What Is Backtesting and Why It Matters for Trading Strategies
Before risking money on a strategy, there is an obvious question: would these rules have worked? Backtesting is the standard way to answer it — the strategy's rules are replayed against historical market data, trade by trade, to produce the performance record the strategy would have had.
Backtesting is genuinely valuable and systematically overrated at the same time. Both halves matter.
What a backtest does
A backtest engine takes three inputs — the strategy's rules, historical price data, and assumptions about costs — and simulates the strategy through time: every signal, every entry, every exit. The output looks like a real track record: total return, Sharpe ratio, win rate, maximum drawdown, trade count.
Done well, this answers important questions cheaply:
- Does the logic have any historical edge at all? Most candidate strategies fail here, at zero cost.
- What is the strategy's character? Trade frequency, texture of wins and losses, and drawdown depth show up in simulation (see How to Read a Trading Strategy's Stats for the metrics).
- How did it behave in different conditions? A backtest spanning bull, bear, and sideways periods shows regime sensitivity — where the strategy earns and where it bleeds.
Where backtests mislead
The dangers are well-documented and worth knowing by name:
- Overfitting. Tune a strategy's parameters long enough and it will fit the past beautifully — because it has memorized history's accidents, not learned a repeatable pattern. Overfit strategies produce spectacular backtests and mediocre live results. This is the single most common failure mode.
- Cost underestimation. Simulated fills are clean; real fills pay fees, slippage, and funding. High-frequency strategies can be profitable in simulation and net-negative live on costs alone.
- Look-ahead and survivorship bias. Subtle engineering mistakes let the simulation "know" things it couldn't have known in real time, inflating results.
- Regime dependence. A backtest covering only a favorable period proves little about the next period — strategy families like trend following and mean reversion have close to opposite regime profiles, so the test window decides which one looks good.
None of these are arguments against backtesting — they are arguments for treating a backtest as a filter, not a promise.
The higher standard: live verification
The gap between simulated and real performance is exactly why live results carry more weight. A live-verified track record is measured from actual trading in real markets: real fees, real slippage, real stress, no hindsight. Uptime — how long a strategy has traded live — is effectively a measure of how much of this hard evidence exists.
The mature evaluation sequence for any strategy:
- Backtest — establishes the logic had historical merit and reveals its character.
- Live verification — the strategy trades in real conditions, and its statistics are measured from actual account activity.
- Ongoing display — live stats stay visible so followers can see current behavior, not just a snapshot.
Strategies presented for following should show where they sit in this sequence. On the ONYX strategy market, strategies are backtested during development and then display live-verified statistics — Sharpe, win rate, Max DD, uptime — measured from real trading, updated as they run.
A reader's checklist
When you see any performance claim, three questions sort most of it out:
- Is this simulated or live? (If unlabeled, assume simulated.)
- Are costs included?
- How long is the live period, and does it include more than one market condition?
A strategy that answers all three cleanly is offering evidence. One that answers none is offering a story.
This article is for general information only and is not financial advice. Futures trading carries risk of principal loss and liquidation. Past results do not guarantee future performance.