What if a simple signal could give you an edge in predicting which stock markets are about to outperform? Imagine spotting economic trends before they hit the headlines and letting that guide your trades. In this article, I show how long-term interest rates, a surprisingly clean macro signal, can be harnessed to create a systematic trading strategy that captures these moves.

Why Rates Matter

Markets don’t move in isolation. Long-term interest rates are one of the cleanest signals we have about how investors view growth, inflation, and monetary policy. When 10-year yields in one country move higher relative to another, it tells us something about relative economic strength and often about how their stock markets might behave.

This article walks through how I built and backtested a relative value (RV) macro strategy that trades equity indices against each other, guided by 10-year rate spreads.

The Core Idea

The strategy is simple in spirit:

• Compare the 10-year yields of two countries.

• Standardize that spread into a Z-score (so we know if it’s unusually wide or tight).

• When the spread is stretched too far, take a position:

  • If Country A’s yield is rising vs. Country B’s → long Country A’s equity, short Country B’s.

  • If the opposite happens → reverse the trade.

Think of it as betting that interest rate differentials eventually show up in equity performance.

Trading the Signal

There are two ways to interpret the spread:

1. Momentum view: if yields are diverging, equities may continue to diverge too.

2. Mean reversion view: an extreme spread could snap back, pulling equities with it.

In this project, I leaned on the momentum interpretation: ride the divergence when the Z-score breaks out past a threshold.

Rolling Mean of the Spread over last 10 days:

Rolling Standard Deviation of the Spread over last 10 days:

Z-score:

Signal Generation:

Execution and Risk Management

Once a signal fires, the trade goes on at that day’s close: long one index, short the other. A few important rules shape the risk:

• Position sizing: based on volatility and hedge ratios (rolling beta between the two indices). This keeps the pair balanced.

• Stop-loss / take-profit: max loss per trade capped at 1% of assets under management; take profit set at 2x that risk.

• Holding period: no trade stays open beyond 32 trading days, even if profitable.

• Signal reversal: if the Z-score flips sides, the trade flips too.

A Quick Dive into the Math

(This section will be developed in details in the next articles)

If you’re into the details:

• Volatility is measured with an exponentially weighted moving average (EWMA) of log returns.

• Hedge ratios are calculated with a rolling Ridge regression between the two assets’ returns—regularization helps avoid noisy fits.

• Position size is scaled so that risk is balanced across both legs of the trade.

These rules make sure no single leg dominates the portfolio’s risk.

The Backtest Setup

For this study, I focused on a handful of equity indices and bond yields:

• Rates: US 10Y, Germany 10Y, Japan 10Y

• Equities: Nasdaq, Germany Equity, Spain Equity, Japan Equity

Currencies were converted into USD so everything was on the same scale. For example:

• German equities × EURUSD

• Spanish equities × EURUSD

• Japanese equities ÷ USDJPY

I built five relative value models pairing different equity indices with their respective 10Y spreads.

Each country is used together to its 10y rate, we assumed the German 10y rate as the European rate to be used with the Spain Equity.

Stress Testing the Strategy

Every model was tested across:

• Z-score thresholds: 0.5, 1.0, 1.5, 2.0, 2.5

• Holding limits: 3, 5, 7, 10, 15, 20, 25, 32 days

That’s 200 simulations in total, each evaluated on both a training and test set:

• Training: June 2010 – Jan 2020

• Testing: Jan 2020 – Feb 2025

The aim wasn’t just to find a profitable model but to see which thresholds and holding rules consistently hold up across regimes, a big deal when you’re trading macro, where cycles can shift fast.

Why This Matters

This kind of strategy sits at the intersection of macro fundamentals and systematic trading. It’s not just chart patterns or price momentum, it’s about structuring trades that reflect how economies diverge or converge.

In the next piece, I’ll share the backtest results, which models held up best, and whether this type of RV strategy could survive in today’s post-zero-rate world.

Train Dataset Results

After running the 200 simulations, I analyzed how the Sharpe ratio behaved in relation to both the maximum holding period and the Z-score threshold. The results are as follows:

After running the 200 simulations, I selected the top two configurations for each asset pair based on the highest Sharpe ratios. The selection was made to balance performance comparison across pairs while retaining a manageable subset for further analysis. The results are summarized below:

Conclusions from test results

1. Stronger Signals = Better Risk-Adjusted Returns

Across multiple asset pairs, higher Z-score thresholds (e.g., 2.0 or 2.5) consistently led to higher Sharpe ratios compared to lower thresholds (e.g., 0.5 or 1.0). This trend implies that:

• Extreme interest rate spreads are more predictive of future equity divergence.

• Entering trades only when macro signals are strong helps avoid noise or false positives.

• The strategy behaves more like a momentum filter, activating only in high-confidence scenarios.

For example:

• For Nasdaq vs Spain, the top configurations used a Z-score of 2.5, producing Sharpe ratios of 0.87 and 0.80.

• For Germany vs Nasdaq, the best results came from a Z-score of 2.0, with lower thresholds underperforming.

This aligns with momentum theory: stronger macro dislocations tend to persist, and capital often flows toward higher-yielding markets when the spread is large enough to justify reallocation.

2. Optimal Holding Period: Short to Medium-Term (3–7 days)

The most effective holding periods were generally in the 3 to 7-day range. This suggests that:

• Market reactions to rate spreads are relatively quick and front-loaded.

• Extending holding periods beyond 10–20 days often leads to decaying returns or reversals.

• The edge captured by the interest rate signal dilutes over time, likely due to mean reversion or other macro catalysts.

Supporting examples:

• Japan vs Spain produced Sharpe > 1.0 with tl_3 and tl_5, while longer holding times degraded performance.

• Nasdaq vs Japan had Sharpe = 0.94 for 2.0_tl_3, and 0.79 for 1.0_tl_3 — reinforcing both the value of short horizons and strong signals.

This supports the view that macro information gets priced in quickly, particularly when linked to yield spreads.

3. Regional Trends: Japan & Spain Show Strong Momentum Response

Several pairs involving Japan and Spain yielded sharply higher Sharpe ratios. Potential reasons include:

• Greater volatility in their interest rates, creating more pronounced signals.

• Higher sensitivity of local equities to macro variables like rates and inflation expectations.

• Possibly more inefficient equity pricing, allowing macro signals to have greater predictive power.

Examples:

• Japan vs Spain: Sharpe ratios of 1.27 (0.5_tl_3) and 1.05 (0.5_tl_5) – top across all simulations.

This may justify regional model tailoring, with more aggressive sizing or allocation to these country pairs if the spread signal remains persistent.

4. Mixed Performance for Germany & Nasdaq

Pairs involving Germany or Nasdaq generally showed more moderate Sharpe ratios (typically in the 0.3–0.5 range), suggesting that:

• These markets may exhibit more efficient pricing, leaving less room for macro-driven strategies to extract persistent alpha.

• The interest rate spread signal may carry less predictive power on its own, potentially overshadowed by local factors such as earnings growth in tech-heavy indices (Nasdaq) or broader economic fundamentals (Germany).

However, an interesting exception emerged with the Germany vs Japan pair:
It was the only case where longer holding periods (20–32 days) led to better Sharpe ratios (0.39 and 0.32, respectively). This may indicate:

• A slower adjustment dynamic between these two economies.

• Or that persistent macro divergence between Germany and Japan provides more lasting trade opportunities compared to other pairs.

Even in these less responsive pairs, configurations with higher thresholds and well-chosen holding times still outperformed others, in line with the broader trend.

Portfolio from Train Results

Lastly, I created an equally weighted portfolio out of the selected models, in the following you can see the results annualized by 252 trading days:

Test Performances

After having run the train and selected the best performing models, I re-ran the backtest with the test window going from 01/01/2020 to 05/02/2025:

Train/Test Performance Consistency Analysis

Results indicate that most selected models preserved reasonable risk-adjusted returns in the test set:

• Japan vs Spain maintained strong performance (train Sharpe: 1.27 / 1.05 → test Sharpe: 0.83 / 0.81), suggesting high robustness and consistency.

• Germany vs Japan also improved in test Sharpe (0.39 / 0.32 → 0.68 / 0.45), indicating potential underestimation during training.

• Nasdaq vs Japan showed performance degradation (0.94 / 0.79 → 0.51 / 0.30), though the Sharpe ratios remained positive, showing partial generalization.

• Germany vs Nasdaq exhibited moderate degradation (0.54 / 0.38 → 0.33 / 0.13), yet preserved profitability.

• Nasdaq vs Spain was the only pair to turn negative in the test (0.87 / 0.80 → -0.26 / -0.45), highlighting structural changes in asset behavior.

Overall, 4 out of 5 pairs retained positive Sharpe ratios out-of-sample, validating the selection process. While some decline in Sharpe is expected due to regime shifts and the absence of lookahead bias, the majority of the strategies continued to deliver favorable risk-adjusted returns.

Nasdaq vs Spain may require further investigation and it will be removed in the final OOS portfolio construction.

The portfolio construction has been the same, in the below you can see the performances of the test datasets of the selected models:

Conclusion

In this analysis, we explored a relative value macro strategy that leverages divergences in 10-year government bond yields to generate trades between equity indices. By standardizing yield spreads into Z-scores and combining them with disciplined risk management—including stop-loss, take-profit, and time-based exits—we were able to backtest multiple models and evaluate their performance across different thresholds and holding periods.

The results show that thoughtful integration of macro signals with robust portfolio rules can provide meaningful Sharpe and Sortino ratios while keeping drawdowns under control. This framework not only highlights the value of relative value strategies in macro-driven markets but also provides a reproducible methodology for further experimentation and optimization.

For those interested in digging deeper, all the code and data used for this analysis are available on my GitHub:

git_hub_link

In the free article, we saw how a simple, rule-based approach could deliver solid results on a single asset and composition of the three {BTC, ETH, SOL}.

In this advanced edition, available exclusively to paid subscribers, we push the research further with a more realistic, risk-aware framework:

• Expanded asset universe → reducing dependency on a single market regime.

• Proprietary risk framework → dynamic volatility-based position sizing and an adaptive trailing stop loss.

• Controlled parameter exploration → small parameter bucket to avoid overfitting.

• Realistic validation → rolling window out-of-sample testing to verify robustness across different market conditions.

Read more

Originally designed for SPY, we adapted the logic to crypto markets and ran it through our high-quality, minute-level historical datasets. The results? Let’s just say a near-1 Sharpe ratio emerged — purely from the raw rules the subscriber provided.

We tested the strategy on three major crypto pairs:

• SOLUSDT

• BTCUSDT

• ETHUSDT

Below is the cleaned-up, systematized version of the strategy we used, presented for full transparency and reproducibility.

Subscribe now

1. Entry Logic: 15-Minute ORB with Volume Confirmation

• 1.1 Define the Opening Range: Use the first 15-minute candle of the session to determine the range (high and low).

• 1.2 Entry Trigger: Enter long if a 5-minute candle closes above the 15-minute ORB high.

• 1.3 Short Setup: mirror the long logic in reverse.

• 1.4 Volume Filter: The breakout must be confirmed by higher volume compared to the last three 5-minute candles.
  → If the breakout candle has higher volume than the previous 3-5 minute candles, enter long.

2. Position Management

• 2.1 Target: Aim for a profit target between 1R and 1.5R.

• 2.2 Stop Loss Options:
  Several stop-loss levels were considered:

  • 2.2.1 The low of the ORB (conservative SL).

  • 2.2.2 Based on ATR (Average True Range) at entry.

• 2.3 Exit Rule:
  Hold the position until price closes below the 21 EMA on the 2-minute chart. This dynamic exit allows trades to ride momentum while keeping risk managed.

3. Market Conditions Filter: Asset Context Required

To increase the probability of success, this strategy is only activated under specific asset conditions:

• 3.1 If the asset is trending upwards:
  Use the 21 EMA on a higher timeframe (e.g., 15m or 1H) to confirm bullish trend.

• 3.2 If asset is moving sideways:
  Use the RSI (Relative Strength Index) on a higher timeframe; only trade if it is between 40 and 60, indicating a neutral but stable market.

(ORB) Introduction

The Opening Range Breakout (ORB) is a time-tested intraday strategy, most commonly used in equities and commodities. It typically focuses on the first 15–30 minutes after market open — a window where volatility and liquidity spikes as institutional players and news-driven flows hit the tape.

In traditional markets like equities, this "opening spike" occurs right at the market open (e.g., 9:30 AM EST for NYSE/NASDAQ). In commodities, the same principle applies, tied to futures market opens.

But what about crypto, where the market never sleeps?

In our past article: Inside the Killzone: Timing Crypto Like a Pro we explored how crypto markets experience their own "pseudo-opening", typically between 14:00 and 15:00 UTC. This aligns with the overlap of U.S. and European trading hours, and often sees a surge in liquidity and volatility, making it the ideal window to apply ORB logic even in a 24/7 market.

In this article, we backtest a 1h ORB strategy with volume confirmation, submitted by one of our subscribers. We’ve refined the rules, defined clear entry/exit criteria, and applied it across historical data to see how it performs.

Let’s dive in.

Backtest Results on Crypto Assets

Although the subscriber originally trades this strategy on SPY, we didn’t have access to reliable 1-minute data for SPY at the moment. Instead, we applied the same strategy to three major crypto assets: BTC, ETH, and SOL, using our high-quality 1-minute historical datasets.

If you have 1-minute data for other assets you'd like us to test this (or other) strategies on, feel free to send it over — we’re happy to backtest community-submitted strategies.

Parameters Used for This Backtest

• ORB Range: First 1h candle after 14:00 UTC

• Entry Long: First 5-minute candle that closes above the ORB high, with volume higher than the last 3 x 5-minute candles

• Entry Short: First 5-minute candle that closes below the ORB low, with volume higher than the last 3 x 5-minute candles

• Stop Loss Long Positions : Low of the ORB

• Stop Loss Short Positions : High of the ORB

• Take Profit (TP): 3R

• Holding Period: 1D

• Market Condition Filter: Only trade when Asset trends up or moves sideways (approximated via a proxy in crypto), EMA 21 used.

• Time for Trading: from 15:00 to 17:00 UTC (Based on article cited above)

Performance Results

We ran the strategy independently on ETH, BTC, and SOL.

ETH Annualized Performance

• Return: 6.78

• Volatility: 19.65

• Sharpe Ratio: 0.43

• Sortino Ratio: 0.73

• Calmar Ratio: 0.16

• Max Drawdown: -29.06%

• Period: 2024/08/08 → 2025/06/25

📉 Comment: ETH shows modest profitability with moderate volatility. The Sharpe ratio is on the lower end, suggesting limited risk-adjusted return, and the drawdown is quite deep, showing that risk control could be improved.

BTC Annualized Performance

• Return: 8.19

• Volatility: 18.96

• Sharpe Ratio: 0.51

• Sortino Ratio: 0.91

• Calmar Ratio: 0.33

• Max Drawdown: -17.02%

• Period: 2022/03/27 → 2022/06/05

📉 Comment: BTC yields better risk-adjusted performance than ETH, with a higher Sharpe and notably lower drawdown. The Sortino and Calmar ratios suggest a more favorable return-to-risk profile, even over a shorter period.

SOL Annualized Performance

• Return: 16.51

• Volatility: 19.03

• Sharpe Ratio: 0.90

• Sortino Ratio: 1.59

• Calmar Ratio: 0.58

• Max Drawdown: -19.03%

• Period: 2022/11/24 → 2023/11/28

📈 Comment: SOL stands out as the best performer. Strong return combined with controlled risk results in excellent Sharpe, Sortino, and Calmar ratios. The strategy appears to have worked particularly well on this asset, possibly due to cleaner intraday trends or more frequent breakouts.

📊 Portfolio Backtest (Equally Weighted)

We also tested the performance of an equally weighted (EW) portfolio, combining the three individual models into a single strategy.

• Return: 11.31

• Volatility: 14.33

• Sharpe Ratio: 0.92

• Sortino Ratio: 1.43

• Calmar Ratio: 0.55

• Max Drawdown: -13.95%

• Period: 2022/03/27 → 2022/06/05

📌 Comment: As expected, combining strategies across assets significantly improved the risk metrics. Diversification helps smooth the equity curve, reduce drawdown, and increase the Sharpe ratio. This illustrates the importance of cross-asset model blending, which we emphasize in our research.

Conclusion & What’s Next

What makes these results particularly striking is what we didn’t do.

We used a static, unoptimized and unexplored set of parameters, applied the same logic across all assets, and ran the strategy without any custom risk framework or model fine-tuning. And yet — we still observed a Sharpe ratio close to 1.

This suggests the underlying idea has real edge. But it also means there's significant untapped potential.

Where This Can Go Next

With the right process, performance could be pushed much further. Here's what future iterations should explore:

• Out-of-sample testing to validate robustness

• Systematic parameter exploration (volume thresholds, exits, filters...)

• Asset-specific tuning and broader cross-asset diversification

• Dynamic position sizing and risk frameworks

• Ensembling multiple models for stability and consistency

All of these are part of the advanced research services we provide our clients. If you're looking to turn raw strategy ideas into robust, risk-adjusted trading systems — check out what we offer at www.stoic-rt.com.

We're also preparing to release an enhanced version of this model, powered by our internal risk engine and allocation logic, with significantly improved metrics. This release will be available soon to our paid subscribers only.

Submit Your Strategy for Backtesting

We’re always open to testing more strategies from our community.
If you have an idea — simple or complex — send it our way!
We’ll systematize it, test it, and possibly feature it in a future article.