week 1-2 march testing.
adaptive strategy live.
mixed results.
testing approach #
week 1 (mar 4-8):
paper trading only.
logging all signals.
no real money.
week 2 (mar 11-15 so far):
live trading tiny size.
$100 per trade max.
testing regime detection.
paper trading results (week 1) #
trades: 12
wins: 8
losses: 4
win rate: 67%
paper pnl: +$2,840 (simulated on full size)
looked promising.
live testing results (week 2 so far) #
monday-tuesday (mar 11-12):
trades: 4
wins: 2
losses: 2
win rate: 50%
real pnl: +$18 (on $100 size)
issues found:
- regime whipsaw on monday
- false signal tuesday morning
- slippage worse than expected
- fill delays on entries
reality different from paper trading.
regime detection issues #
monday mar 11:
VIX opened 17.8 (medium vol).
by 11am: jumped to 20.4 (high vol).
regime switched mid-day.
my confirmation logic requires 3 days.
stuck in wrong regime for 6 hours.
lost 2 trades because parameters didn’t match conditions.
need faster regime detection for intraday.
parameter adjustment #
original confirmation: 3 days
updated: 3 hours for intraday regime shifts
tested tuesday mar 12:
VIX moved 18.2 → 19.8 → 18.5 in 4 hours.
3-hour confirmation prevented false regime change.
worked better.
comparing to february static #
february (static params):
mean reversion: 57% win rate
total trades: 37
net: +$1,400
march week 2 (adaptive, tiny size):
mean reversion: 50% win rate (so far)
total trades: 4
net: +$18
too early to judge.
need more trades.
slippage reality check #
paper trading: assumed 1 tick slippage
live trading: getting 2-3 tick slippage
on $100 trades: not material
on full size ($1,500): would be $40-60 per trade
annual impact if continues:
assuming 300 trades/year.
extra 2 ticks = $12,000-18,000 drag.
significant.
need to add slippage buffer to entry/exit logic.
account status #
mar 1: $394,800
mar 12: $394,818
basically flat.
testing phase = capital preservation.
good.
therapy tuesday (3/12) #
dr. r: “testing new strategy. how’s it going?”
me: “mixed. paper trading looked great. live trading showing issues.”
dr. r: “what kind of issues?”
me: “regime detection too slow. slippage worse than expected. reality different from simulation.”
dr. r: “frustrating?”
me: “yeah but expected. testing phase is for finding this shit.”
dr. r: “how’s different from february?”
me: “february i was trading full size when strategies failed. now testing tiny size before committing.”
dr. r: “learning from experience.”
exactly.
living with A. - coding help #
sunday night (mar 10):
stuck on regime detection bug.
couldn’t figure out why confirmation logic failing.
A.: “what’s wrong?”
me: “regime filter isn’t working. keeps false-triggering.”
A.: “show me the code.”
spent 30 minutes reviewing my logic.
A.: “you’re checking VIX once at market open. need to check every bar.”
me: “…fuck. you’re right.”
A.: “also your 3-day confirmation is calendar days not trading days.”
me: “…double fuck.”
fixed both bugs in 10 minutes.
this is why living together works.
she understands code.
can actually help with trading problems.
what i’m learning #
1. paper trading optimistic
no slippage reality.
no fill delays.
perfect executions.
real trading has friction.
2. regime detection hard
intraday volatility swings.
false signals.
confirmation vs responsiveness tradeoff.
3. testing phase essential
tiny size = cheap lessons.
finding bugs before committing capital.
worth the time.
next steps #
rest of week 2 (mar 13-15):
continue $100 size.
focus on regime detection tuning.
log all slippage.
week 3 (mar 18-22):
if win rate >60%: increase to $200 size.
if win rate <55%: back to paper trading.
week 4 (mar 25-29):
if still working: $400 size.
if broken: pause and reassess.
no rush to full size.
monthly projection #
march target: flat to +1% ($0 to $4k)
currently: +$18 (basically flat)
fine.
testing month = not about profit.
about validating strategy before committing.
comparing to past disasters #
march 2023:
tested strategy for 2 weeks on paper.
looked great.
went live full size immediately.
lost $40k in 3 weeks.
march 2024:
testing strategy for 4+ weeks.
starting tiny size.
finding issues before they cost real money.
growth.
tonight #
week 1-2 march testing.
paper trading: 67% win rate.
live trading: 50% win rate.
reality different from simulation.
finding bugs before they matter.
testing phase working as designed.
3:15am wednesday. adaptive strategy testing. week 1 paper trading looked good (67% wr). week 2 live at $100 size showing reality (50% wr). regime detection too slow, slippage worse than expected. A. helped fix bugs in confirmation logic. account flat at $394,818. testing phase = capital preservation.
-AK