An earlier internal study found that MNQ’s 2024-2026 price move (16,300 → 29,900, +83%) made fixed-point TP/SL targets look like they were “improving” over time when they weren’t — the index was just getting bigger under a static number. That study covered one instrument over one 2.5-year window. This is the follow-up: 26 years of daily data across NQ, ES, and YM, run externally to check whether the finding generalizes or is an MNQ-specific artifact.
The core problem
Setting take-profit, stop-loss, breakeven, and trailing stops as fixed index points (a 45-point target, a 20-point stop) causes rapid parameter obsolescence as an index appreciates. The question is whether switching to % of price or a rolling ATR multiple actually fixes that, or just trades one problem for another.
Three findings
1. Point volatility scales linearly with price. Correlation between price level and ATR in raw points is strong and positive across all three instruments — NQ r = +0.888, ES r = +0.758, YM r = +0.705. When NQ re-rated 8.35x since 2000 (3,482 → 29,094), its point-ATR expanded 25.6x (21.4 → 547.4 points). Point range tracks price almost directly.
2. Expressed as % of price, volatility is stationary — no secular drift. Correlation between price level and ATR-as-%-of-price is close to zero (NQ r = -0.139, ES r = -0.158, YM r = -0.139), and a linear regression of %ATR against time finds essentially no secular trend (R² = 0.0115 for NQ, 0.0193 for ES, 0.0082 for YM — all near zero). NQ’s median daily ATR sits at 1.69% of price with a stable 1.33%-2.47% interquartile range across the full 26 years.
3. Volatility still moves in real regime cycles — %/ATR normalization doesn’t erase that, only the price-level problem. NQ’s ATR-% ranged from 0.95% (2017, “Historical Vol Suppression”) up to 6.21% (2000, dot-com bust), with COVID (2020), the 2022 rate-hike selloff, and 2008’s GFC all showing 2-4x expansions over quiet-year baselines. A %-of-price stop absorbs price re-rating but is still blind to these regime shifts — it doesn’t widen during a genuine volatility spike or tighten during a quiet one. An ATR multiple absorbs both.
What a fixed 45/20-point bracket actually became, on NQ, at four points in time
| Period | NQ price | Avg daily ATR | 45pt TP as % of ATR | 20pt SL as % of ATR | Behavioral identity |
|---|---|---|---|---|---|
| 2004-2006 | ~1,500 | 23.8 pts | 189.1% | 84.0% | Multi-day swing trade |
| 2014-2016 | ~4,200 | 61.2 pts | 73.5% | 32.7% | Multi-hour intraday trend trade |
| 2020 (COVID) | ~10,300 | 268.6 pts | 16.8% | 7.4% | 1-5 minute scalp |
| 2026 | ~27,500 | 547.4 pts | 8.2% | 3.7% | Sub-minute microstructure noise |
Without touching a single line of strategy code, the same 20-point stop went from absorbing 84% of a day’s average range in 2005 to 3.7% of a day’s range in 2026 — silently mutating from a swing-trade stop into something that gets clipped by noise.
Where the three approaches actually differ
| Dimension | Fixed points | % of price | Rolling ATR multiple |
|---|---|---|---|
| Long-term price re-rating | Fails — points shrink relative to price noise | Survives — scales linearly with price | Survives — ATR itself scales with price (r ≈ 0.7-0.9) |
| Cyclical volatility regime shifts | Fails — too tight in crises, too loose in quiet markets | Fails — static % doesn’t widen when daily range doubles | Survives — multiplier widens in crisis regimes, tightens in quiet ones |
| Parameter shelf-life | Weeks to months | Multi-year, still vulnerable to macro vol | Decades — a dimensionless, stationary parameter |
Supporting literature
This isn’t a novel claim in the broader quant literature:
- Kaminski & Lo (2014), Journal of Financial Markets — fixed stopping thresholds underperform because they create asymmetric absorbing boundaries regardless of actual conditional variance; volatility-adjusted thresholds preserve expected return while controlling tail drawdowns.
- Harvey, Hoyle, Russell et al. (2018/2020), AQR/Man Group — futures strategies scaling position size and stop distance by rolling volatility eliminate negative return skew and lift Sharpe ratios by +0.20 to +0.35 across 90+ years of futures history.
- Robert Carver, Systematic Trading — defines “volatility drag”: point-based brackets suffer exponential parameter decay;
Stop Distance = k × Daily Instrument Volatility(k ∈ [0.5, 3.0]) is presented as the only mathematically stationary specification. - Perry Kaufman (2019), Trading Systems and Methods — a comparative backtest across 500+ futures contracts over 30 years found fixed-point parameters had an average profitable shelf life under 14 months, while ATR-normalized multipliers retained positive expectancy over the full 30 years without retuning.
Practical formulation
TakeProfit = EntryPrice ± (k_TP × ATR_14)
StopLoss = EntryPrice ∓ (k_SL × ATR_14)
BracketPts = max(k × ATR_14, FloorPts)The floor matters in practice — an ATR multiple with no minimum can produce a degenerate sub-10-point bracket during illiquid overnight hours.
Answering the original question directly
Does a %-of-price or ATR-multiple coefficient, once set, hold up structurally across long-term index re-rating, or is periodic retuning still required?
Both survive the price-re-rating problem that breaks fixed points. Only ATR multiples also survive cyclical volatility regime shifts — %-of-price is price-level-stationary but still needs retuning across a real vol regime change (e.g., transitioning into or out of a 2022-style rate-hike selloff), while a rolling ATR multiple absorbs both automatically. This directly informs Tokyo Drift’s and Drift VWAP’s TP/SL Mode selector: ATR Multiple is the structurally soundest default of the three modes for anything meant to keep working years out, not just Points.