Crypto backtesting:
test more than the chart.

Start with rules you can reproduce

Specify the market, timeframe, entry and exit conditions, sizing and risk limits before running the test. State whether a condition uses a completed candle or information available during the candle. A rule that sees a candle's closing price cannot also assume it entered earlier in that same candle without a defensible execution model.

In Strategy Arena, Blueprint Studio expresses strategies as connected nodes, with Pine Script v6 import also available. Keep the blueprint or imported rule version with the run. Record every parameter change rather than overwriting the settings behind a promising result.

Check the data before measuring the strategy

Confirm the exchange, market type and time interval. Spot data and futures data are not interchangeable. Check timestamps, missing candles, duplicated records and the history available for newly listed or delisted assets. A universe built only from assets that survived to today can omit failures that a trader could actually have encountered.

• Keep a record of the dataset and date range used by each test.
• Define what the engine should do when a candle or order book update is missing.
• Avoid filling gaps with future information.
• Distinguish exchange timestamps from the time a system received the data.
• Document any exclusions rather than silently discarding difficult periods.

Include fees, spread, slippage and funding

Gross return ignores the cost of moving into and out of positions. Model maker or taker fees as appropriate to the order type. Market orders cross a spread and may consume multiple levels of depth. Futures positions may also incur funding. If the data does not support a realistic fill model, make that limitation explicit.

For example, a hypothetical $1,000 buy and $1,000 sell with a fee of 0.1% on each leg would pay $2 in fees before spread, slippage or funding. That is an arithmetic illustration, not an exchange fee quote or an Apex return estimate. Small predicted edges can disappear when you include realistic costs.

Use the Trading Terminal and order book tools to inspect the liquidity that your assumptions are meant to represent.

Separate development from evaluation

Use one period to develop the rules and another, untouched period to evaluate them. When the evaluation period informs another parameter change, it becomes part of development and is no longer an independent test. Walk-forward research repeats a defined train-and-test sequence over time; it does not license using future information.

Compare the result across market regimes, nearby parameter values and more demanding cost assumptions. Keep the number of candidates you tried. Selecting the best of thousands of runs can produce an impressive curve by chance. Apex OS Quant Labs and Backtest Foundry include out-of-sample testing and overfitting statistics; these support diagnosis and do not certify profitability.

Stress the result and test it forward

Inspect drawdown, turnover, exposure, trade count and the concentration of returns. If a few trades account for almost all the result, understand those trades before extrapolating. Monte Carlo analysis explores variation under a chosen model; it does not remove errors in the source data or the strategy assumptions.

Then run the frozen rules on new data in Shadow and Paper. Compare the signals, costs and simulated fills with what the backtest assumed. The paper trading guide explains why a current order book still does not guarantee a real fill.

Find the right research tools in Apex OS

Strategy Arena and Blueprint Studio are included from Trader for Shadow and Paper. Research modules differ by plan, and Premium adds live strategies and the full research suite. Review the plan comparison for the exact feature allocation.

The public live demo shows market and terminal apps, rather than a full backtesting account. Request access if you want a walkthrough of the strategy and research workflow.