Pipsgrowth EX18068 TrendFollow

Pipsgrowth EX18068 T3Trend is a single-position trend follower that builds its signal from a Tillson T3 moving average computed entirely in software. Rather than loading a custom indicator or calling iCustom, the EA spins up six identical-period EMA handles in OnInit and runs the remaining smoothing passes inside CalcT3Series() / CascadeEMA(). The T3 itself is the standard Tillson construction: six nested EMAs of close, blended with the volume factor v (InpT3Coeff, default 0.7) using the four Tillson coefficients c1 = -v^3, c2 = 3v^2 + 3v^3, c3 = -6v^2 - 3v - 3v^3, c4 = 1 + 3v + v^3 + 3v^2. All six EMAs share the same period input (InpT3Period, default 14), so the user-facing knobs are just the period and the volume factor. The cascade uses a 200-bar working window (T3_LOOKUP=200), which is well past the warm-up required for the sixth EMA to stabilize. Because the EA reads iMA handles for the first pass and then does the rest in code, the chart and the strategy tester produce identical T3 values, and there is no .mqh or .ex5 to distribute alongside the .mq5.

The signal is the slope of the T3, not its value. T3Slope() runs an ordinary-least-squares regression over the last three T3 points (T3_SLOPE_BARS=3) and returns the regression coefficient. Signal() combines that slope with two additional checks on the closed bar (shift=1): a long is armed only when close is above the current T3, the prior T3 is below the current one (i.e. the T3 is rising bar-on-bar), and the regression slope is positive. Shorts mirror the same three conditions with the signs flipped. The monotonicity test is what separates EX18068 from a raw slope-sign-flip system: the EA is not satisfied with a momentary uptick, it wants the T3 itself to be climbing. Confidence is reported but not gated on — it is the magnitude of the slope, normalized by the current ATR(14) and capped at 100, returned alongside the direction so external dashboards and the strategy tester can rank trades, but a low confidence score does not block a valid directional match.

A six-state regime classifier runs on every tick before any entry logic is touched. CalcRegime() reads ADX(14), ATR(14) over a 100-bar window, and Bollinger Bands(20, 2.0), and ranks the live ATR against its own 100-bar distribution and the live band-width against its own 100-bar distribution via Percentile(). The state map is: ADX <= 15 forces REG_CHOPPY, ADX >= 25 with either percentile >= 80 is REG_BREAKOUT, ADX >= 25 alone is REG_STRONG_TREND, ADX >= 20 is REG_WEAK_TREND, ATR-percentile <= 20 is REG_COMPRESS, the same percentile or band-width test with lower ADX is REG_EXPAND, and the residue is REG_RANGE. CHOPPY is the only state that triggers a hard no-trade veto through the rest of the pipeline. The other five are diagnostic — they get stamped on each tick for the journal and for the OnTester pass, but the T3 slope is the only directional input that opens trades.

Confirm() stacks four gates on top of an allowed regime. With InpRequireHTF=true (default), the last closed H1 candle must be on the same side of an H1 EMA(21) (hHTF_EMA, period HTF_T3_PERIOD=21) as the proposed direction. With InpRequireSession=true, the broker-server hour must be inside 7-21 (SESSION_START_HOUR=7, SESSION_END_HOUR=21), a 14-hour window. The spread gate compares the current spread to a rolling 50-bar EMA of spreads (g_spreadAvg) maintained inside OnTick with alpha = 0.05, and rejects entries when the live spread is more than 3.0x that average (MAX_SPREAD_MULT=3.0). Finally, the reward-to-risk ratio is computed from SL_ATR_MULT * ATR(14) for the stop and TP_ATR_MULT * ATR(14) for the target, and the trade is refused if the resulting ratio falls below InpMinRR=1.5. With the hard-coded 2.0 / 4.0 ratio (SL = 2.0 x ATR, TP = 4.0 x ATR, R:R = 2.0), the InpMinRR check is effectively a sanity gate rather than a real constraint.

The capital allocation cap is the EA's most distinctive design choice and the lever most worth understanding before going live. EffCapital() returns min(InpCapAmount, account equity) minus today's realized P&L, but only when InpCapAmount is greater than zero. The default value of 0.0 disables the cap entirely, so out of the box the EA sizes off the full account equity. The cap turns the system from "size off whatever is in the account" into "treat the account as a $X sandbox" — set InpCapAmount=1000 and the EA will never risk more than it thinks it has 1000 dollars of, regardless of how much the account actually holds. InpCapFloor (default $50) is a separate kill: when EffCapital() drops below that level, NoTrade() returns true with reason "capital_below_floor" and the EA stops opening new positions, which is how the EA defends against a runaway losing streak on a capped account. All daily-loss and weekly-loss math is also taken from EffCapital(): a 3% daily loss limit (InpDailyLossPct=3.0) and a 9% weekly loss limit (InpWeeklyLossPct=9.0) halt entries the moment those thresholds are breached in closed-trade P&L.

Sizing rides on the same capital figure. CalcLot() takes EffCapital() * min(InpRiskPercent=0.5%, KELLY_CAP_PCT=5.0%) as the dollar-at-risk budget, then divides by the per-lot risk computed from slDistPrice * (TickValue / TickSize). The resulting lot is rounded down to the broker's LotsStep, floored at LotsMin, capped at LotsMax, and refused outright if OrderCalcMargin says free margin cannot cover the resulting position. InpMaxConcurrent=1 and InpMaxSymbolPos=1 are both hard-coded to 1, so the EA is strictly a one-trade-at-a-time, one-symbol-at-a-time system — running a second chart of XAUUSD with the same magic will be rejected by CountSymbolPositions().

ManageExits() runs on every tick before the entry pipeline, so the exits are always up to date before the next signal is even considered. When unrealized profit reaches 1R (the initial SL distance), the stop is ratcheted to entry + 50 points for longs and entry - 50 points for shorts (BE_OFFSET_POINTS=50), and 50% of the position is closed at the same trigger (PARTIAL_AT_R=1.0, PARTIAL_RATIO=0.5), provided the residual volume is at least 2 * LotsStep. The trailing stop runs continuously: at every tick the EA recomputes the trail at 2.5x ATR(14) (TRAIL_ATR_MULT=2.5) from the current bid/ask, and only updates the stop if the new value is at least 30 points (TRAIL_STEP_POINTS=30) inside the previous one, which avoids spamming the broker with re-modify calls in low-volatility drift. An opposite T3 slope signal closes the position outright via the Signal() function, and a 240-bar time stop (MAX_BARS_IN_TRADE=240) closes anything still open at roughly 20 hours on M5 or 10 days on H1. Order execution goes through three-attempt retry helpers — TryClose_EX18068, TryClosePartial_EX18068, TryModify_EX18068 — that back off 200ms on close-style failures (requote, timeout, price-off, price-changed) and 100ms on modify-style failures (requote, timeout only), which is the right cadence for a busy ECN gold book.

InpDryRun defaults to true, which means the EA logs every intended BUY / SELL / modify / close to the journal as DRYRUN lines but does not place orders until you flip that input to false. InpKillSwitch (default false) is the second safety input: when toggled true, both new entries and most management actions stop. OnTester() ranks optimization passes by (net profit * profit_factor) / (1 + balance_dd_percent / 100) and returns 0 for any test with fewer than 30 trades, so the optimizer only grades runs that have a statistically meaningful sample size. EX18068 is built for traders who want a single, methodical T3 trend position on gold during the active part of the day, sized as a 0.5% slice of effective capital, with the capital cap as the lever that turns a gold-trend EA into a sandbox-sized, equity-isolated test bed.