough over the selected horizon.