POWER ENGINE · MNQ FUTURES · JULY 9, 2026 · AN ECLAWMERCE PRODUCT
Part 1 · What We Changed & Why
v4.2.5 is a data-driven rebuild based on forensic analysis of ~150 trades across 4 accounts. Every change below traces directly to a finding in the trade data — not intuition, not preference. The engine's core thesis (trade power imbalance, ride momentum, protect capital) is unchanged; what changed is how aggressively it filters for real momentum vs. fakes, and how it sizes risk.
The root cause this version addresses
Analysis of the prior session (v4.2.4) revealed a specific, measurable problem: 30 of 45 trades held 2 minutes or less, and those short-hold trades went 13W/17L (43% win), losing −79 points net. Meanwhile, the 15 trades that survived beyond 2 minutes went 11W/4L (73% win) and were strongly profitable. The engine was entering on single-bar momentum spikes that looked real but died immediately. The average loser (−12.6 pts) was bigger than the average winner (+9.7 pts) — the win/loss size ratio had inverted to 0.77×, when a profitable system needs it above 1.0×.
In plain terms: the bot was firing too fast on signals that hadn't proven themselves. The entries that survived two bars were the real ones. v4.2.5 is built around that single finding.
1 · 2-Bar Consecutive ConfirmationEntry Quality
The power gates (pressure, ADX, grade, bar quality) must now remain true for 2 consecutive bars before the engine enters. If bar 1 fires the signal but bar 2 doesn't confirm, the entry cancels. This directly filters the single-bar spikes that produced 17 of the session's 21 losses.
DATA: trades surviving 2+ bars won 73%. Trades dying within 2 bars won 43%. The 2-bar gate targets this exact split.
2 · Rising ADX RequiredEntry Quality
ADX must now be higher than the prior bar (trend strengthening, not fading). Previously, ADX only needed to be above a floor (23), so a falling ADX — a trend exhausting — still passed. Now the trend must be actively building. This filters late-cycle entries where momentum technically exists but is weakening.
DATA: several losing entries fired at ADX levels that were above threshold but declining from recent highs — the tail end of a move, not the beginning.
3 · Stronger Confirming BarEntry Quality
The bar that triggers entry must now have a real body (≥ 0.35 ATR) and close in its strong zone (top 40% for longs, bottom 40% for shorts). Previously, a tiny doji that technically closed green could pass. Now the bar must show genuine commitment in the trade's direction.
4 · Position Sizing: 3 / 5 / 7 with $400 CapRisk
Contract sizing raised to B = 3, A = 5, A/Storm = 7 for meaningful risk on confirmed setups. Governed by a hard $400 per-trade maximum loss cap: before entry, if stop × contracts × $2 exceeds $400, contracts are automatically reduced. An earlier $5,000 drawdown was traced entirely to a TradersPost signal multiplier set at 7× instead of 1× — amplifying every trade beyond the code's risk controls. All accounts are now confirmed at 1×; sizing is controlled exclusively inside the Pine script where the cap governs it.
DATA: at intended 1-3 contract sizing, the same session that lost $5,000 would have been approximately −$96 — a normal, survivable fluctuation.
5 · C-Grade Trades EliminatedSelectivity
Minimum grade raised to B only; C removed from the options entirely. C-grade trades were the only negative-expectancy bucket in the dataset.
DATA: C-grades won 30% and lost ~$7.83/contract each after costs, while A and B trades were net positive. Removing this bucket is worth ~+$75/ctr/session.
6 · Early-Out Loosened to 5 BarsManagement
The early-exit window widened from 3 to 5 bars. With the stricter entry gate (2-bar confirmation + quality filters), the trades that get in deserve more room to develop. The entry filter now does the quality work; the early-out is the safety net.
7 · FVG Mapping on ChartVisual
Fair Value Gaps drawn as boxes on the chart — teal (bullish) and coral (bearish), auto-removed when filled. Display-only; zero effect on trading logic. Carried forward from v4.2.4.
What did NOT change
The core Power Engine thesis — trade on power imbalance, skip balanced/choppy conditions.
The stepped stop ladder — breakeven at +9 pts, lock half at +18, stepped trail at ~7 pts behind best.
Runner logic — big winners are uncapped. The +48-point, +34-point, +33-point winners were not affected.
Pullback-gated re-entry — no chasing back in at the exit extreme.
All safety systems — latched lockout, blackouts, pre-halt, broker disaster stops.
Part 2 · Recent Trade Analysis & Summary
Analysis based on the most recent session export (91 fills, 45 reconstructed trades, single account), covering Asia through early NY on July 8–9, 2026. All figures below are stated at intended sizing (dividing the 7× multiplier out) to show the strategy's true performance.
45
Trades
53%
Win Rate
0.77×
Win/Loss Ratio
−$96
After Cost
The win rate is sound (53%) but the win/loss size ratio inverted (0.77×). Average winner: +9.7 pts. Average loser: −12.6 pts. The losers are bigger than the winners, which turns a healthy win rate into a breakeven-or-worse result. The prior session's ratio was 1.46×. This inversion is directly caused by the single-bar spike entries dying immediately for full losses while winners are cut modestly. v4.2.5's consecutive-bar gate targets this specific problem.
By Session
Session
Trades
Win%
Gross/ctr
After Cost
Read
Asia
32
53%
+$102
−$5
High volume, marginal after costs
London
7
57%
+$40
+$22
Cleanest session — fewer trades, better quality
New York
6
50%
−$92
−$114
Choppy, give-back signature
London was the model session — 7 trades, 57% win, net green after costs. That's what the whole engine should look like: fewer, higher-quality entries. Asia took 32 trades and churned its edge away on costs. v4.2.5's filters are designed to make every session look more like London.
The Critical Finding: Hold Time Predicts Outcome
Hold Time
Trades
W/L
Win%
Net Points
≤ 2 minutes
30
13W / 17L
43%
−79.3 pts
> 2 minutes
15
11W / 4L
73%
profitable
This is the table that drove v4.2.5. The trades that survive the first two minutes are overwhelmingly profitable (73% win). The ones that die immediately are the drag (43%, net negative). The 2-bar consecutive confirmation is designed to stop the engine from entering on signals that won't survive — by requiring the signal to prove it's still there one bar later before committing capital.
Selectivity Progress
Version
Trades/Session
Reduction
v4.2.3 (legacy engine)
~83
—
v4.2.4 (selectivity gates)
45
−45%
v4.2.5 (2-bar + quality, projected)
~15–25
~−50% further
Each version has progressively filtered for higher-quality entries. v4.2.5 is expected to cut volume roughly in half again from v4.2.4, but with materially higher win rates and — critically — a win/loss size ratio above 1.0× again.
Grade Performance (cumulative data, ~150 trades)
Grade
Trades
Win%
After-Cost Expectancy
Action
A
33
42%
+$0.28/ctr/trade
Keep
B
28
53%
+$0.25/ctr/trade
Keep
C
10
30%
−$7.83/ctr/trade
Eliminated in v4.2.5
Power (tagged)
68
51%
+$0.39/ctr/trade
Core engine
Best & Worst Trades
Time (UTC)
Side
Points
$ at intended size
Held
Best
23:34
LONG
+31.25
+$188
8 min
Worst
02:36
SHORT
−25.25
−$101
3 min
At intended sizing, the worst single trade is −$101 — well within the $400 cap and entirely survivable. The best single trade was +$188 on a clean 8-minute momentum run. The worst loss streak was 3 consecutive losses totaling −$96.
The Multiplier Lesson
A critical operational finding: TradersPost signal multipliers on all 4 accounts were set to 7× instead of 1×. This amplified every trade — entries, exits, wins, and losses — by 7×, completely bypassing the in-code $400 loss cap and confidence-based sizing. A session that was −$96 at intended sizing became approximately −$5,000 at 7×. The multiplier has been corrected to 1× on all accounts. All sizing is now controlled exclusively inside the Pine script, where the $400 cap governs it. The TradersPost multiplier must remain at 1× permanently.
Outlook & Validation Plan
What We Expect from v4.2.5
Fewer trades, higher quality: projected 15–25 trades per full session (down from 45).
Win/loss size ratio back above 1.0×: the fake-spike entries that inverted the ratio should be filtered by the 2-bar gate.
Net positive after costs: fewer trades = lower commission drag; higher-quality entries = better per-trade expectancy.
The $400 cap actually working: at 3/5/7 sizing with 1× multiplier, the cap governs the high-contract / wide-stop combinations.
This version runs untouched for 24 hours regardless of results. The validation discipline is absolute: the parameters are frozen, the prediction is written, and the tape writes the verdict. If the data confirms, v4.2.5 becomes the operational baseline. If it doesn't, we read what the data says and adjust one variable. No mid-session tuning. No shotgunning.
The Honest Bottom Line
The strategy is sound. Across all sessions and versions, the Power Engine finds real momentum, catches runners (the +48, +33, +31-point winners prove it), and contains losses when the safety systems are correctly configured. The problems this week were operational (a 7× multiplier), not strategic. v4.2.5 adds the entry-quality layer the data demanded — the consecutive-bar confirmation — and sizes risk meaningfully on the setups that earn it. The edge is real. Now we prove it cleanly.
