Why Backtesting Bitcoin Is Different

Bitcoin trading isn’t just “stocks without earnings.” It’s a 24/7, globally fragmented market with unique microstructure and cost drivers. To backtest effectively, you need to account for factors that don’t exist—or matter less—in traditional markets.

• 24/7 trading and weekend liquidity: No closing bell means signals can trigger at any hour. Spreads and depth vary by time of day and global session, affecting fill quality.
• Exchange fragmentation: Price discovery occurs across many venues. Canadian BTC/CAD books can be thinner than USD pairs, impacting slippage and capacity.
• Perpetual futures funding: If your strategy uses perps, funding payments add a recurring cost (or credit) you must model accurately, not as a flat fee.
• Fees and maker/taker dynamics: Taker fees, rebates, and VIP tiers vary by venue. Assume the wrong schedule and your P&L can be dramatically overstated.
• Stablecoin and fiat rails: Capital flows via stablecoins, wires, or Interac e‑Transfer have timing and limits. Funding latency can delay entry and distort results if ignored.
• Data artifacts: Crypto data may contain outlier ticks, partial candles, or exchange-specific anomalies. Cleaning is mandatory.

Building a Clean, Credible Data Set

High-quality data is the foundation of any backtest. In crypto, it’s common to use multiple sources and reconcile them to reduce venue-specific bias and detect errors.

Choose the Right Market and Quote Currency

• BTC/USD vs BTC/CAD: Many Canadian traders default to BTC/CAD for convenience, but spreads can be wider and liquidity thinner. If your strategy doesn’t explicitly require CAD fills, consider testing on BTC/USD for cleaner depth and then incorporate FX when you map to realized CAD results.
• Spot vs Perpetuals vs Futures: Each requires different costs: trading fees only for spot; fees plus funding for perps; fees plus roll/expiry effects for dated futures.

Timeframes and Resampling

• Use UTC for consistency: It avoids daylight savings mismatches and makes global session analysis easier.
• Resample carefully: When converting trades to OHLCV, choose volume‑weighted prices and ensure your resampling logic handles missing intervals without inventing data.
• Validate completeness: Check that your provider didn’t skip days or minutes. Gap tests and candle count checks are essential.

Data Cleaning Checklist

• Remove impossible prices or volumes outside reasonable bands.
• De‑dupe overlapping records from multiple sources.
• Align timestamps, rounding to a consistent precision.
• Document every transformation so results are reproducible.

If you test a strategy on Canadian exchanges specifically, note that maintenance windows, liquidity troughs, and CAD pair quirks can create outliers you must filter or model explicitly (e.g., delayed candles during overnight maintenance on lower‑liquidity venues).

Get Costs Right: Fees, Spreads, Slippage, and Funding

Nothing inflates a backtest like underestimating trading costs. Model costs conservatively and you’ll avoid painful surprises.

Fees and Rebates

• Taker vs Maker: If your signals use market orders, assume the taker schedule. Only apply maker fees or rebates if your logic includes passive orders and realistic fill criteria.
• Tier assumptions: Don’t assume VIP fee tiers unless you’ve historically met their volume thresholds.
• CAD venues: Canadian platforms often publish transparent fee tables, but spreads may be larger than on deep USD books. Model both fee and spread impacts.

Spreads, Depth, and Slippage

At minimum, deduct half the quoted spread when crossing with small orders. For larger orders, simulate partial fills by walking the book or using a volume‑based slippage model. On Canadian BTC/CAD markets, consider a size‑aware function that increases slippage outside peak liquidity hours.

Funding for Perpetuals

Funding can flip a positive backtest negative. Use historical funding timestamps and rates, and apply them prorated to holding periods. If you carry positions through volatile windows, stress test scenarios where funding spikes.

Financing, Borrow, and Cash‑and‑Carry Nuances

If your strategy borrows stablecoins or BTC for margin, model borrow costs. If you run a basis trade (spot vs futures/perps), incorporate custody, withdrawal fees, on‑chain fees, and delays—especially relevant if moving funds between Canadian and global venues.

Execution Realism: From Paper to the Order Book

“Assuming you get the close price on every bar is the fastest path to overfitting.”

Execution assumptions determine whether a backtest reflects tradable reality. Design your simulator to mimic how your orders actually work.

• Market orders: Cross the spread and add size‑based slippage.
• Limit orders: Fill only if price trades through your level and available size is sufficient. Consider queue position and partial fills.
• Child orders (TWAP/VWAP/POV): Break large trades into time‑sliced or volume‑targeted chunks. Apply deterioration if your flow becomes a large share of interval volume.
• Latency and rate limits: Incorporate delays for signal generation, API calls, and exchange throttling, especially during high‑volatility windows.
• Venue selection: If you route across Canadian and global exchanges, simulate best‑execution logic and fees per venue.

For Canadian traders funding accounts via Interac e‑Transfer, model the cash‑in latency by disallowing trades until funds are confirmed. For withdrawals to self‑custody, simulate on‑chain fees and expected confirmation times; otherwise, your backtest may assume capital mobility that doesn’t exist in practice.

Signal Design Without Overfitting

The more parameters you add, the more likely you’re discovering noise. Keep indicators simple, document the hypothesis driving each signal, and prefer techniques that generalize across regimes.

• Start with a small, robust set of features (trend, momentum, volatility, or mean‑reversion). Avoid stacking many correlated indicators.
• Use regime filters sparingly—e.g., a volatility or trend filter—to determine when to deploy a tactic.
• Favor signals that are interpretable. If you can’t explain the edge in plain language, it may not persist.

Position Sizing, Risk, and In‑Bar Assumptions

Risk management starts inside the simulator. Define how stops, targets, and position sizes behave within each bar or tick.

• Volatility targeting: Size positions so the expected daily or weekly volatility contribution remains stable over time.
• Fixed fractional: Risk a small, consistent percentage of equity per trade; include compounding effects.
• Stops and gaps in a 24/7 market: In crypto, “gaps” are rarer but rapid moves happen. Decide whether stops fill at touch or require trade‑through, and account for slip on fast breaks.
• In‑bar logic: If testing on hourly candles, avoid assuming you can buy the low or sell the high. Use realistic order priority or intra‑bar simulations.

Walk‑Forward Testing and Robustness Checks

Backtests that shine only in a single historical window are fragile. Walk‑forward testing (WFT) forces your process to repeatedly retrain or refit on a rolling in‑sample, then evaluate on the next out‑of‑sample slice. It reveals how the strategy behaves as the market evolves.

A Simple Walk‑Forward Workflow

1. Split your data into N sequential folds (e.g., 12 months train, 3 months test, rolling).
2. Optimize parameters only on the train slice using conservative constraints.
3. Freeze parameters and run the test slice with live‑like costs and execution.
4. Roll the window forward and repeat. Aggregate out‑of‑sample results across all folds.

Additional Robustness Tools

• Parameter sweeps and heatmaps: Look for broad, flat regions of profitability rather than narrow peaks.
• Trade reshuffling: Bootstrap your sequence of trades (with replacement) to estimate variance around performance metrics.
• Noise injection: Add small random delays or price noise to test sensitivity to slippage and latency.
• Venue rotation: Re‑run on multiple exchanges or on BTC/USD and BTC/CAD to ensure the edge isn’t venue‑specific.

Evaluate What Matters: Beyond Raw Returns

Headline returns don’t tell the whole story. Focus on risk‑adjusted, capacity‑aware metrics and operational factors.

• Max drawdown and time under water: Know the worst historical pain and how long recoveries took.
• Sharpe and Sortino: Judge returns relative to volatility, with and without penalizing downside more heavily.
• Hit rate and payoff ratio: Strategies with low hit rates can still win if the average winner dwarfs the loser.
• Turnover and capacity: Higher turnover means higher cost and tax drag. On thinner CAD books, size limits can arrive sooner than on deep USD venues.
• Execution footprint: If your fills represent a large share of interval volume, expect worse outcomes live than in backtest.

From Backtest to Paper Trading to Live

Treat deployment as a staged process with explicit checks and balances.

1. Paper trading: Run the strategy in real time with the exact same order logic and costs. Compare expected vs realized fills and slippage.
2. Small‑size pilot: Go live with minimal size. Verify order routing, funding flows, and reporting work as designed.
3. Monitoring and alerts: Build dashboards for P&L, risk, latency, and data health. Include alerts for funding spikes, fee tier changes, or exchange incidents.
4. Post‑trade analytics: Attribute slippage and errors. Update your simulator with real‑world distributions so it stays honest.

Canadian traders should also confirm that their chosen platform supports the order types their strategy requires (e.g., post‑only, reduce‑only, or native stop orders) and understand any domestic account protections and operational constraints. If you interact with multiple venues, document custody flows between Canadian regulated platforms and global exchanges, including withdrawal whitelists and cooldowns.

Canadian Tax and Compliance Considerations in Your Backtest

While this article does not provide tax or legal advice, incorporating tax and reporting assumptions into your analysis will make your backtest more realistic—especially if you’re a Canadian resident.

• Income characterization: Crypto profits in Canada may be treated as business income or capital gains, depending on your circumstances and intent. Model both scenarios to see how outcomes differ after tax.
• Adjusted Cost Base (ACB): If treated as capital property, Canada’s ACB method affects gain/loss calculations when you dispose of part of your holdings. Your backtest should track lots and ACB changes, not just average prices.
• Superficial loss rule: Consider how repurchasing substantially identical property within a short window might impact the deductibility of capital losses. Model the possibility of deferred losses in your scenario analysis.
• Recordkeeping: Keep thorough records of trades, fees, funding payments, and transfers across wallets/exchanges. A robust trading journal helps reconcile tax outcomes with strategy performance.
• Foreign exchange and reporting: If you use non‑Canadian platforms, consider whether holdings may fall under foreign property reporting thresholds and factor this administrative reality into your operational plan.

Finally, stay aware of Canadian regulatory expectations around platforms and custody. Even if you’re an individual trader, understanding what your platform must comply with can inform your counterparty risk assessment and withdrawal planning.

Tooling: From Data to Simulation

You don’t need a hedge‑fund stack to run credible tests, but you do need reliable building blocks and a repeatable workflow.

Data Access and Normalization

• Use reputable data vendors or exchange APIs. For resilience, keep an independent historical archive rather than relying on on‑demand queries.
• Normalize symbols and timezones. Map BTC/CAD and BTC/USD where needed and ensure consistent decimal precision.
• Version your data sets so you can reproduce results even after a data provider revises history.

Backtesting Libraries and Analytics

• General purpose: dataframes and numerical libraries for quick iteration.
• Strategy testing: frameworks that support event‑driven or vectorized simulations, with hooks for custom cost and execution models.
• Order book simulation: libraries or your own modules to model depth, partial fills, and queue position for more advanced execution testing.
• Visualization: plotting libraries to produce equity curves, drawdown charts, and parameter heatmaps.

Keep the stack lightweight early on. As your strategy matures, add complexity—such as multi‑venue routing or options/funding schedulers—only where it materially improves realism.

A Step‑by‑Step Backtesting Checklist

1. Define the hypothesis: What market behavior are you trying to capture (trend, mean‑reversion, breakout on volume, funding dislocations)?
2. Pick markets and instruments: Spot, perps, or futures; BTC/USD or BTC/CAD; which exchanges and why.
3. Assemble and clean data: Source, reconcile, de‑noise, and document.
4. Design signals: Keep them simple and interpretable. Cap the number of tunable parameters.
5. Model costs: Fees, spreads, slippage, funding, borrow, on‑chain fees, and cash‑in/out delays (e.g., Interac e‑Transfer timing).
6. Build execution logic: Market vs limit, partial fills, queue effects, child orders, and API latency.
7. Risk and sizing: Volatility targeting, stops, and portfolio constraints.
8. Run walk‑forward testing: Rolling train/test, parameter stability checks, and noise injection.
9. Evaluate rigorously: Risk‑adjusted performance, drawdowns, capacity, and sensitivity to assumptions.
10. Stage deployment: Paper, small size, full size—with monitoring, post‑trade attribution, and continuous model validation.
11. Document tax assumptions: Model after‑tax P&L under reasonable Canadian scenarios and maintain detailed records for reconciliation.

Common Pitfalls—and How to Avoid Them

• Optimizing on the test set: Once you’ve seen out‑of‑sample results, that slice is no longer “clean.” Resist the urge to iterate using test feedback.
• Ignoring funding and borrow costs: Perps and margin trades can look great before costs; mediocre after.
• Assuming perfect liquidity: Especially problematic on CAD books or during off‑peak hours.
• Slippage underestimation: Use stress scenarios from live fills to calibrate your model.
• Data snooping: If you try enough ideas, something will appear to work by chance. Keep a research log and limit simultaneous hypotheses.
• Operational blind spots: Ignoring deposit/withdrawal delays, maintenance windows, or custody policies can derail a “good” strategy.

Example: Turning a Simple Idea into a Robust Test

Suppose you want to test a basic momentum strategy that goes long when a fast moving average crosses above a slow one and exits when it crosses back. Here’s how to approach it rigorously:

1. Choose BTC/USD spot on a liquid global venue for cleaner execution assumptions. Later, replicate on BTC/CAD to gauge CAD‑specific slippage.
2. Source 1‑minute candles and verify completeness; resample to 15‑minute bars for signals, but execute with micro‑slippage assumptions at the next bar’s open plus half‑spread.
3. Model taker fees; only claim maker rebates if you enforce post‑only logic and realistic fill criteria.
4. Apply volatility targeting so position size scales down during high‑volatility regimes.
5. Run a 12‑month train / 3‑month test walk‑forward over several years. Record Sharpe, Sortino, max drawdown, and turnover per fold.
6. Inject ±50–100 ms order latency and a small probability of partial fills to reflect busy markets.
7. Validate on BTC/CAD with higher slippage and evaluate capacity limits relative to typical Canadian book depth.
8. Estimate after‑tax results under both capital gains and business income assumptions, using ACB tracking for the former.

By the end, you’ll know not just whether the signal “works,” but whether it works after costs, within your liquidity constraints, and with operational frictions that mirror your real setup.

Operational Readiness: What to Confirm Before Scaling

• Counterparty and custody: Where are assets held? What are withdrawal policies, whitelists, and cooldowns?
• Capital flows: For Canadian bank rails, confirm daily limits and typical Interac e‑Transfer timing. For global venues, confirm stablecoin deposit speeds and on‑chain fee budgets.
• Order types and risk controls: Ensure the live venue has the exact order semantics you modeled (reduce‑only stops, post‑only, trigger conditions).
• Monitoring: Live dashboards for positions, P&L, latency, errors, and funding events. Include kill‑switches.
• Recordkeeping and reconciliation: Automate trade exports, fees, and funding logs. Keep an auditable trail to support both performance reviews and Canadian tax reporting.