Pipsgrowth EX18026 TrendFollow

Pipsgrowth EX18026 TrendRider is a per-bar trend-following EA that decides direction from a self-built Kaufman Adaptive Moving Average (KAMA), not from a stock MQL indicator. The KAMA is reconstructed inside RefreshKAMA() over a 200-bar history window using the canonical Kaufman recurrence: an efficiency ratio over a 14-bar change-vs-total-path window drives a smoothing constant that morphs between a fast EMA equivalent (fast SC = 2 → α = 2/3) and a slow EMA equivalent (slow SC = 30 → α = 2/31), and the output is squared before being applied to the running KAMA value. The smoothing constant squared (α²) is what gives KAMA its character — it stays almost frozen in chop and snaps toward price in trends. Once the KAMA series is in hand, the EA does not look at the line directly. Instead it computes a per-bar signed score in RefreshKAMA() from three primitives: the KAMA slope across the last 3 bars (SLOPE_BARS=3), the difference between the most recent close and the KAMA on that bar, and the bar-over-bar change in close. Those three primitives are combined into a 4-quadrant scoring table. When the KAMA is rising and the close is above it and the close is also rising, the score is +1.0 × trendStrength × posStrength. When the KAMA is rising and the close is above it but the close is falling, the score drops to +0.5 × trendStrength. The opposite side is symmetric with negative values. The resulting per-bar signal is then clamped to [-1, +1] and stored in g_kamaSignal[].

That per-bar signal is aggregated by AggregateSignal() using a decay-weighted mean across the most recent 3 closed bars (SIGNAL_PERSISTENCE=3, SIGNAL_DECAY=0.8). The weights are 1.0 for the most recent bar, 0.8 for the bar before that, and 0.64 for the bar before that, then scaled into a -100..+100 percentage. A long fires when the aggregated score is at or above +70% and a short when it is at or below -70%; the InpSignalThreshold input exposes that 70% dial. The persistence+decay structure is what separates TrendRider from a raw KAMA crossover: a single noisy bar that just touches +100 cannot cross the threshold on its own, and a one-bar flip in the underlying KAMA direction has to persist for two more closed bars to drive the aggregate across the line. That gives the system natural whipsaw filtering without a separate confirmation indicator.

Regime gating happens before the signal is even consulted. UpdateRegime() runs each new bar and classifies the current state into one of seven buckets based on the ADX(14) reading, the ATR(14) percentile over a 100-bar lookback, and the Bollinger Bands(20, 2) width ratio on the prior closed bar. A Strong Trend needs ADX ≥ 30; a Weak Trend needs ADX ≥ 20. Breakout is reserved for ATR-percentile ≥ 0.85 paired with the Bollinger width growing at least 15% bar-over-bar. Expand is ATR-percentile ≥ 0.70 without the breakout-width growth. Compress is ATR-percentile ≤ 0.25 with a Bollinger-width-to-midline ratio below 0.03. Range is ATR-percentile ≤ 0.40. Everything else is Choppy. Choppy and Range are both hard-blocks inside NoTrade() — NoTrade() returns true and the EA refuses to open anything, with the reason string printed to the log. That is a deliberate design choice for a trend-follower: if the market is not trending, do not trade it.

Confirmation in ConfirmEntry() is a two-factor check. The first factor is the higher-timeframe EMA side: TrendRider builds an HTF EMA(50) and an HTF EMA(200) on whatever timeframe InpHTF resolves to through MapTimeframeInt(). With the default InpHTF=8, the input falls through the switch statement and the EA uses PERIOD_H1 by default — same default-fallback pattern as the rest of this family. A long is only permitted when the HTF EMA(50) is above the HTF EMA(200); a short is only permitted when the fast HTF EMA is below the slow. The second factor is the close of the most recent completed bar against the inlined KAMA on that same bar: a long requires close > KAMA, a short requires close < KAMA. If either factor fails, the entry is skipped and a [CONFIRM] ... line is printed.

The 22 inputs are organised into five groups. IDENTITY holds InpMagic=22218026, the trade comment string, the master enable switch InpEnableTrading=true, InpDryRun=true (default-on paper mode — positions are logged but not sent), and InpKillSwitch=false which forces the EA into close-only mode when flipped. SIGNAL (KAMA) exposes the ER period (14), the fast smoothing constant base (2), the slow smoothing constant base (30), and the threshold (70.0). RISK & SIZING holds the per-trade risk percent (0.5% of effective capital), the per-side max-open-trades (5), the daily loss limit percent (3.0%), the capital cap amount (0 = off, falling back to account equity), and the capital cap floor ($50). EXIT / MANAGE sets the ATR period (14), SL multiplier (2.0), TP multiplier (3.0), and trail multiplier (2.5). REGIME / CONFIRM holds the ADX period (14) and the HTF integer (8 → H1 by default). EXEC holds the spread cap (50 points) and the slippage tolerance (10 points).

Risk and exits are sized off ATR(14) and clamped through ClampMinStop() so SL and TP always respect the broker's stops-level floor plus the 2pt minimum-stop offset. The pre-trade tpDist/slDist ratio is checked against MIN_RR=1.5 — if the geometry cannot deliver at least 1:1.5, the entry is rejected. Position sizing is a clean effective-capital formula in CalcLots(): risk$ = effCap × 0.5%, perLotLoss = (slDist / tickSize) × tickValue, lots = risk$ / perLotLoss, normalised to the broker's volume step and clipped to the volume min/max, then verified against free margin via OrderCalcMargin() before any ticket is sent. ManageExits() runs every tick on each open position. The first thing it does is compute R-multiple from the recovered initial risk distance. If R ≥ 1.0, the stop is ratcheted to the open price (break-even, one-shot). If R ≥ 1.5, 50% of the position is partially closed via TryClosePartial_EX18026() (also one-shot). On every tick, a 2.5×ATR(14) forward-only trail is ratcheted up beyond the open. The trail only ever moves in the profitable direction. A 120-bar time exit (MAX_BARS_IN_TRADE=120) force-closes any position that has been held for 120 closed bars on the current timeframe. An opposite-signal exit closes the position if the aggregated signal crosses to the opposite threshold while the trade is open. A regime-change exit closes the position if the regime flips to Choppy while the trade is open. All three of those exit helpers retry up to three times on requote/timeout/price-off, sleeping 200ms between attempts.

The book-keeping layer is what makes the system safe in continuous use. UpdateRealized() runs every tick and walks the deal history from the start of the current day to compute g_todayRealized and from seven days back to compute g_weekRealized. UpdateLossStreak() walks the last 30 deals in chronological order and counts how many consecutive losses end the current streak, storing the bar time of the most recent loss in g_lastLossBar. NoTrade() checks those four numbers on every new bar: if daily loss ≤ -3% of effective capital, the EA goes quiet for the rest of the day; if weekly loss ≤ -9% of effective capital, the EA goes quiet for the rest of the week; if there was a recent loss and the elapsed bars since the last loss bar are fewer than 5 (COOLDOWN_BARS_AFTER_LOSS=5), the EA skips entries; if the current bar matches g_lastEntryBar, the EA refuses to double-enter on the same bar; if buy or sell count has reached InpMaxOpenTrades=5, the EA refuses additional same-side entries. The session window is hard-coded to 7–21 server time, and Saturday/Sunday is always blocked. Filling mode is auto-detected (FOK → IOC → RETURN). Trade retry on initial send is single-attempt on requote/timeout/price-off with a slippage tolerance of 10 points. The OnTester formula is (netProfit × profitFactor) / (1 + balanceMaxDD) with a 30-trade floor, so the strategy optimiser scores on net profit and profit factor while penalising drawdown.

In backtest the EA behaves like a slow-burn KAMA trend follower: it sits out Choppy and Range regimes almost entirely, lets the regime classifier wait for a clear Breakout or Expand, then lets the decayed aggregate cross the 70% threshold and the HTF EMA side confirm. Once in, the position is held through the 1R break-even ratchet, the 1.5R partial close, and a 2.5×ATR forward trail — so the typical trade either exits quickly on a stop out, gets partially banked at 1.5R, or runs to a 3×ATR TP and a continued trail. The 120-bar time exit and weekly 9% loss cap are the final safety net if the trend fails to materialise. The 22 functions cleanly separate regime, signal, sizing, entry, and management, so any failure surfaces as a tagged log line ([NOTRADE], [CONFIRM], [SIZING], [EXEC], [MANAGE]) rather than as a silent miss.