Walk-Forward Backtest Results

Walk-forward validation is the gold standard for testing quantitative investment strategies. The concept is straightforward: you split your data into two non-overlapping periods. The first period (in-sample) is used to design, calibrate, and optimize your model. The second period (out-of-sample) is used to test it on data the model has never seen.

This approach directly addresses the most dangerous problem in quantitative finance: overfitting. With enough parameters, any model can be made to look spectacular on historical data. A model that “predicts the past” with 99% accuracy might be completely useless going forward because it memorized noise rather than discovering genuine patterns.

Walk-forward validation prevents this by separating discovery from confirmation. If a strategy performs well on data it was specifically designed for, that tells you very little. If it performs well on data it has never seen, that is meaningful evidence that the underlying signals are real.

Most retail trading tools skip this step entirely. They show you one set of backtested numbers without disclosing whether the model was designed on the same data being used to evaluate it. That is like a student writing the test questions and then claiming a perfect score proves intelligence.

AlphaStocks split the January 2021 through February 2026 period into two distinct windows:

The in-sample period includes the post-COVID rally, the 2022 bear market, and the 2023 recovery — a diverse set of market conditions that stress-tested the formula across multiple regimes.

Once the formula was finalized, it was locked. The out-of-sample period (2024 through early 2026) includes the AI-driven tech rally, sector rotations, and several correction events. The formula was not modified in response to any of these conditions.

The backtest uses the v22 composite scoring formula, which blends four axes into a single 0–10 rating for every S&P 500 stock:

The simplest test: take the 10 highest-scoring stocks at the start of the backtest period (January 2021) and hold them through February 2026 with no rebalancing. This is a pure buy-and-hold test that eliminates any benefit from monthly rotation.

The top-10 portfolio nearly doubled the S&P 500's return over the same period. An annual alpha of +8.4% means that each year, on average, the portfolio outperformed the benchmark by 8.4 percentage points. Over five years, that compounds into a substantial gap.

This buy-and-hold result is particularly significant because it requires no ongoing management. The formula identified high-quality companies at attractive prices with strong momentum, and those characteristics persisted over the following five years.

The more rigorous test uses a larger portfolio with monthly rebalancing. Each month, the 30 highest-scoring stocks are selected, equally weighted, and held until the next rebalance. This tests whether the formula can consistently identify outperforming stocks over time, not just at a single entry point.

Several observations stand out from this data. The top-30 portfolio outperformed the S&P 500 in both the in-sample and out-of-sample periods. The out-of-sample alpha (+8.4%) is lower than the in-sample alpha (+13.9%), which is expected — no model works as well on unseen data as on its training data. However, the alpha remained substantial and positive.

The full-period cumulative alpha of +34.2 percentage points demonstrates the power of consistent, modest annual outperformance compounded over multiple years.

Raw returns tell only half the story. The Sharpe ratio measures return per unit of risk. A higher Sharpe means the portfolio delivered more return for each unit of volatility endured.

The out-of-sample Sharpe ratio of 1.12 is notably higher than the in-sample Sharpe of 0.61. This is the opposite of what you see with overfit strategies, where risk-adjusted returns typically degrade dramatically on unseen data. An improving Sharpe ratio out-of-sample is one of the strongest signals that a model has captured genuine market patterns rather than noise.

For context, a Sharpe ratio above 1.0 is generally considered good for a long-only equity strategy. The S&P 500's historical Sharpe ratio typically falls between 0.3 and 0.7 depending on the measurement period.

Maximum drawdown measures the largest peak-to-trough decline during the test period. It answers the question every investor actually cares about: how bad did it get at the worst point?

The top-30 portfolio experienced roughly 40% less drawdownthan the S&P 500. A maximum drawdown of -14.3% versus -23.9% means the formula-selected stocks fell significantly less during the worst market conditions of the test period.

This matters more than most investors realize. A 24% drawdown requires a 31.6% recovery to break even. A 14% drawdown requires only a 16.3% recovery. Smaller drawdowns mean faster recoveries and less temptation to panic-sell at the bottom.

The lower drawdown is likely a consequence of the Quality axis receiving the largest weight (40%). High-quality companies with strong balance sheets and consistent profitability tend to decline less during market stress. The formula is inherently biased toward resilient businesses, and that bias shows up most clearly during corrections.

The distinction between in-sample and out-of-sample performance is the single most important concept in quantitative finance testing. Many backtests you encounter online are exclusively in-sample: the model was designed and tested on the same data. This guarantees nothing about future performance.

Here is what our walk-forward split reveals:

Consider what an overfit strategy would look like: spectacular in-sample returns followed by zero or negative out-of-sample alpha. The Sharpe ratio would collapse. Drawdowns would increase. Our results show the opposite pattern — risk-adjusted returns improved and drawdowns remained controlled.

The backtest results suggest — but do not prove — that the AlphaStocks composite formula identifies genuinely useful investment signals. Specifically:

The combination of quality, value, momentum, and timing factors appears to produce returns above the benchmark across multiple market conditions. The formula outperformed in the post-COVID rally, the 2022 bear market, and the 2024–2025 AI-driven bull run. It underperformed in sideways periods.

The value-trap killer (Timing = min(Value, Momentum)) appears to reduce drawdowns. By requiring both cheapness and rising prices, the formula avoids the classic trap of buying stocks that are cheap and getting cheaper. This likely explains the substantially lower maximum drawdown.

The out-of-sample results are weaker than in-sample, as expected, but remain meaningfully positive. A strategy that maintains alpha on unseen data, with improving risk-adjusted returns, is a stronger signal than spectacular in-sample numbers that collapse forward.

These results do not mean the formula will continue to outperform. They mean it has demonstrated the ability to do so in a controlled, honestly disclosed test.

Backtest results are a starting point, not a conclusion. Here is how to incorporate them into a sound investment process: