Pullback Sniper on Micro Gold (MGC)
Should you use it?
Don't useDon't use it on MGC
The best set the sweep found does not hold up. The settings and results below are shown for reference only, not as a recommendation.
- Misses the brief (profit factor above 1 and net profit at least 3× max drawdown) in NinjaTrader's own Strategy Analyzer.
Best set found
MGC · 30-minute bars · Long · EMA-100 trend filter on both sides · full session, no flatten · 95-tick (9.5 pt, $95) stop · 32-tick (3.2 pt, $32) target
Net profit
+$2,292
1 MGC contract · after commission
Profit factor
1.16
$1.16 won per $1 lost
Max drawdown
$807
net / drawdown 2.8 : 1
Win rate
77%
747 trades · 1.4 per day
Per month
+$91
largest loss −$125
One contract · NinjaTrader 8 Strategy Analyzer, Aug 2024 – Sep 2026, after commission.
What the numbers say
- It falls short of the brief. Target: profit factor above 1.0 and at least 3 : 1 net profit to max drawdown. Result (NT8 Strategy Analyzer, with the “NinjaTrader Brokerage Free” commission template): PF 1.16 and 2.8 : 1 on 747 trades.
- Profitable in every calendar year of the test (2024: +$408, 2025: +$1,464, 2026: +$1,360); 17 of 26 months positive; best 2024-09 (+$523), worst 2026-07 (−$263).
- Long / short split. Longs +$2,292 on 747 trades, shorts $0 on 0.
- Exposure. Average 51 minutes in the market, 1.4 trades per session; every trade closed within the Globex session it was opened in (0 at the 15:55 flatten, 42 at the 17:00 session close), nothing carried across the 17:00 break or a roll — but note that overnight entries are held through the night session.
- Walk-forward. The in-sample winner failed forward at 3 of 4 cut-offs. The recommended set was positive forward at 4 of 4 cut-offs (forward PF 1.27–1.36, net/DD 3.7–6.2) and would have passed the in-sample gate at 1 of 4 of them — positive forward but not something an earlier search would have selected.
- Plateau. 3 of 72 stop/target cells in the same family pass PF > 1 and net/DD ≥ 3 over the full period — an isolated optimum.
- Verified. The replica behind the sweep matched NT8’s trade count on the validation window (744 vs 747); headline figures come from NT8’s own engine.
The best settings found (reference only)
Not a recommendation
These are the settings the sweep liked best on this instrument, and they still fall short. They are here so you can see what was tried.
Add ToredaPullbackStrategy to a MGC 30-minute chart (or the Strategy Analyzer) and change only these inputs — everything else stays at the strategy default (shown for reference; see the verdict above):
| Input | Value | Note |
|---|---|---|
| Data series | MGC, 30 Minute, Last | Front-month contract; NT8 merge policy “MergeBackAdjusted” supplies the history. |
| Direction (Pullback) | Long | Longs 747 / shorts 0 trades. |
| Bull / Bear EMA Length | 100 | Default 50. |
| Use Bull EMA Filter / Use Bear EMA Filter | true / true | Default off. |
| Stop Unit / Stop Ticks (Risk) | Ticks / 95 | 9.5 MGC points = $95 per contract before slippage. |
| Target Unit / Target Ticks (Risk) | Ticks / 32 | 3.2 MGC points = $32. |
| Use Breakeven / Use Trail | false / false | |
| Use Time Window (Session) | false | Whole ETH session. |
| Use Flatten Time / Flatten | false | Positions run to stop, target or the 17:00 session close. |
| Max Trades Per Day | 0 | 0 = unlimited. |
| Analyzer: order fill resolution | High, 1 Minute | Matches the runs in this report; commission template “NinjaTrader Brokerage Free” (≈ $2.40 per round turn). |
1 contract vs 3 contracts
One position at a time, so size scales cleanly: the ratios stay the same and every dollar figure is three times larger.
| 1 contract | 3 contracts | |
|---|---|---|
| Net profit (after commission) | +$2,292 | +$6,877 |
| Gross profit / gross loss | $16,822 / −$14,530 | $50,466 / −$43,589 |
| Commission paid | $1,793 | $5,378 |
| Profit factor | 1.16 | 1.16 |
| Max drawdown | $807 | $2,422 |
| Net / max drawdown | 2.8 : 1 | 2.8 : 1 |
| Profit per month | +$91 | +$274 |
| Average winner / loser | $29 / −$84 | $88 / −$253 |
| Largest loss | −$125 | −$376 |
| Worst losing streak (3 trades) | −$253 | −$760 |
| Risk per trade at the stop | $95 + costs | $285 + costs |
| Suggested risk capital (≈ 3 × max drawdown) | $2,422 | $7,265 |
Year by year
| year | trades | net | PF | win |
|---|---|---|---|---|
| 2024 | 128 | +$408 | 1.17 | 76% |
| 2025 | 379 | +$1,464 | 1.20 | 77% |
| 2026 | 253 | +$1,360 | 1.29 | 79% |
| in-sample (→ 2025-12-31) | 508 | +$1,903 | 1.20 | net/DD 3.6 |
| out-of-sample (2026) | 253 | +$1,360 | 1.29 | net/DD 3.7 |
How robust is it?
Walk-forward check
The in-sample winner held up forward at 1 of 4 cut-offs; the recommended set was positive forward at 4 of 4 and would have passed the gate at 1 of 4. It works forward but was identified with the whole period in view — discount the headline accordingly.
Monte Carlo
The observed drawdown ($528) sits at the low end of what the same daily results could have produced — 85% of resampled histories draw down further, and 1 in 20 reaches $1,281 or more. The edge is not in doubt in this sample: 0.5% of resampled histories end below zero and the 3 : 1 brief holds in 72% of them. Reordering the same trades (second row) moves the drawdown a lot: the 95th-percentile sequence-risk drawdown is $1,349. Size the account on the 95th-percentile drawdown ($1,281 per contract) plus margin, not on the observed $528, and expect up to 195 trading days without a new equity high.
At the default settings
At its defaults the strategy loses on every timeframe tested.
| configuration (defaults) | trades | net | PF | max DD | win |
|---|---|---|---|---|---|
| 1m Both S40t T80t noEMA ETH | 19,910 | −$54,264 | 0.90 | $54,306 | 33% |
| 2m Both S40t T80t noEMA ETH | 14,922 | −$31,656 | 0.92 | $31,745 | 34% |
| 3m Both S40t T80t noEMA ETH | 12,153 | −$24,766 | 0.93 | $25,404 | 34% |
| 5m Both S40t T80t noEMA ETH | 9,073 | −$7,774 | 0.97 | $9,494 | 35% |
| 10m Both S40t T80t noEMA ETH | 5,689 | −$15,290 | 0.90 | $16,314 | 33% |
| 15m Both S40t T80t noEMA ETH | 4,126 | −$8,663 | 0.92 | $9,464 | 34% |
| 30m Both S40t T80t noEMA ETH | 2,234 | −$5,066 | 0.92 | $6,078 | 34% |
Know before you trade it
- Selection bias. 44,100 configurations were searched; the walk-forward is the only evidence not contaminated by the search. The recommended set was positive forward but earlier searches would not have chosen it.
- Plateau width. 3 of 72 neighbouring stop/target cells pass — the set is an isolated optimum and should be treated as unproven.
- Instrument-specific. Stops and targets are in MGC ticks; do not transfer them to another instrument — each instrument has its own report.
- Fill assumptions. Both engines fill stops at the stop price and targets on touch; live fills will be worse.
- Two years, one regime. The sample is Aug 2024 → Sep 2026; behaviour in other regimes is untested.
Getting started
- Do not chase the leaderboard. Re-optimising to whatever tops the in-sample ranking is the procedure the walk-forward check shows failing.
Pullback Sniper on other instruments
Backtest results are hypothetical. They were produced with the benefit of hindsight on historical data, do not represent actual trading, and cannot fully account for slippage, liquidity or the emotional pressure of live trading. Past performance is not indicative of future results. This content is for educational purposes only and is not financial advice. Trading futures involves substantial risk of loss. See our risk disclosure for details.