Why Results Differ

​

When “different results” are normal

Different backtest results are not always a bug. Results can differ because of:

• different data sources and coverage
• different evaluation timing (close vs intrabar)
• different execution assumptions (fills, fees, slippage)
• different indicator implementations or defaults
• different position sizing rules

This page shows the most common causes and a simple diagnostic flow.

---

​

1) Data source and coverage differences

Two platforms rarely use identical historical data. Differences may include:

• missing candles or gaps
• different exchange source
• different candle construction rules
• different listing start dates

Fix:

• confirm the backtest range and data coverage

Data coverage

Understand what data your test is using.

---

​

2) Timeframe mismatch

A strategy is defined on a timeframe. Even a small mismatch (15m vs 1h) changes:

• indicator values
• crossover points
• trade count
• volatility profile

Fix:

• confirm the exact timeframe used on both sides
• do not compare a “similar” timeframe

---

​

3) Candle-close vs intrabar evaluation

Many platforms allow intrabar logic or simulate fills inside candles. Trinigence evaluates logic on candle close by default. Impact:

• entries and exits may happen later than expected
• crossovers trigger only when present at close
• stop/take profit behavior may differ if another tool assumes intrabar fills

Fix:

• compare on candle-close mode where possible
• validate trades, not just metrics

How backtesting works

Learn the evaluation model behind the simulator.

---

​

4) Indicator defaults and parameter differences

Indicators often have:

• default lengths
• smoothing rules
• internal calculation variants

Example:

• one tool uses EMA smoothing for RSI
• another uses Wilder’s smoothing (standard)

Fix:

• explicitly specify parameters in your strategy description
• avoid relying on “platform defaults”

Indicator parameters

See how parameters and defaults change behavior.

---

​

5) Execution assumptions (fees, slippage, fill model)

Even with identical trades, results differ due to execution modeling:

• maker/taker fees
• slippage assumptions
• limit vs market fill behavior
• partial fills vs full fills

Fix:

• compare settings side-by-side
• ensure fees and slippage are aligned

---

​

6) Position sizing differences

PnL depends on sizing. Differences can come from:

• fixed size vs % of equity
• compounding vs static sizing
• leverage assumptions
• one-trade-at-a-time vs multiple positions

Fix:

• compare normalized metrics (profit factor, expectancy)
• align sizing rules when comparing PnL

---

​

7) Schedule & filters differences

Time filters can silently eliminate trades. Examples:

• “weekdays only” applied in one platform but not in another
• timezone differences shift session windows

Fix:

• compare schedule definitions
• confirm timezone

Time & session logic

Understand session/time rules and timezone impact.

---

​

A fast diagnosis workflow

When results differ, do this:

---

​

Common patterns you’ll see

---

​

What to read next