Words can hide things; pictures can't. Here is our research program in six charts — real CME MNQ data, ~1,900 sessions, realistic costs, all reproducible from the methods described across this series.
Thin, real, unglamorous: ~$15k over 7.3 years at minimum size. The shape matters more than the total — persistent grind, no hockey stick.
Worst excursion ~$4,600, and — honesty compels — the deepest one is the most recent. A drawdown chart is a strategy's true biography.
Two flat years, five good ones, a soft current year. A backtest where every year is green is a backtest you should question.
Total profit by profit-target. The 0.5R version wins 66% of trades and makes a third of the money. Everything from 1R up is statistically identical (we tested).
| 5 | 10 | 15 | 20 | 25 | 30 | 35 | 40 | 45 | 50 | 55 | 60 | |
| 0 | -0.21 | +0.28 | +0.58 | +0.77 | +0.78 | +0.67 | +0.74 | +0.71 | +0.54 | +0.70 | +0.79 | +0.78 |
| 0.05 | -0.21 | +0.28 | +0.58 | +0.77 | +0.78 | +0.67 | +0.74 | +0.71 | +0.54 | +0.70 | +0.79 | +0.78 |
| 0.1 | -0.19 | +0.25 | +0.59 | +0.77 | +0.78 | +0.67 | +0.74 | +0.71 | +0.54 | +0.70 | +0.79 | +0.78 |
| 0.15 | -0.08 | +0.27 | +0.56 | +0.80 | +0.80 | +0.67★ | +0.75 | +0.72 | +0.54 | +0.71 | +0.79 | +0.78 |
| 0.2 | -0.03 | +0.29 | +0.57 | +0.71 | +0.79 | +0.69 | +0.75 | +0.73 | +0.49 | +0.69 | +0.80 | +0.77 |
| 0.25 | -0.06 | +0.27 | +0.54 | +0.83 | +0.89 | +0.81 | +0.89 | +0.81 | +0.53 | +0.72 | +0.80 | +0.79 |
| 0.3 | -0.30 | +0.42 | +0.73 | +0.86 | +0.77 | +0.67 | +0.77 | +0.71 | +0.54 | +0.71 | +0.81 | +0.78 |
| 0.35 | -0.29 | +0.47 | +0.52 | +0.70 | +0.66 | +0.66 | +0.58 | +0.54 | +0.35 | +0.48 | +0.72 | +0.82 |
| 0.4 | -0.03 | +0.47 | +0.65 | +0.60 | +0.58 | +0.70 | +0.82 | +0.55 | +0.38 | +0.55 | +0.71 | +0.81 |
Sharpe across the parameter neighborhood (opening-range minutes → · volatility filter ↓). Bright plateau = many nearby settings work; ★ = the configuration we trade — inside the ridge, not on a lone spike.
Five full CPCV out-of-sample reconstructions of history (annualized Sharpe each). All positive, tightly clustered — honestly-chosen parameters kept working on data they never saw, 15 splits out of 15.
None of this proves the future. It proves we didn't fool ourselves about the past — the most a backtest can honestly claim, and more than most ever establish.