Pipsgrowth EX16004 Trend

The 2-bar reversal EA built for traders who don't trust indicators.

Pipsgrowth EX16004 Trend runs on a deliberately minimalist premise: read two consecutive closed bars, and act only when the second one fails to follow through on the first. There is no moving average, no oscillator, no multi-timeframe stack — just bar[1] and bar[2] doing the talking, and a 14-period Average True Range doing the math for the stop distance and trailing ratchet. That compactness is the EA's defining feature. In a PipsGrowth corpus full of multi-indicator trend followers, EX16004 stands out as the price-action native — the strategy that argues two candles of honest rejection are worth more than five overlapping filters.

The two patterns are mirrors of each other, and both encode a specific market story: a failed breakout in one direction, immediately followed by a bar that closes back inside the prior range. The 2-bar reversal is, in market-microstructure terms, a liquidity grab — a stop hunt beyond the obvious level that produces a wick on bar[1] and a body close back on the right side of the prior bar's extreme.

A BUY fires when bar[2] is bearish (close[2] < open[2]) AND bar[1] makes a new low (low[1] < low[2]) but bar[1] closes back above bar[2]'s low (close[1] > low[2]). The bear pattern is already in motion; bar[1] extends it to a fresh extreme; but the buyers reclaim the bar and close back inside. Buyers won.

A SELL fires symmetrically: bar[2] is bullish, bar[1] takes out the prior high (high[1] > high[2]), but bar[1] closes back below bar[2]'s high. The bull pattern was the bait; the bar[1] upper wick is the trap; the close is the rejection.

Both signals use fully closed bars (bar[1] and bar[2]), so the entry fires only on the open of bar[0] — no repaint, no live-bar re-evaluation. The same closed-bar discipline applies to the ATR(14) read in CheckSignal(): SL/TP are computed off bar[1]'s ATR, not the still-forming bar[0].

Position management: the 14-period ATR does triple duty

Initial stop is ATR × InpATR_Multiplier (default 2.0), and the take-profit is stop × InpRiskReward (default 2.0), giving a fixed 1:2 reward-to-risk on the static level placed at entry. Once in profit, the ManageOpenPositions() function runs every tick and applies a two-stage ratchet:

Breakeven — when profit distance ≥ ATR × InpBEStart_ATR (default 1.0), the stop moves to open + InpBEOffset × Point for buys (and open − offset for sells). The 30-point offset means the breakeven stop is placed slightly past entry, not exactly at it, so a small spread-induced whipsaw can't re-stop the trade after the BE fires.

Trailing stop — when profit distance ≥ ATR × InpTrailStart_ATR (default 1.5), the stop follows price at a distance of ATR × InpTrailDist_ATR (default 0.5). A tighter trail (0.5× ATR) is intentionally less generous than the breakeven trigger (1.0× ATR), so the trade has room to breathe before the trail kicks in. The trailing stop never goes below entry (the trailSL > openPrice check on line 447 enforces this).

Both stages are wired through TryModify_EX16004(), which retries up to 3 times on REQUOTE / TIMEOUT before giving up — a small but real protection against broker requote storms on D1-bar opens when volatility spikes.

Pyramiding (off by default)

InpEnablePyramid = false is the shipped default. When the trader turns it on, the EA permits up to InpMaxPositions = 2 in the same direction, but every additional entry is gated by IsGridSafe() — a guard that walks the open-position list and refuses to add unless every existing same-direction position is in profit by at least InpMinProfitPoints = 300 points. The 300-point floor (3 pips on a 5-digit pair, 30 points on XAUUSD) prevents the EA from averaging into a loser; it can only scale into a confirmed winner. A blocked pyramid logs a Pyramid BLOCKED line to the journal with the actual profit distance and the required threshold.

Drawdown protection

CheckDrawdownLimits() compares two real-time metrics each tick:

• Daily loss — InpMaxDailyLoss = 3.0% of g_dailyStartBalance (the balance at the start of the trading day, refreshed at CheckNewDay() at server midnight).
• Equity drawdown from peak — InpMaxEquityLoss = 10.0% of g_peakEquity (the highest equity seen since the EA attached, updated each tick).

Either threshold trips g_ddLimitHit, which halts all new entries for the rest of the day. If InpCloseOnDDLimit = true (the default), the EA also fires CloseAllPositions() to flatten the book. The day reset happens at server midnight via CheckNewDay(), which compares a YYYY.MM.DD string to g_lastDayTime — so a Sunday re-attachment after a weekend flat does not inherit a stale daily-loss flag.

Time filter

CheckTimeFilter() enforces a Monday-to-Friday window. By default it accepts all 24 hours (InpStartHour = 0 to InpEndHour = 23), but the trade decision happens on D1 bars, so the actual decision point is the daily bar open — not the minute clock. InpFridayCloseHour = 20 calls CloseAllPositions() on Friday at 20:00 server time, which means the EA is flat going into the weekend. Weekend trading is blocked by the day_of_week == 0 || day_of_week == 6 check.

Slippage, spread, retries

InpSlippage = 30 points is the cap passed to trade.SetDeviationInPoints(). InpMaxSpread = 250 points (2.5 pips on a 5-digit FX pair, 25 points on XAUUSD) is the no-trade gate. For D1/H4 trading this is loose — it filters out the worst illiquidity events but lets through normal spreads on a standard ECN book. InpMaxRetries = 3 is the in-loop retry count for OpenPosition(); after each failed attempt the code Sleep(100), refreshes the symbol, and tries again.

Sizing and account cap

CalculateLotSize() resolves a risk-based lot: riskMoney = balance × InpRiskPercent / 100 (default 1%), divided by the per-lot loss at the stop distance, then NormalizeLot()-clamped to the symbol's lot step, between minLot and MathMin(maxLot, InpMaxLotSize = 5.0). Setting InpRiskPercent = 0 falls back to InpFixedLot = 0.01 for the all-fixed-size mode. Before sending the order, OpenPosition() calls OrderCalcMargin() to confirm the account has enough free margin — no partial fills, no rejected orders from under-margined attempts.

Fill policy

SetProperFillingMode() reads SYMBOL_FILLING_MODE at OnInit and chooses FOK, IOC, or RETURN based on what the broker advertises. Exness / IC Markets / Pepperstone support is therefore automatic — no input needed and no per-broker template required.

Chart panel

A 220×140 dark-slate-gray rectangle in the top-left corner with six labels: title (gold), magic number, current ATR, spread, buy/sell position counts, and live P&L (lime if positive, red if negative). The panel is created in CreateInfoPanel() and refreshed on every OnTick() by UpdateInfoPanel(). All chart objects are prefixed EA_ and cleared in OnDeinit() so that removing the EA from the chart leaves no orphan objects behind.

What it does not have

There is no OnTester() function in the source — the brief in the file header lists "OnTester" as one of 12 layers, but the .mq5 does not actually implement a custom optimization criterion. There is no multi-timeframe read (the EA only sees _Period bars), no news filter, no session-specific behavior beyond the Mon-Fri and Friday-close gates, and no hedging logic — IsGridSafe() operates only on same-direction positions, so a buy and a sell can coexist on the same symbol if both patterns fire on adjacent bars. TryClosePartial_EX16004 is defined as a 3-attempt retry helper but is not wired into any live call in this source — the EA never partial-closes.

What to expect in backtest

D1 reversal patterns are rare — typical frequency is 1-3 signals per pair per month on liquid FX, more on XAUUSD where the daily range is wider. The 1:2 RR and the trailing ratchet are designed to make the few winners count: a single trend day that runs 3× ATR after the trail activates (which kicks in at 1.5× ATR) can return 2-3× the initial stop. The 3% daily DD cap means the EA will not fight a bad day past the third percent; the 10% equity-peak cap means the equity curve can give back up to a tenth of its high-water mark before the EA goes quiet for the session.