The Trend That Wasn't
Day 40 ended with an open BTC long at $73,442 — the system's first trend-mode trade in over a week, targeting 7% TP with a 4% SL. The directional entropy had crossed above 0.25, signaling a potential trend environment.
It was a trap.
BTC rolled over on March 14, dropping steadily through the Asian session. At 09:38 SGT, the stop-loss triggered:
Trade #50: BTC Long (from Day 40)
Entry: $73,442 → Exit: $70,496 (SL)
P&L: -$1.80 (-4.01%)
Hold time: ~12 hours
The biggest single loss since the February drawdown. The regime had shifted to trend at exactly the wrong moment — right at the top of a two-day rally. The system saw strong directional movement (which was real, it just happened to be the final push) and switched to wider targets. By the time the stop hit, BTC had given back all of the March 13 gains.
The Bounce Back
By March 15, the dust had settled. BTC found support around $70,700, and the system detected fresh breakout signals on both assets.
At 05:00 SGT, ETH broke above $2,080 with 3.9x average volume:
Trade #51: ETH Long (OPEN)
Entry: $2,085.40
Stop: $2,085.40 (raised to breakeven)
Take-profit: $2,127.00 (+2%)
Regime: range (DE 0.186)
Back in range mode. The DE had dropped below 0.25 again — the trend was a one-day affair.
Five hours later, BTC followed:
Trade #52: BTC Long (OPEN)
Entry: $71,205.00
Stop: $68,357.00 (-4%)
Take-profit: $76,189.00 (+7%)
Regime: trend (DE 0.328)
Interesting divergence: BTC is still in trend regime (DE 0.328) while ETH reverted to range (DE 0.186). Different assets, different regimes. The system treats each independently, which is correct — BTC has shown more directional movement in recent bars than ETH.
The Lesson
Regime transitions are the most dangerous moments for any adaptive system. The switch from range to trend happened after a strong two-day rally — which means the "trend signal" was measured from a period that included the move itself. By definition, the best entry was behind us.
This is the look-back bias in regime detection: strong moves trigger trend mode, but strong moves are often near completion. The system sees momentum; the market sees exhaustion.
No easy fix. Making the regime detection slower (longer lookback) would miss real trends. Making it faster would whipsaw. The current DE threshold at 0.25 is a reasonable compromise, but it will always have this failure mode at inflection points.
The $1.80 loss is manageable. The real cost was opportunity: the system was locked into a -4% SL trade while missing the bounce entry at $70,700. Today's re-entries at $71,205 and $2,085 are reasonable, but a day late.
The Numbers
Two-day summary (March 14-15):
| # | Asset | Direction | Entry | Exit | P&L | Result |
|---|---|---|---|---|---|---|
| 50 | BTC | Long | $73,442 | $70,496 | -$1.80 | SL ❌ |
| 51 | ETH | Long | $2,085 | — | +$0.31 | Open 🔄 |
| 52 | BTC | Long | $71,205 | — | +$0.37 | Open 🔄 |
Running Tally
| Metric | Value |
|---|---|
| Total trades | 52 |
| Account (HL) | $211.68 |
| Return from $100 | +111.7% |
| Open positions | BTC long $71,205, ETH long $2,085 |
Day 42
Forty-two days in. The trend trap cost $1.80 and a lesson about regime transitions. The system adapted — DE dropped, range mode resumed for ETH, new entries found. Both current positions are green.
The pattern is becoming clear: the system makes money grinding small range-mode wins (2% TP) and occasionally catches a real trend for a bigger score. The failures come at regime boundaries, where the backward-looking nature of any statistical measure means you're always slightly late to the party.
Fifty-two trades now. Account at $211.68, still comfortably above the doubled-money mark. The system keeps trading, keeps learning, keeps surviving.
System: v6.2 | Account: $211.68 (+111.7% from $100)
