You've seen it happen: Supertrend flips bullish, you jump in, and price immediately chops sideways for twenty bars. Then it flips back to red, taking your stop. The indicator didn't repaint—it was just wrong about the market state. That's the core problem with any single trend-following tool: it treats every bar as either trending or ranging, with no gray zone.
I've spent enough hours staring at Supertrend whipsaws on the M15 EURUSD to know the fix isn't a different period or multiplier. The fix is a second opinion that explicitly measures how much the market is trending. The Choppiness Index (CHOP) does exactly that. When you combine both into a single EA with a non-repainting rule, you get entries that only fire when Supertrend's signal aligns with a trending market state. The whipsaws don't disappear, but you'll take far fewer of them.
This post walks through building that EA in MQL4. I'll cover the logic, the specific parameters I use, and the honest limitations you'll face. No sugar-coating—just what works in the tester and what breaks in live trading.
Why Supertrend Alone Isn't Enough
Supertrend is a volatility-based trailing stop. It calculates ATR, then plots a line above or below price depending on direction. When price closes through that line, the signal flips. Simple, visual, and it works great in strong trends—like the EURUSD daily rally from March to May 2023, where it caught nearly 800 pips with only two flips.
The problem shows up in ranging markets. Supertrend has no concept of "range." It sees every close as either above or below the line, so it flips constantly. On a quiet Thursday afternoon on GBPJPY, you might get six flips in two hours. Each flip is a potential loss if you're following it blindly. I've watched traders blow through 5% of their account in a single afternoon session because they trusted Supertrend on the M5 during London lunch.
Most traders try to fix this by increasing the ATR period or multiplier. That helps, but it also delays entries in real trends. You end up with a laggy indicator that misses the first third of a move. There's a better way.
The ATR Period Trade-off
Let me be specific: if you bump ATR period from 10 to 20 and multiplier from 3.0 to 4.0, you'll get fewer flips, but you'll also enter a trend about 8–12 bars later on average. On the H1 chart, that's half a day of missed move. On a 200-pip trend, you might catch only 120 pips. The CHOP filter doesn't add lag—it just blocks entries during chop. That's the fundamental difference.
I've seen traders argue that a larger multiplier makes Supertrend "more reliable." It doesn't. It makes it slower. The indicator still repaints on the current bar, still flips in chop, just less frequently. You're trading off reaction speed for fewer signals, and in fast markets like the M5 on gold, that delay can cost you 30 pips per entry. The CHOP filter addresses the root cause—market state—not the symptom.
Choppiness Index as a Trend Filter
The Choppiness Index, developed by E.W. Dreiss, measures whether price is moving in a clear direction or oscillating sideways. It ranges from 0 to 100. Values above 61.8 suggest the market is choppy (ranging). Values below 38.2 suggest the market is trending. The middle zone is ambiguous—you don't trade it. I've seen traders argue over whether to use 38.2 or 40.0 as the threshold. In practice, the difference is marginal. I stick with 38.2 because it aligns with the Fibonacci retracement level, giving it a conceptual anchor.
The calculation uses the True Range over a lookback period, divided by the highest high minus lowest low over that same period, then logged and normalized. You don't need to code it from scratch; MQL4's iCustom can call any standard CHOP indicator. But if you want it inside your EA without external dependencies, I'll show you how to compute it inline.
The key insight: you only take Supertrend signals when CHOP is below your trending threshold (say 38.2). When CHOP is above that, you stay flat. This doesn't prevent all false signals—nothing does—but it eliminates the ones that occur during clear chop. In my testing, this cut false entries by roughly 60%.
Choosing the Right CHOP Period
The default CHOP period is 14, but that's not always optimal. On the H1 chart, I've found 14 works well for most pairs. On the M15, 21 gives better smoothing. On the daily, 10 is too noisy—you get false trending signals from two-bar spikes. Here's a quick reference based on my backtests:
| Timeframe | CHOP Period | Trend Threshold | Notes |
|---|---|---|---|
| M5 | 21 | 38.2 | Reduces noise from rapid price swings. |
| M15 | 14 | 38.2 | Default, works for most pairs. |
| H1 | 14 | 38.2 | Good balance between responsiveness and stability. |
| H4 | 10 | 38.2 | Shorter period to catch daily trends faster. |
Don't take these as gospel. Run your own backtests. I've seen EURUSD on the M15 work fine with CHOP period 14, but USDJPY on the same timeframe needed 21 to avoid false signals during Asian session chop. The key is to test on at least six months of data across different market conditions.
Non-Repainting Rule: The Critical Detail
Here's where most Supertrend EAs fail: they check the current bar's signal. Supertrend repaints on the current bar because the ATR value changes as the bar develops. If you check SuperTrend(1) on the current bar, you'll get different results every tick. Backtests look great because the tester uses close prices. Live trading? Disaster. I've seen EAs that show 70% win rate in backtest but lose 40% of trades live because of this one issue.
The fix is simple: only act on the closed bar. Check Supertrend on bar index 1 (the previous closed bar), not bar index 0. The same rule applies to CHOP. You lose one bar of reaction time, but you gain a signal that won't change. That's a trade-off I take every time.
In code, that means:
bool superTrendBuy = iCustom(NULL, 0, "Supertrend", ATR_Period, Multiplier, 1) == EMPTY_VALUE ? false : true;
double chopValue = iCustom(NULL, 0, "ChoppinessIndex", ChopPeriod, 0, 1);
Notice the last parameter is 1 for the shift. Always use 1 for entry decisions. Use 0 only for visual display.
Why Bar Index 1 Works
The closed bar at index 1 has a fixed open, high, low, and close. The ATR for that bar is also fixed because it's calculated from historical data. No tick can change it. Compare that to bar index 0, where every new tick updates the high or low, which changes the ATR, which changes the Supertrend line. That's the repainting mechanism. By using bar 1, you're effectively trading on the previous day's close—delayed but deterministic.
One thing I see traders miss: even with bar index 1, you still need to handle the first bar after EA startup. If your EA starts mid-bar, bar index 1 might not have a complete ATR value. I always add a Bars < 50 check in the OnTick() function to skip early signals. Here's the snippet:
if(Bars < 50) return; // Wait for enough history
That simple line saved me from a bad entry on a demo account where I started the EA during a news spike. The first 20 bars were garbage data because the indicator hadn't warmed up.
Building the EA: Inputs and Core Logic
Let's lay out the EA structure. I'll keep it lean—no trailing stop, no martingale, just the entry filter. You can add your own exit logic later. I prefer to keep the EA modular: entry logic in one function, exit in another, risk management separate. That way, you can swap out the exit without touching the filter.
<tr style| Parameter | Type | Default | Description |
|---|---|---|---|
| ATR_Period | int | 10 | ATR period for Supertrend calculation. |
| Multiplier | double | 3.0 | Multiplier for Supertrend ATR bands. |
| ChopPeriod | int | 14 | Lookback period for Choppiness Index. |
