Yodacom Research · Paper № 1 · May 2026

Intelligent Grid Trading: What 13 Years of Crypto Data Actually Shows

A Walk-Forward Validation of Regime-Gated Grid Strategies Across 17 Assets

Jeremy J. Black · Yodacom Research

Quantitative analysis by Yodacom AI Research Team

We ran a walk-forward backtest of an adaptive grid trading strategy across 17 major cryptocurrencies, 13 years of daily data, and 135 valid test folds. The uncomfortable number: the grid underperforms buy-and-hold in 59% of folds, with average alpha of −578%. That figure is dominated by bull-run years where assets returned 1,000–8,000% and no finite-return strategy competes. What the data actually shows is a strategy with a 100% win rate against buy-and-hold in confirmed bear-market folds, meaningful capital preservation when markets fall, and better calibration after correcting two measurement errors in our own system. Forward confidence (authors' subjective assessment, not a return forecast): 6.5/10.

Back to Research
Section 1

Key Findings

Bear-Fold Win Rate (2011–2025, n=135)
100%
Win rate in every fold where buy-and-hold lost more than 20%
135 valid folds 17 assets 172,588 daily bars
  • 135 valid walk-forward folds across 17 assets and 13 years, yielding 172,588 daily bars
  • 41.5% of folds beat buy-and-hold — the rest are bull-run years where assets returned 200–8,000%
  • Bear-fold win rate: 100% — every fold where buy-and-hold lost more than 20%, the grid outperformed
  • FIS v2 regime gating improved alpha from −578% to −464%
  • 2025 Sharpe: −0.786 — third-worst behind 2018 (−1.641) and 2022 (−1.291)
  • Forward confidence: 6.5/10 (subjective assessment, not a return forecast)
Section 2

Methodology

This study uses a walk-forward validation framework: a 2-year training window followed by a 1-year test window, sliding forward by 1 year at each step. This structure prevents look-ahead bias by ensuring that every parameter set tested was calibrated exclusively on data preceding the evaluation period.

Training window2 years
Test window1 year
Slide1 year forward
Total folds152 (135 valid, 17 skipped for insufficient history)
SymbolsBTC, ETH, XRP, ADA, LTC, BCH, DOGE, XMR, XLM, EOS, ZEC, DASH, TRX, NEO, XTZ, MANA, STX
Date range2011–2025
Data sourceTiingo daily bars (172,588 observations)
Cost modelRetail Binance.US — 0.40% maker / 0.60% taker / 0.05% slippage

Regime gating is provided by the FIS™ (Fuzzy Inference System): a 10-rule Mamdani engine reading three input signals — price relative to the 200-day moving average, volatility ratio (current vs. trailing), and momentum — and producing an engagement scalar from 0.0 (fully disengaged) to 1.0 (fully deployed). When the engagement scalar falls below threshold, the grid does not open new positions.

FIS v2 introduces EWMA volatility decay with a 7-day half-life, replacing the raw-window volatility input used in v1. This change smooths regime transitions and reduces false disengagement signals during short-term volatility spikes that do not represent structural regime shifts.

Section 3

The Aggregate Numbers

MetricValue
Mean annual return−9.7%
Median annual return+1.6%
Mean Sharpe ratio−0.468
Median Sharpe ratio−0.303
Mean Calmar ratio0.038
Mean max drawdown27.4%
Win rate vs buy-and-hold41.5%
Average alpha vs buy-and-hold−578%

The −578% average alpha figure is the most frequently misread number in this dataset. It requires context. During bull-run years — 2017, 2020, 2021, 2024 — assets in this universe returned 200% to 8,000% in individual folds. No finite-return strategy competes with 5,000% buy-and-hold returns on a small allocation. The grid strategy, which generates return from oscillation within a range, is structurally incapable of matching unidirectional bull-run appreciation. Alpha in those folds is deeply negative by construction, not by failure.

The more meaningful signal is the bear-fold record: a 100% win rate in every fold where buy-and-hold lost more than 20%. Capital preservation in declining markets is the primary value proposition of a regime-gated grid strategy. An investor who avoids deep drawdowns in bear markets does not need to compete with 3,000% returns in bull years to achieve superior long-run compounding — they need only to survive the bear years with capital intact. The grid does that.

Section 4

Sharpe by Year

The chart below shows average Sharpe ratio per calendar year across all valid folds, sorted worst-to-best. The pattern reflects crypto market regimes cleanly: sustained bear markets (2018, 2022, 2025) produce the worst Sharpe figures; the only positive-Sharpe year is 2020, which combined post-COVID recovery volatility with the early Bitcoin bull run — an ideal oscillation environment for a grid strategy.

Chart 1 — Sharpe by Year
Average Sharpe Ratio per Calendar Year (135 valid folds)

Backtested data — not investment advice. Sharpe = (annual return − 0%) / annualized vol; risk-free rate set to 0 for crypto context.

Section 5

What FIS Gating Does

A measurement artifact requires explanation before reading the FIS comparison table. When the FIS gate is applied and the grid is disengaged for a full fold, that fold produces a return of approximately 0% (cash, no deployment) rather than a deeply negative return. Relative to a buy-and-hold that returned, say, +300% in a bull-run fold, a 0% result is still a loss of alpha. The FIS gate therefore converts bull-run folds from "somewhat losing" to "fully losing" on the alpha metric, which mechanically drops the raw win rate from 41.5% to 28.9%. The gate is working correctly — it is preserving capital in bear folds — but the win-rate metric penalizes it for not chasing bull runs.

ConfigurationWin RateAvg Alpha vs B&H
Always-on (no gate)41.5%−578%
FIS v1 (raw volatility)26.7%−486%
FIS v2 (EWMA vol, hl=7d)28.9%−464%

The correct metric is expected return per unit of engaged capital. The gate correctly identifies hostile regimes, preserves capital, and redeploys it when conditions are favorable. The −578% to −464% improvement represents $114 of alpha gained per $100 of buy-and-hold.

Section 6

What We Fixed

Fix 1: Alpha Factor Demotion (Weight 0.15 → 0.00)

The production composite score included an alpha-vs-buy-and-hold factor at 0.15 weight. As documented above, alpha vs. buy-and-hold is structurally negative in bull-run years by design — not by poor strategy execution. Including it as a positive selection signal was selecting against exactly the configurations that perform best in deployable regimes. Removing the alpha factor improved composite-to-Sharpe correlation significantly.

Weight Candidater(composite, Sharpe)r(composite, −drawdown)
Production (alpha=0.15)0.9010.674
Alpha-demoted (alpha=0.00)0.9500.761

The improvement is statistically significant at the 95% confidence interval ([+0.026, +0.076], N=2,000 bootstrap resamples).

Fix 2: Portfolio Volatility Correction (0.30 → 0.16)

The optimizer used a hardcoded portfolio volatility assumption of 0.30 (30% annualized). Actual measured volatility for the always-on grid is 20.97% annualized; for the FIS v2 gated configuration it is 16.07%. The hardcoded figure overstated portfolio volatility by 9–14 percentage points, causing the mean-variance optimizer to systematically underweight the grid sleeve in multi-asset portfolio construction. The corrected input of 0.16 produces allocation recommendations consistent with the actual risk contribution of the strategy.

A third fix is in progress: Hurst exponent threshold recalibration. The current gate uses H < 0.45 as the mean-reverting regime threshold. Zero mean-reverting folds in 13 years of crypto data is a red flag — it suggests the threshold is too tight and is excluding deployable folds. We are testing H < 0.50 and H < 0.52 as candidate thresholds. Results will be incorporated in the forthcoming technical addendum.

Section 7

Forward Confidence: 6.5/10

Authors' subjective assessment — not a return forecast
6.5/10
Walk-forward confidence rating — Paper № 1 · May 2026

The current market context: Bitcoin formed a death cross in November 2025, declining from approximately $110,000 to a trough near $66,000 in February 2026. The current regime reads as −2 (Bear), with 3 of 4 regime indicators bearish. Historical crypto bear markets average 12–18 months in duration; the current bear is approximately 6 months in as of this writing.

The Sharpe pattern by period provides useful framing for what forward conditions look like across the regime cycle:

PeriodAvg SharpeRegime context
2020+0.187Only positive-Sharpe year (ideal oscillation)
2019−0.455Consolidation post-2018 bear
2023–2024 avg−0.140Recovery / late-cycle bull
2025−0.786Sustained bear (current)

Upward-confidence factors: The historical pattern holds — bear markets end and are followed by consolidation and recovery phases where grid performance improves. We are within the historical duration window for a regime shift. Three measurement fixes (alpha demotion, vol correction, Hurst recalibration) improve the strategy's calibration going forward. FIS v2 capital preservation compounds over full market cycles: surviving a bear market with less drawdown means more capital available for the subsequent recovery phase.

Downward-confidence factors: The bear market has not confirmed a bottom. Zero mean-reverting folds across 13 years of crypto data means the strategy has operated almost entirely in trending regimes — 67% of all folds are trending regime, a structural characteristic of cryptocurrency markets that limits the core mechanism of grid strategies. Macro and regulatory unknowns remain elevated.

The 6.5 reflects a strategy we believe is correctly designed, honestly measured, and positioned for regime recovery — but one currently operating in a headwind whose favorable conditions are not dominant in crypto's historical record.

Section 8

What This Means for Practitioners

1. When to run a grid

Macro regime not confirmed bear, and the asset is not in a confirmed uptrend. The ideal condition is lateral consolidation or mild oscillation within a defined range. Post-bear consolidation phases (analogous to 2019 and 2023) are the highest-probability deployment windows based on historical data.

2. When to stand aside

Confirmed bear markets and confirmed uptrends. In both regimes, the directional move dominates the oscillation signal. Grid strategies in sustained trends produce the worst Sharpe readings in this dataset (−1.641 in 2018, −1.291 in 2022). The FIS gate exists precisely to identify these conditions and preserve capital.

3. How to read Sharpe in crypto context

Crypto assets carry 60–150% annualized volatility as a baseline. A Sharpe of 0.0–0.5 is not failure in this asset class — it represents respectable risk-adjusted return given the vol base. The −0.786 Sharpe of 2025 should be read against the asset-class context: this is a strategy doing its job (preserving more capital than buy-and-hold in a bear) in conditions structurally hostile to grids.

The most undervalued feature of a regime-gated grid strategy is what it does not lose.

Section 9

Methodology Appendix

Full methodology details — click to expand

Strategy: AdaptiveGrid v3DynamicMode, geometric spacing, 55 grid levels, 2.59% grid spacing. Fill model: GTX (limit orders fill on price cross, not touch).

Data: Tiingo daily bars, 172,588 total observations across 17 symbols from 2011 through 2025. Data sourced directly from Tiingo API; no survivorship adjustment was applied beyond the 17-symbol universe selection, which was fixed prior to the study.

Walk-forward structure: 2-year training window / 1-year test window / 1-year slide. Training window optimizes grid spacing and FIS thresholds. Test window evaluates out-of-sample performance. No parameter re-optimization during the test window.

Cost model: 0.40% maker / 0.60% taker / 0.05% volatility-conditional slippage, reflecting Binance.US retail tier. All returns are net of these costs.

FIS (Fuzzy Inference System): 10-rule Mamdani engine with centroid defuzzification. Inputs: (1) price vs. 200-day MA, (2) volatility ratio (14-day / 63-day), (3) 20-day momentum. Output: engagement scalar 0.0–1.0. FIS v2 replaces raw volatility ratio with EWMA-smoothed vol (half-life = 7 days).

No look-ahead bias: All parameter sets are calibrated using only data preceding the test window. FIS rule weights are fixed before the study and not updated during test periods. Regime signals use only data available at the time of each daily bar.

Section 10

Important Disclosure

Hypothetical and Backtested Performance. All return and performance figures in this paper are based on historical backtested simulation using the CoinRoc walk-forward research harness. Backtested results do not represent actual trading results and were achieved by means of the retroactive application of a model designed with the benefit of hindsight. Hypothetical performance results have many inherent limitations and no representation is made that any account will or is likely to achieve profits or losses similar to those shown.

Past Performance. Past performance, whether actual or indicated by historical tests, does not guarantee or predict future results. The performance of any investment strategy can and often will differ materially from backtested results.

Not Investment Advice. This paper is for informational and research purposes only and does not constitute investment advice, a solicitation, or an offer to buy or sell any security or investment product. Yodacom Research is not a registered investment adviser. Nothing in this paper should be construed as a recommendation to buy, sell, or hold any asset. Readers should consult a qualified financial professional before making any investment decisions.

Forward-Looking Statements. This paper contains forward-looking statements and projections, including the forward confidence rating and regime outlook. These statements are based on the authors' subjective assessment of historical data and current conditions. They are not forecasts, predictions, or guarantees of future performance. Actual results may differ materially from any forward-looking statement made herein.

Digital Asset Risk. Cryptocurrency and digital asset markets are highly volatile, largely unregulated, and subject to rapid and unpredictable changes in value. Grid trading strategies involve substantial risk of loss, including the potential loss of all invested capital. The strategies described herein are not appropriate for all investors.

Regulatory Status. Cryptocurrency assets are subject to unique and evolving regulatory risks. Regulatory changes in any jurisdiction could materially affect the viability, legality, or profitability of grid trading strategies. No regulatory approval or review of this paper or the strategies described herein has been obtained or sought.

Citation: Black, J. J. (2026). Intelligent Grid Trading: What 13 Years of Crypto Data Actually Shows. Yodacom Research pre-print. Quantitative analysis by Yodacom AI Research Team.

Author: Jeremy Black is CEO of YodaCom.com and the creator of CoinRoc, an AI-powered crypto grid trading analysis platform.

Reproduction with attribution permitted.

Back to Research