Back to Research
Strategy Research
New

Adaptive Grid v2.1: Does Volatility-Adjusted Spacing Work?

Testing ATR-Based Grid Spacing Against the Proven Static 2.59% Baseline

Published: February 23, 2026 | 13 Cryptocurrencies | 3 Rolling Windows | 3 Spacing Variants

The Hypothesis

Our rolling backtest study found that grid range coverage is the #1 driver of returns — assets that stayed within their grid range performed dramatically better. The v2.1 hypothesis:

"If we adapt grid spacing to each crypto's actual volatility (ATR), we can keep more assets in range and improve risk-adjusted returns."

We tested this by running three spacing variants through the same blind forward test:

Three Variants Tested

v1

Static 2.59%

Baseline (Current)

Fixed spacing for all assets. The proven production configuration used in CoinRoc today.

v2

Full Adaptive

Φ × ATR, 1.8% floor

ATR-based spacing per asset. Can go tighter or wider than 2.59%. Low-ATR assets (BTC, ETH) get tighter grids; high-ATR assets (DOGE, SOL) get wider.

v2b

Wider-Only Adaptive

Φ × ATR, 2.59% floor

ATR-based spacing but never tighter than the proven 2.59%. Only widens for high-volatility assets. Preserves the baseline as a safety net.

  Adaptive Spacing Formula:

  Full Adaptive:    Spacing = clamp(Φ_grade × ATR%, 1.80%, 6.50%)
  Wider-Only:       Spacing = clamp(Φ_grade × ATR%, 2.59%, 6.50%)

  Φ by Grade: A=0.85 | B=0.95 | C=1.05 | D=1.20
  ATR calculated from Year 1 data only (blind forward test)

Portfolio Results (Calmar CDaR)

Portfolio-level returns using Calmar CDaR optimization across all three windows:

W1 (Bull)
Static
77.2%
Adaptive
70.9%
Wider-Only
75.1%
W2 (Mixed)
Static
45.8%
Adaptive
39.8%
Wider-Only
46.1%
W3 (Bear)
Static
10.8%
Adaptive
8.3%
Wider-Only
9.7%

Result: Static 2.59% won in W1 and W3. Wider-Only was closest to static across all windows and slightly ahead in W2. Full Adaptive underperformed consistently due to tighter spacing on low-ATR assets.

The Critical Finding

Tightening Below 2.59% Hurts Performance

For low-ATR assets (BTC, ETH, BNB, LTC), the Full Adaptive variant tightened spacing to 1.80%. This consistently produced worse returns than the static 2.59%:

AssetATR%StaticAdaptive (1.8%)Wider-Only (2.59%)
BTC1.29%4.7%3.7%4.7%
ETH1.59%20.2%18.6%20.2%
BNB1.68%16.7%9.6%16.7%
LTC1.66%184.6%196.9%184.6%

Exception: LTC — the one low-ATR asset that benefited from tighter spacing in W1 (196.9% vs 184.6%). But this didn't repeat in W2 or W3, suggesting it was window-specific rather than systematic.

Why the Wider-Only Variant Is Most Promising

The Wider-Only variant preserves the proven 2.59% baseline and only widens for volatile assets. This means:

Low-ATR assets unchanged

BTC, ETH, BNB, LTC stay at 2.59% — proven and safe

High-ATR assets get wider grids

DOGE, ADA, POL, AVAX get more room to breathe

No downside risk

The worst case is identical to static (2.59% floor)

W2 portfolio outperformed static

46.1% vs 45.8% — small but in the right direction

Per-Window Asset Results (W1 — Bull Market)

AssetATR%SpacingStaticAdaptiveWider-Only
LTC1.66%1.80%184.6%196.9%184.6%
XRP2.01%2.11%98.8%82.0%98.8%
ADA2.10%2.52%70.0%84.5%70.0%
LINK2.43%2.92%57.6%63.3%63.3%
UNI2.73%3.27%37.5%38.6%38.6%
ETH1.59%1.80%20.2%18.6%20.2%
BNB1.68%1.80%16.7%9.6%16.7%
SOL2.50%2.63%13.5%13.5%13.5%
AVAX2.47%2.96%7.7%2.5%2.5%
DOT1.96%2.06%6.1%10.6%6.1%
BTC1.29%1.80%4.7%3.7%4.7%
DOGE2.39%2.87%4.4%12.6%12.6%
POL3.13%3.29%-8.1%-10.4%-10.4%

Green bold = outperformed static. Red = underperformed. Assets sorted by static return descending.

Why 2.59% Is Already Near-Optimal

An important finding: our Dynamic Grid v3 with trailing-up already provides significant volatility adaptation. The grid trails prices upward and recycles depleted buy levels as new sells. This built-in mechanism means:

1

The grid already widens upward as prices rise (trailing-up adds new levels above)

2

Level recycling means the grid doesn't exhaust capital during drops — depleted buys become new sells

3

Adding ATR-based spacing on top of this may be redundant — the trailing mechanism already handles much of what adaptive spacing aims to solve

Implication for v2.1 Implementation:

The adaptive spacing concept is sound in theory, but the existing Dynamic Grid already captures much of the benefit. A Phase 2 implementation with continuous ATR monitoring (recalculating spacing as markets shift) might perform differently from this Phase 1 test, which uses a one-time ATR from Year 1 data.

Limitations

1

Static Φ values are heuristic

The rating-based multipliers (0.85–1.20) are initial estimates, not optimized. Different Φ values could change the results.

2

ATR calculated once

ATR is fixed from Year 1 — no mid-Year-2 adaptation. A continuously adaptive version (Phase 2) may behave differently.

3

Same grid range for all variants

Both methods use the same Year 1 upper/lower limits. Only spacing density changes, not the range itself.

4

Small sample (3 × 13 = 39 observations)

Directional but not statistically definitive. More windows and assets would increase confidence.

5

Floor/ceiling clamps limit some assets

Some assets hit the 1.8% floor (Full) or 6.5% ceiling, capping the adaptive benefit.

Conclusions & Recommendations

Full Adaptive (1.8% floor) is not recommended. Tightening spacing below 2.59% for low-ATR assets consistently hurt performance. The static baseline is already well-calibrated.

Wider-Only (2.59% floor) shows promise. It preserves the proven baseline while adding a volatility safety net for high-ATR assets. Results are approximately neutral with potential upside in mixed markets.

Static 2.59% remains the production recommendation. Our Dynamic Grid's trailing-up mechanism already provides significant volatility adaptation. The static spacing is near-optimal for this mechanism.

Phase 2 (continuous recalculation) merits further study. The one-time ATR from Year 1 may go stale during Year 2. Continuously adaptive spacing that responds to real-time volatility shifts could perform differently.

Related Research

Explore Grid Trading on CoinRoc

The proven Dynamic Grid v1.0 with 2.59% spacing is available now. Start with Discovery to find B- rated cryptos and build your optimized portfolio.

This study tested three grid spacing variants using blind forward testing — ATR and all parameters derived from Year 1 data only. No Year 2 information was used in spacing decisions.

Study date: February 23, 2026 · 13 cryptocurrencies · 3 test periods · 3 spacing variants