A backtest metric summarizes a particular return history. Read it with the run's dates, inputs, benchmark and assumptions. Numerical examples below illustrate arithmetic; they are not measured investment results.
This public card changes as the model updates. Open the linked record and inspect the current result before quoting it.
Compound annual growth rate
CAGR is the constant annual rate that connects the starting and ending value through compounding. In an illustrative example, growth from 100 to 250 over 12 years corresponds to about 7.9 percent annually.
ENSEMBLE compounds the selected daily return series and annualizes using the number of return observations divided by the reported bars per year. CAGR does not describe the path or the size of losses along the way. Compare matched windows and the same treatment of costs and cash flows.
Annualized volatility
Volatility describes how widely returns vary. The current performance calculation uses the sample standard deviation of daily returns multiplied by the square root of the annualization factor. The result exposes bars_per_year: 252 for an exchange calendar or 365 for an everyday calendar inferred from the dates.
For an illustrative exchange-calendar series, a daily standard deviation of 1 percent scales to about 15.9 percent annually. This scaling is a convention, not a guarantee about future dispersion. It treats positive and negative moves symmetrically and can obscure tail risk.
Sharpe convention
Sharpe ratios are often described using mean excess returns relative to a risk-free rate. ENSEMBLE's current headline metric uses CAGR divided by annualized volatility, with risk_free=0. It is not a Treasury-bill-adjusted arithmetic-mean Sharpe ratio. The web calculation and server performance block use this convention.
Do not compare it directly with another service's Sharpe value without reconciling the numerator, risk-free assumption, calendar and window. A high number can reflect a favorable sample or omitted costs; there is no universal acceptable range.
Maximum drawdown
Maximum drawdown is the largest peak-to-trough decline in the cumulative return series over the selected window. An illustrative fall from 200 to 120 is a 40 percent drawdown. Recovery is not required for a decline to count; an ongoing drawdown matters too.
Inspect its duration and the full path. A single worst episode does not describe every loss, and a window without a severe decline cannot establish that the model is safe.
Trading activity and exposure
Trade counts use the available fill ledger or runtime totals. Where a freeform result has no fill ledger, the performance block can fall back to an estimate based on book changes. Rebalance counts and trade counts are different: one rebalance can involve several fills.
Turnover metadata describes the scale of trading, with its convention and available evidence depending on the result. It is not itself a measurement of fees or taxes. Similarly, inspect the allocation history to understand exposure over time rather than assuming that a strategy label proves its time in market.
Compare and keep the research record
Check the run ID, dates, price basis, warnings and any synthetic history before comparing. Review the model code you own, its inputs, and the selected gross or fee-adjusted view. Save both run IDs and compare one change over matching windows. Stored research supports inspection but does not guarantee immutable snapshots of every data dependency.
Use the portfolio walkthrough to carry out that sequence. This illustrative prompt starts a model whose actual implementation you should review:
ENSEMBLE is research software. Backtests are hypothetical and past performance does not guarantee future results.
Frequently asked questions
- What is a good Sharpe ratio?
- There is no universal pass/fail value. Compare identical windows and conventions, inspect drawdowns and costs, and test sensitivity. ENSEMBLE's current headline calculation uses CAGR divided by volatility with a zero risk-free rate.
- Is maximum drawdown more important than volatility?
- They answer different questions. Volatility describes return dispersion; maximum drawdown describes the worst observed peak-to-trough decline. Neither captures every risk.
- What is the difference between CAGR and average annual return?
- CAGR describes compounded growth. In an illustrative two-year example, a 50 percent gain followed by a 50 percent loss leaves 75 percent of the starting value: the arithmetic average is zero, but CAGR is about negative 13.4 percent.
- Are transaction costs included automatically?
- No. Inspect the model and selected gross or fee-adjusted view. Do not assume commissions, spreads, taxes, financing or market impact were modeled unless explicitly included.
Related
- BacktestingHow to backtest a portfolio without writing code
Research two assets, build an allocation model, inspect its saved results, and compare one change. A practical first workflow for DIY investors using ENSEMBLE.
- BacktestingCommon backtesting mistakes and how to avoid them
The ways a backtest lies: lookahead bias, survivorship bias, overfitting, ignored costs, short or regime-limited samples, and misread metrics. What each one looks like, how to detect it, and what a backtesting tool can and cannot protect you from.
- StrategiesThe 60/40 portfolio: rules, variants, and a backtest
Define a stock-and-bond allocation, inspect a public 60/40 model, and test rebalancing or asset changes over matched dates. Learn what the model shows and what it leaves out.
- StrategiesDual momentum: the rules, the evidence, and a backtest
Define relative and absolute momentum separately, inspect a public model, and compare a parameter change. The example explains the research workflow without promising a historical or future result.