๐ŸŽฏ Exit Study โ€” how should we close a trend trade?

Same trades, same entries โ€” we only change how we EXIT, then measure the result on fresh data it never saw. Trend plays only (revbreak_long, trend_ride, open_drive, band_break), 2,114 trades.
How to read this: the money column is dollars of profit per 100 trades, each risking $100, measured on FRESH data (data the test never saw โ€” the honest number for real trading). "The band" = a fixed profit target a set distance from the day's average-price line. "The EMA / trend line" = a line that follows price; we exit when price closes back through it.
๐Ÿ’ก For a trend trade, RIDE THE TREND. Don't grab a small fixed target early โ€” that's where almost all the profit was being left behind.

The comparison (fresh-data profit per 100 trades)

How we exit$ / 100 (fresh)win %what it means
Ride the trend to the EMA break (100%)$+1,47142%Hold the whole position; exit only when price closes back through the 20-minute trend line.
Take 1/3 at the band, ride 2/3$+87946%Bank a little early, ride most. Keeps most of the profit and lifts the win rate a bit.
Take 1/2 at the band, ride 1/2$+70746%More banked early = less total profit.
Take 2/3 at the band, ride 1/3$+53446%Most banked early = most profit left behind.
Current ladder (60% / 20% / 20%)$+16940%Our old mean-reversion exit โ€” wrong tool for a trend play.
Cap 100% at the band (fixed target)$+20044%Grab a small fixed target and quit. Almost no money on a trend play.
Every dollar you take off early at the band is a dollar not riding the trend โ€” so profit falls the more you bank early. The only thing a partial buys you is a slightly higher win rate and a smoother ride (you lock a small win on more trades), at a real cost in total profit.

The swing idea (hold for days, exit at the 5/10/20-day lines) โ€” not reliable yet

Looks amazing, is actually a lottery. Held for days and scaled out at the daily 5/10/20-day averages, the average came out huge โ€” but that's a mirage: the typical (median) trade LOSES (-0.86R), only 34% win, and 67% stop out in the first day or two (the intraday stop is far too tight to hold overnight). The giant average is carried by a handful of rare multi-week runners. With a sensible swing stop it even goes negative on fresh data ($-2,271 per 100). Verdict: intriguing, but a proper swing test needs real swing entries, wider stops, position-sizing, and more history before any number is trustworthy. Don't trade it on these results.
Recommendation.
โ€ข Ride the trend to the EMA break (100%) โ€” the clear winner ($+1,471/100 on fresh data), and it holds up out-of-sample.
โ€ข If you want a smoother, higher-win-rate version that's easier to hold, take a small 1/3 off at the band and ride the rest โ€” you keep most of the profit ($+879/100).
โ€ข Never cap the whole thing at the band on a trend play โ€” it throws away ~90% of the edge.
โ€ข Match the exit to the play: this is for TREND plays. Bounce-back (mean-reversion) plays should still take the fixed target โ€” there's no trend to ride.
โ€ข The swing exit is parked until we can test it properly.
โ† Experiment Log ยท 90-Day Backtest ยท Hub