Skip to content
Education · Plain English

How to read these numbers.

Every strategy page on Asteria shows the same set of metrics: annualized return, drawdown, Sharpe, profit factor, MAR, win rate. They mean the same thing whether the strategy trades crypto perpetuals, spot, equities or futures — only the typical ranges differ. Here is what each one actually means, what's a good value, and the trap each one hides. No formulas you don't need.

First thing to know

In-sample vs Out-of-sample

The big two numbers on every strategy card live next to each other and mean different things.

In-sample (5y research, 2021–2026) — the strategy was developed, tuned and tested against this period. It is not a prediction. It tells you "this is how the idea would have looked on the data we used to design it." It will always look good, because if it didn't we'd have thrown the strategy away.

Out-of-sample (recent 12m OOS) — the strategy runs against the last twelve months after it was tuned. This is the reality-check. We recompute it weekly. If in-sample is 99% annualized and OOS is 30% annualized, the difference is honest decay, not a bug.

Rule of thumb: trust the OOS number more than the in-sample number. Treat in-sample as "the ceiling," OOS as "the working estimate." Live results usually come in 2–3× lower than the in-sample headline.
The headline number

Annualized return

Annualized return is the smoothed yearly return that would take your starting balance to your ending balance over the same period. It makes different strategy samples easier to compare fairly.

Annualized return = (Ending ÷ Starting) ^ (1 ÷ years) − 1

Example: $1,000 → $1,610 in 2 years is 26.9% annualized, not 30.5%, because compounding matters.

≤ 20%Modest. Below most actively-managed strategies.
30–80%Realistic and defensible.
80–150%Strong — but always check OOS and drawdown.
> 200%Almost always curve-fit or unsustainable.
Watch out: a high annualized return on a short window can be one good trade away from looking ordinary. Always check the trade count below.
The number that matters most

Max drawdown

Drawdown is the worst peak-to-trough loss the strategy went through. If your account hit $10,000 and later dropped to $7,800 before recovering, that's a 22% drawdown.

DD = (lowest after peak − peak) ÷ peak

This is the number that will decide whether you stay invested. Looking at return without drawdown is like buying a car based on top speed without checking the brakes.

< 10%Very smooth.
10–25%Typical for futures strategies.
25–40%Painful but survivable.
> 40%Most people quit before recovering. Avoid.
Self-test: imagine your starting capital cut by the max drawdown number. If that loss would force you to turn the strategy off, the position is too big. Reduce it.
Return per unit of pain

MAR ratio

MAR = annualized return divided by max drawdown. It tells you how much return you got for each percent of pain.

MAR = annualized return ÷ |max drawdown|

A MAR of 1.0 means a strategy returned 30% per year and had a 30% drawdown — fair trade. A MAR of 3.0 means it returned 30% per year with only a 10% drawdown — exceptional.

< 0.5Poor risk-adjusted return.
0.5–1.0Acceptable.
1.0–2.0Good.
> 2.0Excellent — verify with OOS.
Volatility-adjusted return

Sharpe ratio

Sharpe asks: how much excess return did you earn for each unit of volatility? It penalises strategies that get the same return through a noisier ride.

Sharpe = (avg return − risk-free) ÷ stdev of returns × √(periods per year)

Sharpe is the metric professional allocators use first. It is also the easiest to game by picking the right time window. Compare Sharpe only between strategies on the same data and same frequency.

< 1.0Mediocre.
1.0–1.5Decent.
1.5–2.0Strong.
> 2.0Outstanding — and rare. Verify.
Trap: Sharpe treats upside volatility (good!) and downside volatility (bad) equally. A strategy with rare big wins can look "less Sharpe" than a steady mediocre one even when it's actually better.
Quality of the wins vs losses

Profit factor

Profit factor = total dollars won ÷ total dollars lost. A profit factor of 2.0 means for every $1 the strategy lost, it made $2.

PF = sum(winning trades) ÷ |sum(losing trades)|
< 1.2Edges too thin to survive fees.
1.2–1.8Working.
1.8–3.0Healthy.
> 5.0Suspiciously high. Check sample size.
How often the strategy is right

Win rate

Win rate is the percentage of trades that closed profitable. By itself, it tells you nothing about whether the strategy makes money.

Trend-following strategies typically have 30–50% win rate with big winners. Mean-reversion strategies typically have 70–90% win rate with small winners and occasional big losers. This holds across asset classes — crypto, futures, equities — only the trade sizes scale.

Always pair win rate with profit factor. A 40% win rate with PF 2.5 is excellent. A 90% win rate with PF 1.1 is one bad day from blowing up.
Sample size

Trade count

The number of trades behind a backtest. The more trades, the more confident we can be that the result isn't an accident.

< 30Treat as a coin flip.
30–100Suggestive, not conclusive.
100–300Statistically meaningful.
> 300Robust enough to draw conclusions.
Reality check

Why your live result will differ from the backtest

Even an honest, properly tested strategy will not match its backtest in production. Reasons:

ReasonTypical impact
Slippage — the price you actually fill at vs the price the backtest assumes.0.1–0.5% per trade
Fees — taker/maker fees + funding on perpetuals.0.05–0.15% per round trip
Latency — milliseconds between signal and order.Negligible for swing strategies, painful for HFT
Liquidity — the backtest assumes you got the price; in real life sometimes the order rests or partially fills.Strategy-dependent
Regime change — the market behaves differently after the strategy was developed.The big one
Plan for: live results in the 30–60% range of the backtest headline over the first year, climbing toward the OOS number as the strategy proves itself on your account.
When to walk away

Red flags to spot

If you see any of the below, pause and think before depositing capital:

Annualized return > 300% on a long backtestUsually overfitting. Probably won't survive.
Max drawdown not shownWalk away. This is the most important number.
Win rate 95%+Almost always means "no stop loss; one bad trade wipes the gains."
Fewer than 30 tradesNot enough data to conclude anything.
Backtest period doesn't include 2022 bear marketSurvivorship — strategy hasn't been tested in pain.
No out-of-sample number publishedYou're looking at the curve they fit, not a forecast.

What Asteria does with all this

Every strategy on the Strategies page shows the in-sample headline next to the recent 12m OOS, with trade count and drawdown alongside. The Risk Center on your Dashboard alerts you if your live drawdown exceeds 10% from your peak balance, and recommends pausing new entries above 20%. The Cockpit page (beta) lets you watch the strategy's decisions tick by tick on any ticker before you commit real capital. Nothing here is investment advice — but everything here is shown the same way every quant fund shows them internally.

Back to strategies →