Glossary

Backtesting terms, in plain English.

38 words you’ll meet when you test a trading strategy, explained without the jargon.

B

Backtest

Running a strategy’s rules over historical data to see how it would have performed. It shows what would have happened under the test’s assumptions, not what will happen next.

C

Commission and fees

What the broker and the exchange charge for each trade, usually per contract for futures. Small on one trade, but large for a strategy that trades often, so a backtest without them flatters it.

D

Daily loss limit

A prop firm rule that caps how much an account can lose in a single day. Depending on the firm, breaking it either ends the attempt or stops trading for the rest of the day, even if the strategy is profitable overall.

Data snoopingalso: Multiple testing

The bias that comes from testing many ideas or versions on the same data and keeping the best one. The more you test, the more likely the winner is simply the luckiest.

Deflated Sharpe ratio

A statistic, developed by researchers David Bailey and Marcos López de Prado, that estimates how likely it is that a strategy’s Sharpe ratio reflects real skill, after adjusting for how many strategies were tried, how much data there is and unusual return patterns. It keeps the best of many tries from being mistaken for skill.

E

Edge

A repeatable advantage: a reason a strategy’s average trade should make money after costs, beyond what luck alone would produce.

Expectancy

The average amount a strategy makes or loses per trade. It combines the win rate with the average win and the average loss, which is why a high win rate alone says little.

F

Fill

The execution of an order. A realistic backtest only assumes fills that could really have happened; for example, a limit order isn’t always filled just because price touched it.

I

In-sample data

The part of history used to build and tune a strategy. Results on it are almost always flattering, because the strategy was shaped to fit it.

L

Look-ahead bias: When a backtest uses information that wouldn’t have been available at the time of the trade, such as using a bar’s closing price to decide an entry at its open. It inflates results in a way no one can repeat live.

M

Market regime

A stretch of time with a particular character, such as trending or choppy, calm or volatile. Many strategies work in some regimes and not others.

O

Out-of-sample data

History the strategy didn’t see while it was being built or tuned, used to check whether it still holds up.

Overfitting

Tuning a strategy so closely to past data that it captures random noise rather than a pattern that repeats. It looks great on the data it was built on and disappoints on new data.

P

Paper trading

Simulated trading with no real money, usually on live prices. Useful for practicing and checking execution, but simulated fills can be kinder than real ones.

R

Risk of ruin

The chance that a run of losses takes an account down to the point where trading has to stop.

S

Sample size

How many trades a result rests on. More trades make luck less likely, but there’s no magic number: trades from one kind of market add less evidence than trades spread across many.

T

Trailing drawdown

A maximum-loss limit that rises as the account makes new highs, so the allowed loss is measured from the best point rather than the starting balance. Common in prop firm evaluations; at many firms it stops rising once it reaches the starting balance, and some count open profits.

W

Walk-forward analysisalso: Walk-forward testing

Tuning a strategy on one window of history, testing it on the next unseen window, then sliding forward and repeating. It shows whether an edge holds across time rather than in one lucky split.

Win rate

The share of trades that make money. On its own it says little: a strategy can win most of its trades and still lose money if its losses are much bigger than its wins.