TRADEWINDS — Deep Analysis

Multi-account session forensics & engine recommendations
INTERNAL · v4.2.3 POWER ENGINE · 4 ACCOUNTS · ~150 TRADES

1 · Per-Account Results

Note: the accounts mirror the same signal stream, but they diverge because they were subscribed at different times, some ran legacy-engine trades before the Power switchover, and one account was intentionally paused mid-session (Lucid eval management). The cleanest read of the current engine is the account that ran pure Power end-to-end (Lucid 100).

AccountTradesWin%Gross/ctrAfter CostEngine
Lucid 1003655%+$268+$178100% Power
Big 15014648%+$334−$29mixed
Lucid Big 25251%+$102−$27mixed
Lucid Big 34850%+$68−$51mixed
Lucid 11643%−$42−$82legacy
The single most important row is Lucid 100: pure Power engine, 55% win, +$178/contract after commissions. That is the current bot, uncontaminated. Every account that ran mixed engines or higher trade counts fared worse after costs — which points the finger squarely at trade volume and legacy-engine contamination, not at the Power strategy itself.

2 · The Cost Problem — The #1 Enemy

The gross-vs-net gap is the story of this dataset. Big 150 traded 146 times and turned +$334/ctr gross into −$29/ctr after ~$360 of commissions. Lucid 100 traded 36 times and kept +$178 net. Same signals, wildly different outcomes — the difference is almost entirely trade count.

At ~$2.48/contract round-turn, every trade must clear ~1.25 points just to break even on costs. A 150-trade session carries ~$370/ctr of commission drag. The edge is real, but over-trading is taxing it away. Fewer, higher-quality trades is the highest-value improvement available — worth more than any stop-logic tweak.

3 · Loss Forensics — Where the Points Leak

Classifying every Power-engine loss by where it exited relative to its stop:

Loss typeCountCostWhat it means
Full-stop13−260 ptsTrade went wrong from entry, hit full stop. The dominant leak.
Tightened (give-back)9−67 ptsWent into profit, then reversed back through. The stepped ladder targets these.
Near-breakeven scratch11smallCut quickly and cheaply. Working as designed.
Key insight: the biggest dollar leak is full-stop losses (−260 pts), not give-backs (−67 pts). These are trades that were wrong immediately. That points to two fixes: (a) entry quality — stricter gates so fewer bad-from-the-start trades get taken; and (b) a hard per-trade dollar cap so the worst full-stops (especially multi-contract) can't run large. This aligns with the reported single ~$700 loss: almost certainly a multi-contract position on a wide stop hitting full-stop.

4 · Winner Analysis — Are We Keeping Them?

Power-engine win distribution: 6 wins over 20 pts (top five: 48, 34, 30, 28, 28), 15 wins of 10–20 pts, 14 wins under 10 pts. The big winners are being captured — the runner logic is working when a real move appears. The concern is the cluster of sub-10-point wins: trades cut before they matured. Some are correct (momentum genuinely faded); some are the trail being too tight and exiting on normal pullbacks.

Verdict on "keeping winners": the system already lands the big ones (that's the 1.46× win/loss ratio). The refinement is letting the mid-size winners breathe a little more — which is a trailing-distance calibration, best tested via the planned exit-model dropdown, not an assumption to hard-code now.

5 · The Hard-Loss / Thrash Check

Across all accounts, 30 exits carried the daily-hard-loss code, concentrated in the mixed/legacy accounts. On the pure-Power account (Lucid 100): zero hard-loss exits. This confirms the latched-lockout fix holds on the current engine — the thrash pattern lives only in the legacy-contaminated accounts and the pre-fix history. Not a current-engine concern, but a reason to get every account onto clean Power.

6 · Session Behavior

SessionPower TradesWin%Gross/ctrRead
Asia3650%+$74Steady, workmanlike
London1747%+$98Choppy but net positive
New York1560%+$23Best win rate, but flat $ — give-back signature

NY's high win rate but flat dollars is the clearest evidence for the give-back refinement: winning often, but the winners aren't being kept large enough. Asia and London carried the session on volume.

7 · Recommendations — Ranked by Expected Value

1. Reduce trade frequencyHIGHEST EV

The cost drag is the dominant leak. Raise the entry bar so the bot takes fewer, higher-conviction trades. Concretely: nudge Min Power 80→85, Power Gap 50→55, and Min ADX 20→23. Target: cut trade count 30–40% while keeping the best setups. Expected effect: turn the mixed accounts' after-cost losses positive. Test one variable at a time.

2. Hard per-trade dollar capHIGH EV

Full-stop losses are the biggest point leak, and a single ~$700 loss occurred. Add a max-loss-per-trade cap (candidate $400): before entry, if stop-distance × contracts would risk more than the cap, reduce contracts or skip. Also verify the broker disaster-stop is never placed wider than the cap. Protects the downside without touching the edge.

3. Get all accounts onto clean PowerMED EV

The performance spread across accounts is mostly engine contamination (legacy trades) and different start times. Once every account runs pure Power v4.2.3 with recreated alerts, the mixed-result noise disappears and the true edge shows on all four. Confirm each account's first trade reason code reads "Power."

4. Calibrate the runner trail (don't hard-code yet)MED EV

The mid-size winners suggest the trail may be slightly tight in some conditions. Rather than guess, build the exit-model dropdown (stepped / 80%-peak / structure / ATR) and A/B them on identical entries. Let data pick the trail, matched to A/Storm confidence (high-conviction trades get more room).

5. Keep what works — do NOT changeHOLD

The near-breakeven scratch cuts, the confidence-based sizing, the latched lockout, and the runner logic that catches the 48-pointers are all working. The 1.46× win/loss ratio is the crown jewel — protect it. Resist the urge to over-engineer a system that is already profitable.

8 · The Honest Bottom Line

You have a genuinely winning bot. The pure-Power account proves it: 55% win rate, 1.46× win/loss ratio, +$178/ctr net. The problems in the other accounts are not strategy problems — they're cost drag from over-trading and contamination from legacy trades. Both are fixable with settings, not redesign. The path from "profitable" to "very profitable" runs through fewer trades, a loss cap, and clean engine deployment — in that order. The edge is real; now protect it from itself.