The backtest reports a couple of dozen numbers per model. They are not a scoreboard to be summed. Each answers one question, and the interesting information is usually where two of them disagree.
The four questions
- How fast did it grow? CAGR
- How bumpy was the ride? Volatility
- What's the worst it got? Max drawdown
- Was the growth worth the risk? Sharpe, Sortino, Calmar
The three ratios in the last line are all "return per unit of risk." They differ only in what they count as risk. Sharpe divides by all the wobble, Sortino by the downside wobble only, and Calmar by the single worst fall. Ulcer Index is the odd one out: not a ratio at all, but a measure of how long you spend underwater.
Why the disagreements matter
A model that leads on CAGR and trails on Calmar bought its growth with a deep hole you'd have had to sit through. One that leads on Sharpe but not Sortino was penalized for swinging upward. When the Strengths panel gives a model four badges and the next model none, that usually means one window, one lucky start date, not a settled answer.
Two habits keep this honest. Compare only over the same window (the page indexes every model from a common start date for exactly this reason), and treat a short backtest as a hint rather than evidence: every one of these numbers is arithmetic on a single historical path, not a forecast of the next one.
Where they come from
All of it is computed from the model's own simulated history: daily returns of the allocation as you recorded it, rebalanced on its cadence. Sharpe and Sortino use a fixed 4% risk-free rate. Hover any metric in the table or on a Strengths card for its one-line definition, or open the term itself for the full entry.