The Scalping Backtest Gap
Standard MT5 Backtest
Modelling Quality
25%
Spread Setting
0 pips (default)
Slippage
0 (default)
Data Source
1-minute OHLC bars
Shown Result
$8,000 profit / 6 months
Real Trading Conditions
Real Tick Data Quality
99%
Real Spread (ECN)
10–15 pips
Real Slippage
1–2 pips per trade
Data Source
Real tick-by-tick data
Actual Result
$2,400 profit / 6 months
(or break-even)
This 3× gap is not a coincidence
Standard backtests with 0 spread and 25% modelling quality are structurally incapable of representing what scalping bots actually experience live. Every single winning trade in the standard backtest is inflated by the missing spread cost. Fixing these settings closes the gap between backtest and live performance significantly — but most traders never take this step.
How to Backtest a Gold
Scalping Bot Before Going Live
Published 15 June 2026 · 11 min read
Standard MT5 backtests with 25% modelling quality and 0 spread are systematically misleading for scalping bots — typically overstating results by 2–5×. The minimum requirements for a trustworthy scalping backtest are: real tick data (99% modelling quality), realistic spread (12–15 pips for XAUUSD ECN), and 2–3 point slippage. Without all three, the results cannot be trusted as a basis for live deployment.
Scalping Backtest Setup Checklist
Before running or trusting any gold scalping bot backtest, verify all 7 items below. Click each item for details on why it matters and how to verify it.
Tick data downloaded (not just 1-minute OHLC)?
Tick data captures every price change within each bar. Without it, intra-candle spikes that trigger SL or TP on a scalping bot are invisible — the bot appears to exit at bar close when in reality it may have exited mid-bar at a much worse price.
Modelling quality ≥ 99% shown in Strategy Tester?
Open Strategy Tester → check the "Modelling Quality" line in the report. Below 90% means reconstructed tick data, not real ticks. At 25% (OHLC only) you are essentially guessing.
Spread set to your real broker spread (not 0)?
Right-click on the XAUUSD symbol in Strategy Tester and manually set the spread. For ECN brokers, use 100–150 points (10–15 pips on standard XAUUSD pricing). Do NOT leave it at 0.
Slippage set to 2–3 points?
In Strategy Tester → Expert Properties → Testing tab, set "Execution" to "Based on Real Ticks" and slippage to 2–3 points minimum. Scalping is disproportionately affected by slippage because it represents a larger fraction of the total profit target.
Testing period covers at least 2 years?
One year is not enough. XAUUSD can spend an entire year trending in one direction, which will flatter any directional scalping bot and mask its true range-bound performance.
Testing period includes one range-bound AND one trending year?
Confirm by looking at a XAUUSD yearly chart. If your 2-year period is 2019–2020, it includes a moderate 2019 and the extreme 2020 COVID spike — that is one type of mix. 2022–2023 gives you the commodity surge and then consolidation — a different mix. Both are needed for a complete picture.
Spread was NOT set to 0 in any externally provided test data?
If you are evaluating someone else's backtest, look for the "Spread" row in the Strategy Tester report. If it says 0 or is missing entirely, the test is unreliable for a scalping bot and you should request a re-run.
Backtest Reliability Score
0%
Unreliable — results cannot be trusted for scalping
0 / 7 items verified
Why Tick Data Is Non-Negotiable for Scalping Bots
When a scalping bot opens a trade and sets a 15-pip stop loss, what happens if price dips 16 pips intra-candle before recovering and closing above your entry? In real trading, the stop loss triggers and the trade is closed at a loss. In a 1-minute OHLC backtest, the bar shows open, high, low, and close — and if the close is above your entry, the trade may appear as a winner, ignoring the low entirely.
Scalping bots are particularly vulnerable to this problem because:
Tight stop losses are close to normal price fluctuation
A scalping bot with a 10-pip SL can have that stop triggered by routine intra-candle noise that a swing trader would never notice. Without tick data, the tester does not know about that spike.
Entry timing within the candle matters
If your scalping bot has breakout entry logic that should trigger when price crosses a specific level, that level may be crossed for milliseconds within a 1-minute bar without appearing in the OHLC summary. With tick data, the exact entry point is modelled. Without it, the entry is approximated to the bar open, which is almost always incorrect for breakout scalpers.
Trailing stops are highly tick-sensitive
An EA with a trailing stop (like Goldie Razor V2.8.4's 6-level trailing logic) needs tick-by-tick price data to simulate accurately when each trail level is triggered. Simulated tick data produces trailing stop exits that are systematically too optimistic.
Exact MT5 Settings for a Scalping Backtest
Here is the exact configuration for a trustworthy gold scalping bot backtest in MT5 Strategy Tester:
Open Strategy Tester
In MT5: View → Strategy Tester (or Ctrl+R). Select your EA from the Expert dropdown.
Select XAUUSD symbol
Choose XAUUSD from the Symbol dropdown. Make sure tick data has been downloaded first via Tools → History Center.
Set Modelling to "Every Tick Based on Real Ticks"
In the Model dropdown, select "Every Tick Based on Real Ticks" — NOT "Every Tick" (which uses simulated ticks from OHLC) and NOT "Open prices only".
Set spread manually
Right-click on XAUUSD in the Symbol dropdown → click the properties icon. Set "Spread" to 100–150 (this is in points — 100 points = 10 pips on standard XAUUSD). Match your ECN broker's actual average spread.
Configure slippage in Expert Properties
Click the Expert Properties button → Testing tab. Set "Deviation" (slippage) to 20–30 points (2–3 pips). This is the maximum slippage the Strategy Tester will simulate on entries and exits.
Set date range to 2+ years
In the From/To date fields, select a minimum 2-year period. Include at least one trending year and one range-bound year for XAUUSD.
Check modelling quality in results
After running, the Strategy Tester report shows "Modelling quality" at the top. Confirm it shows 99%. If it shows 25% or 90%, your tick data was not properly downloaded or imported.
What to Expect When You Switch from 25% to 99% Modelling Quality
The results will almost always be lower. This is not a sign that the EA is bad — it is a sign that your new results are accurate. The original 25% results were the illusion; the 99% results are reality.
What to do with the new results:
Still profitable at 99%
Proceed with caution — results are trustworthy basis for demo testing
Marginal at 99%
Investigate whether spread reduction (better broker) or settings adjustment can improve it before demo
Losing at 99%
Do not proceed to demo — the strategy is not viable under realistic conditions. Revisit fundamentally.
Related Reading
How to Backtest Without Getting Fooled
Detecting misleading backtest results from vendors or third parties.
Backtesting a Gold EA Properly
The complete end-to-end backtesting methodology for any XAUUSD EA.
Demo Account Testing
Validating backtest results on a live demo account before real money.
Can You Really Make Money Scalping Gold?
The profitability maths for gold scalping — what the numbers say.
Slippage Impact on Gold Scalping
Why slippage matters more for scalpers than any other strategy type.
Frequently Asked Questions
Open MT5 → Tools → History Center. Find XAUUSD in the instrument list, select the M1 (1 Minute) timeframe, and click Download. This downloads raw tick data from your broker's server. For older historical data (3+ years), you may need a third-party tick data provider such as Dukascopy, which offers free historical XAUUSD tick data in .csv or .bin format that can be imported into MT5.
Modelling quality is a percentage shown in Strategy Tester reports that indicates how accurately the price data was reconstructed for the test. 99% means real tick data was used — every price update was captured. 90% means high-quality 1-minute bar data was interpolated into ticks. 25% means only OHLC bars were used, with the open, high, low, and close prices interpolated in a fixed pattern. For scalping bots, 25% is effectively useless because the interpolation creates artificial, regular price movement that does not reflect real intra-bar price discovery.
The impact is proportional to how small your profit target is. For a scalping bot targeting 15 pips: a 12-pip spread represents 80% of the target, and 2-pip slippage adds another 13%. In a 0-spread backtest, the same bot shows a 15-pip gain per trade. In a realistic backtest, it shows 15 - 12 - 2 = 1 pip per trade — and many trades that looked profitable become losers. This is why scalping bots commonly look 2–5× better in standard backtests than they perform live: the difference is entirely in execution costs.
Not entirely — it can give you a rough directional indication of whether a strategy is fundamentally viable (e.g. if it loses at 25% modelling quality, it will almost certainly lose at 99%). But for any decision about live deployment, a 25% modelling quality result is insufficient. The absolute numbers — profit, drawdown, win rate — cannot be trusted. Only use 99% tick data results as the basis for go/no-go decisions on a scalping bot.
Goldie Razor V2.8.4 trades M15 breakouts with 20–35 pip targets — it is not a pure scalping bot. But its 6-level trailing stop logic means that positions can be exited at various points within a candle depending on where the trailing stop is triggered. Without tick data, those intra-candle exits are approximated rather than modelled precisely. The spread and slippage settings also matter even for non-scalping strategies: a 12-pip spread on a 25-pip target is 48% of the trade — significant enough to require accurate modelling at 99% quality for reliable results.
Goldie Razor V2.8.4
M15 breakout + H4 EMA filter — built for XAUUSD on MT5