Skip to content

Metrics Glossary ​

Formal definitions for every statistic reported after a MarketHeist backtest. All metrics are computed on the strategy's daily (or bar-frequency) return series unless otherwise noted.


Sharpe Ratio ​

Definition: Annualized excess return divided by annualized volatility of returns.

Sharpe = (μ_strategy - μ_rf) / σ_strategy × √N

Where:

  • μ_strategy = mean daily return of the strategy
  • μ_rf = daily risk-free rate (assumed 0% in MarketHeist)
  • σ_strategy = standard deviation of daily strategy returns
  • N = trading days per year (252 for daily; 52 for weekly; 12 for monthly)

Interpretation: Risk-adjusted return per unit of total (upside + downside) volatility. Higher is better. See Metrics for interpretation ranges.


Sortino Ratio ​

Definition: Like Sharpe, but uses only downside volatility (standard deviation of negative returns).

Sortino = (μ_strategy - μ_rf) / σ_downside × √N

Where σ_downside = standard deviation of daily returns that are strictly negative (upside volatility ignored).

Interpretation: More appropriate than Sharpe for asymmetric strategies where large upside moves are desirable. A Sortino ratio above 1.5 is generally considered strong.


CAGR (Compound Annual Growth Rate) ​

Definition: Annualized geometric return over the full backtest period.

CAGR = (Final portfolio value / Initial portfolio value)^(1/T) - 1

Where T = number of years in the backtest period (fractional years allowed).

Interpretation: The constant annual return rate that would produce the same terminal value. A strategy with CAGR of 12% would double in value roughly every 6 years.


Max Drawdown (MDD) ​

Definition: The maximum peak-to-trough decline in the equity curve, measured from any historical peak to any subsequent trough before a new peak is set.

MDD = max over all t of [ (Peak(0→t) - Trough(t→T)) / Peak(0→t) ]

Interpretation: The largest percentage loss an investor would have sustained if they had entered at the worst possible time. Expressed as a negative percentage (e.g., -35%). A lower absolute value is better.


Calmar Ratio ​

Definition: CAGR divided by the absolute value of Max Drawdown.

Calmar = CAGR / |MDD|

Interpretation: Return per unit of worst-case drawdown. A Calmar above 0.5 is generally acceptable; above 1.0 is strong. Unlike Sharpe, Calmar focuses on the worst-case scenario rather than average volatility.


Omega Ratio ​

Definition: The total of the returns above a threshold return level L (typically 0%) divided by the total of the shortfalls below it.

Omega(L) = ∫[L,∞] (1 - F(r)) dr  /  ∫[-∞,L] F(r) dr

Where F(r) is the cumulative distribution function of returns.

In practice, computed as:

Omega = Sum of (returns - L) when positive / Sum of (L - returns) when negative

Interpretation: Values above 1.0 mean the gains above the threshold add up to more than the losses below it (amounts, not counts of winning and losing periods). Unlike Sharpe, Omega captures all moments of the return distribution (skew, kurtosis, fat tails).


Profit Factor ​

Definition: Sum of all positive daily return values divided by the absolute sum of all negative daily return values.

Profit Factor = Σ max(r_t, 0) / Σ max(-r_t, 0)

Interpretation: How many dollars the strategy wins per dollar it loses, across all bars. A profit factor of 1.5 means the strategy earns $1.50 for every $1.00 lost. Values above 1.0 are net-profitable; below 1.0 are net-losing.


Recovery Factor ​

Definition: Total net profit over the backtest period divided by the absolute value of Maximum Drawdown.

Recovery Factor = Total Net Return / |MDD|

Interpretation: How many times over the strategy "paid back" its worst drawdown through accumulated profits. A recovery factor below 1.0 means the strategy has not yet recovered from its worst loss within the tested period.


% Invested ​

Definition: Percentage of total bars (trading days, weeks, or months) during which the strategy held a long position.

% Invested = (Bars with long signal / Total bars) × 100

Interpretation: A high % invested (close to 100%) means the strategy behaves similarly to buy-and-hold. A low % invested means the strategy spends significant time in cash, reducing compounding but also limiting drawdowns.


Annualized Volatility ​

Definition: Standard deviation of daily returns, annualized.

Ann. Volatility = σ_daily × √N

Where N = 252 (daily), 52 (weekly), or 12 (monthly).

Interpretation: The annualized "wobble" of the strategy. The S&P 500 has historically shown ~15–20% annualized volatility. Lower volatility with similar returns implies higher Sharpe.


Total Return ​

Definition: Cumulative percentage gain or loss over the full backtest period.

Total Return = (Final value / Initial value - 1) × 100

Unlike CAGR, Total Return does not normalize for time. A 200% total return over 20 years (CAGR ~5.6%) is very different from 200% over 5 years (CAGR ~24.6%).


Win Rate ​

Definition: Percentage of trading bars (days the strategy is long) that ended with a positive return.

Win Rate = (Profitable long bars / Total long bars) × 100

Note: Win rate in isolation is not a reliable quality indicator. A strategy can have a 40% win rate and still be highly profitable if wins are much larger than losses (high profit factor).


Assumptions & limitations ​

Metrics are only as meaningful as the assumptions behind them. MarketHeist's backtest engine uses these conventions — keep them in mind when comparing results to a live account or another tool:

  • Risk-free rate is 0%. Sharpe and Sortino treat the risk-free return as zero. In high-rate periods this flatters strategies that sit in cash for long stretches, since idle cash is credited no yield.
  • Leverage is frictionless. A leverage multiplier scales the strategy's return series directly, with no borrowing or financing cost and no margin-call mechanics. Reported numbers show the mechanical effect of leverage, not the cost of obtaining it.
  • target_vol and target_dd are descriptive, not walk-forward. These modes pick a single leverage multiplier from the strategy's realized volatility or drawdown measured over the whole backtest window. Because that uses the full history, the achieved volatility/drawdown target is in-sample — a live strategy sizing in real time would not have known it in advance.
  • Transaction costs are turnover-based, one-way, and idealized. The optional cost (in basis points) is charged in proportion to how much the position changes on each bar. It does not model bid/ask spread, slippage, market impact, or taxes.
  • Returns are dividend-adjusted. Price series use total-return (adjusted close), so dividends are reinvested; there is no separate cash-dividend line.

Further reading ​

MarketHeist Backtest Engine