Paper Trading vs. Backtesting: From Historical Data to Live Execution

KTX
KTX
  • Updated
Paper Trading vs Backtesting — KTX Learn

A strategy can make money on historical data and start losing in paper trading. The explanation may lie in signal calculations, fill assumptions and trading costs. Backtesting applies rules to past data. Paper trading typically uses current market data and virtual money to observe how the strategy executes over time. Used together, they help explain where returns come from and where results diverge.

What is backtesting? Testing trading rules on historical data

A backtest needs defined data, entry and exit conditions, position sizing and a fill model. For example, a BTC strategy might signal when an hourly candle closes above the highest price of the previous twenty completed candles, then enter at the next executable opportunity. The system calculates trades and the equity curve in historical order.

Backtests can quickly compare different market regimes and test sensitivity to fees, slippage and parameters. Results depend on the input data and model. Open, high, low and close data alone cannot reconstruct the full sequence of trades inside a candle. If a candle touches both a stop and a profit target, which triggers first affects the outcome. Finer data reduces ambiguity, but the execution assumptions still need to be stated.

What is paper trading? Observing real-time execution with virtual capital

A real-time paper account generates signals as new data arrives, submits virtual orders and records results. It helps check data timeliness, order states, cancellations and recovery after a disconnection. Discretionary traders can also practise following rules and keeping records.

QuantConnect paper-trading documentation describes combining real-time data with fictional capital and notes that its default model does not simulate slippage in either backtesting or paper trading. Before using a simulator, understand how it handles spreads, partial fills and limit-order queues. A real-time feed alone cannot reproduce every condition of real execution.

The main differences between paper trading and backtesting

Backtesting uses historical data; paper trading uses live data and virtual funds
Comparison Historical backtest Real-time paper trading
Data Historical market data Incoming current market data
Speed Can cover long periods quickly Advances with real time
Purpose Research rules across market conditions Check forward performance and execution
Fills Historical data and an execution model Current data and simulated matching rules
Common biases Look-ahead, overfitting, missing data Idealised fills, short observation, frequent rule changes
Capital Hypothetical funds in a calculation Virtual funds in a simulated account

Why can the same strategy produce different results?

Signal and execution times differ

If a breakout requires the final closing price, wait for the candle to finish and then use a price actually available. Filling a signal known only after the close at that candle's earlier low overstates returns. The best price visible on a historical chart may have been unavailable when the signal became actionable.

Fees, slippage and funding differ

Suppose a round trip earns 0.40% before costs. Entry and exit fees of 0.05% each and adverse slippage of 0.03% on each side leave approximately 0.24%, ignoring compounding. An additional 0.02% funding payment reduces that to about 0.22%. These figures illustrate the calculation. Actual costs depend on the product, account tier, holding period and fills. Avoid counting the same spread cost again as slippage.

Parameter selection and market samples differ

Repeatedly tuning moving averages, stops and filters on the same history can select a combination that happens to fit past noise. Reserve data excluded from tuning and then observe subsequent markets. If paper results lead to parameter changes, record the new version and effective time, and restart its forward evaluation.

How to connect backtesting with paper trading

  1. Document the rules, including data source, candle time zone, calculation time, order type, position sizing and exits.
  2. Split research and out-of-sample periods chronologically. Define evaluation methods in advance to reduce changes made after seeing results.
  3. Include reasonable execution costs and test sensitivity to slippage, fees and derivatives funding.
  4. Freeze the version for paper trading. Save signal times, order prices, fill prices, cancellations and exceptions.
  5. Explain differences trade by trade. Check data gaps, latency, minimum order size, price precision, margin and limit-order fill rules.

When checking price ranges and volatility, use KTX BTC/USDT market to observe market conditions. Keep the pair, product type and time frame consistent when comparing tests.

What should you measure?

Alongside net returns, examine maximum drawdown, drawdown duration, trade count, average win and loss, turnover and risk exposure. Check whether a few trades generate most profits and compare trending, ranging and sharp-fall periods. A high win rate can be offset by a few large losses. A short profitable paper run may cover only one market regime.

Match the observation period to trading frequency. A strategy trading once a week provides little evidence after a few days. Define the number of trades and market scenarios needed before assessing results, reducing premature conclusions based on short-term profit or loss.

Frequently asked questions

Should beginners start with backtesting or paper trading?

Paper trading is useful for practising order entry and learning order types. For a defined strategy, backtest the rules first, then observe a fixed version in paper trading.

Does historical replay count as paper trading?

Some platforms call historical replay simulated trading. Check whether the data is historical or real time. Replay is closer to manually advancing a historical test.

If both tests are profitable, will live trading be profitable?

Real trading also involves actual order queues, market impact, system reliability and behaviour under financial pressure. Identify which conditions the simulation covers.

Can backtesting eliminate look-ahead bias completely?

Check indicator publication times, candle completion and signal execution individually. A day's final close, for example, becomes usable only after the relevant time.

Should paper trading include fees?

Yes. Fees and execution costs affect the net returns of frequently trading strategies. Record any costs omitted by the platform separately and adjust the evaluation.

Was this article helpful?

0 out of 0 found this helpful

Have more questions? Submit a request