Model
Starting PER_POS: $1,000 (same for all 4 scenarios).
Daily P&L: backtest base P&L × (current PER_POS / $1K). Scales linearly with position size.
Reinvest rule: PER_POS[t+1] = PER_POS[t] × (1 + L × daily_rate[t]).
L is the leverage multiplier on growth speed.
Interpretation: L=1 reinvests profits as they come (vanilla compound).
L=4 borrows against profits to grow PER_POS 4× faster — losses also amplified 4×.
Account BP: not tracked — assumed infinite (per user spec).
Result: exponential divergence between curves. Day 240 PER_POS:
L=1 ≈ $1.8K · L=2 ≈ $3.3K · L=3 ≈ $6K · L=4 ≈ $11K.
Cumulative profit roughly tracks PER_POS curve area.
Key numbers by leverage
| Scenario |
Starting PER_POS |
PER_POS at day 120 |
PER_POS at day 240 (median) |
Cum profit day 120 |
Cum profit day 240 (median) |
Day 240 5/95 band |
MaxDD |
Worst single day |
Sharpe |
| 1x compound · vanilla | $1,000 | $1,084 | $1,175 (+17%) | $+2,654 | $+5,475 | $+2,746 – $+9,012 | $-1,423 | $-553 | 2.45 |
| 2x compound · moderate | $1,000 | $1,160 | $1,317 (+32%) | $+2,618 | $+5,554 | $+2,281 – $+9,710 | $-1,424 | $-559 | 2.39 |
| 3x compound · aggressive | $1,000 | $1,226 | $1,483 (+48%) | $+2,558 | $+5,732 | $+2,143 – $+11,174 | $-1,422 | $-564 | 2.33 |
| 4x compound · very aggressive | $1,000 | $1,278 | $1,669 (+67%) | $+2,474 | $+5,955 | $+2,390 – $+12,825 | $-1,419 | $-568 | 2.26 |
How to read: all scenarios start at $1K per position.
L=1 is plain compound — profits get rolled forward at the natural rate (≈0.25%/day).
L=2/3/4 amplify the compound rate by borrowing against accumulated profit, so PER_POS grows 2-4× faster.
On bad days that compounding works backwards — at L=4 a -2% backtest day becomes a -8% drop in PER_POS.
Sharpe is identical across leverages in this model — both edge and risk scale together,
but geometric (compound) return suffers a small volatility-decay drag at higher L.
Capacity ceiling at high L: at L=4 after 6 months, PER_POS reaches ~$11K per position
× 13 positions = $143K per-day notional in <$10 PM-only names.
Those tickers have PM volume 50-500K shares — your $11K positions = 1,000-2,000 shares = 0.2-2% of PM tape.
Real slippage at that size will eat ~10-30% of the modelled edge. Model is upper bound.
Look-ahead caveat: backtest uses 5-min bars labelled by start time, so the
`09:10` bar actually closes at 09:14:59 — backtest sees ~5 minutes of future at decision time.
Real live Sharpe is lower than backtest. Treat all numbers above as upper bounds.