Smarter Grids: Adapting to Market Volatility

How ATR-Based Grid Spacing Compares to Fixed 2.59%

Date: 2026-02-23

The Idea

Our current Dynamic Grid strategy uses a fixed 2.59% spacing between buy and sell orders. But not all cryptos move the same way — Solana swings 5% in a day while Ethereum might only move 1.5%.

What if we adapted the grid spacing to match each crypto's actual volatility?

That's what Adaptive Grid v2.1 does. It uses the Average True Range (ATR) — a standard measure of how much an asset moves — to set the grid spacing for each crypto individually.

  Fixed Grid (v1.0)         Adaptive Grid (v2.1)
  ─────────────────         ────────────────────
  Same spacing for          Wider spacing for
  every crypto              volatile assets (SOL, DOGE)
  (2.59% always)            Tighter spacing for
                            stable assets (ETH, BNB)

How We Tested It

We ran both approaches — fixed and adaptive — through the same 3 test periods with 13 cryptocurrencies. Same data, same blind forward test, only the spacing changed.

The Results

Across all assets and test periods (39 total observations):

  Adaptive vs Fixed — Per-Asset Results
  ─────────────────────────────────────────────────────
  Return win rate:     10/39 (26%)
  Mean return change:  -2.5pp
  Mean Calmar change:  +0.04
  Range:               -17.4pp to +14.5pp

Portfolio-Level (Calmar CDaR Optimized)

  Calmar CDaR Portfolio: Fixed vs Adaptive
  ─────────────────────────────────────────────────────
  W1       Fixed: 77.2%    Adaptive: 70.9%    Δ -6.2pp
  W2       Fixed: 45.8%    Adaptive: 39.8%    Δ -6.0pp
  W3       Fixed: 10.8%    Adaptive: 8.3%     Δ -2.5pp

What This Means

In this test, the adaptive approach did not consistently outperform fixed spacing on raw returns. However, it may still offer benefits for risk management and drawdown reduction that require further tuning.

The key insight: one spacing doesn't fit all. High-volatility assets like SOL and DOGE need more room to breathe, while stable assets like ETH can profit from tighter grids that capture smaller price movements.

Honest Limitations

1. The ATR is calculated once from historical data — it doesn't update during trading.
2. The multiplier values (Φ) are initial estimates, not optimized.
3. Three test periods is directional, not definitive.
4. Past performance ≠ future results.

What's Next

This is Phase 1 validation — proving the concept works with a one-time ATR calculation. Phase 2 would implement continuous ATR monitoring that adjusts the grid as market conditions change in real time.

Study date: 2026-02-23 | 13 cryptocurrencies | 3 test periods | Blind forward test

×Click anywhere to close