MADX Cobra
A trend-continuation filter built from two moving-average bands and a fast ADX, entered on a resting limit order.
How it works
The setup is a trend filter made of two bands and a momentum check. A 200 period band, the black zone, decides direction: while price closes above it you may only buy, while it closes below you may only sell, and while it closes inside you do nothing. A 10 period band, the blue zone, times the entry. A 5 period ADX above 40 is the confirmation that the move is trending rather than drifting.
The verdict. Across five years and 10,692 trades it made $80,720 on a $50,000 account. That sounds like a result until you look at what it cost to get there. The average trade returned 0.018R, which is close enough to nothing that costs decide the outcome. The deepest drawdown was $88,440, which is more than the entire five year profit and more than the account it was traded on.
Would a bigger account fix it? No, and the reason matters more than it first appears. Scale everything up and nothing changes: ten times the size turns $80,720 of profit into $807,200 and turns the $88,440 drawdown into $884,400 alongside it. Hold risk at $500 instead and simply double the account, and the drawdown is still 88% of it. For a funded account it is worse again, because a $100,000 evaluation typically carries a trailing drawdown limit somewhere near $3,000, and this strategy exceeds that by more than twenty times. Account size is irrelevant when the rule that ends you triggers at 3%, not at 100%.
The rules we tested
- Build the black zone from a 200 period average of typical price, with the 200 period averages of the highs and the lows as its edges
- Build the blue zone the same way, from 10 period exponential averages
- Buy only while price closes above the black zone, sell only while it closes below, and stand aside while it closes inside
- Take a signal when a candle closes beyond the blue zone in that direction and the 5 period ADX is above 40
- Do not buy that close. Rest a limit order 8 ticks better than it, below for longs and above for shorts
- Fill only when a later candle trades a full tick through that price. A candle that touches the level is not a fill
- Leave the order working for six candles, thirty minutes, then cancel it
- Cancel early if price closes back through the blue zone, if the black zone regime is lost, or if a contract roll lands on a working candle
- Put the stop at the opposite outer edge of the blue zone, measured from the price you actually filled
- Exit when a candle closes on the far side of the blue zone. There is no profit target
- Instrument
- NQ
- Timeframe
- 5 minute
- Session
- 24 hour
- Test period
- Aug 9, 2021 to Aug 7, 2026
- Account size
- $50,000
- Risk per trade
- $500
The results
| Year | Trades | Win rate | Avg per trade | Total | Net P&L |
|---|---|---|---|---|---|
| 2021 Partial | 836 | 32.3% | +0.10R | +86.86R | $48,783 |
| 2022 | 2,187 | 31.0% | +0.01R | +15.32R | -$22,840 |
| 2023 | 2,128 | 28.3% | -0.04R | -83.66R | -$41,179 |
| 2024 | 2,154 | 29.9% | +0.04R | +90.15R | $20,018 |
| 2025 | 2,126 | 31.1% | +0.01R | +27.83R | $38,968 |
| 2026 Partial | 1,261 | 33.5% | +0.05R | +60.71R | $36,971 |
| All Total | 10,692 | 30.7% | +0.018R | +197.21R | $80,720 |
Largest drawdown over the period: 162.7R.
Net of costs: $4.00 round turn commission per contract, 1 tick slippage on the exit.
Example trades
How we tested it
This is the part that decides whether the numbers above mean anything.
Data. NQ 1-minute data from Databento, combined into 5-minute candles and run around the clock rather than in session, because a 200 period band needs a continuous series to mean anything.
Costs. $4.00 round turn commission per contract, and one tick of slippage on the exit only. The entry is a resting limit order, and a resting order fills at its price or better, so charging it entry slippage would be charging for something that cannot happen.
Fills. A limit counts as filled only when a later candle trades a full tick through it. A candle that merely touches the level does not count. That distinction is not pedantry: allowing touch fills turns this strategy's result from a loss into a profit, which is exactly why the stricter test is the only defensible one.
Micro against full size. The gross result is identical on both contracts at +495.58R, because the strategy cannot tell which one you are trading. Everything separating them is cost, and the micro is the more expensive of the two:
- NQ at $20 a point: $9.00 per contract per round turn, and +197.21R after costs
- MNQ at $2 a point: $1.24 per contract per round turn, and +84.49R after costs
The micro looks cheaper and is not. Commission does not shrink by the same factor the contract does, so on MNQ you pay $0.37 of commission for every dollar per point of exposure against $0.20 on NQ. Ten micros cost more than one full-size contract for the same position.
Data: Databento NQ 5-minute, 24-hour bars.