Quant Rules for the Nikkei 225: How Statistical Edge Compares to a Buy-and-Hold

The Problem with Black-Box AI in Stock Analysis

Most 'AI-powered' investment tools produce recommendations without explaining the reasoning behind them. 'Stock X is a Buy' with a confidence of 82% tells you nothing about why — what conditions triggered the signal, how many similar historical situations exist, or how reliable that pattern has been over time.

Kabu Prediction takes a different approach: every rule is explicit. 'Buy 8035 (Tokyo Electron) when VIX > 30' is a complete, testable hypothesis. We know exactly how many times this condition occurred in the past 10 years (29 times), how often the stock rose within 1 week (66%), and the average return when it did (+4.2%).

What Makes a 'Good' Statistical Rule?

We evaluate rules on four dimensions. First, win rate: what percentage of signals resulted in the expected price direction within the target horizon (3 days, 1 week, 1 month, 3 months). Second, edge: the win rate minus 50%, representing the statistical advantage over a random coin flip. Third, Sharpe ratio: risk-adjusted return relative to volatility. Fourth, sample size: rules with fewer than 20 historical instances are flagged as statistically insufficient.

A rule with 65% win rate, +8% edge, Sharpe of 1.4, and 40 instances over 10 years is considered reliable. A rule with 80% win rate but only 5 instances is dismissed as noise.

Rule Categories and Their Edge

Technical Rules (~20 rules)

Based on price and volume patterns: RSI oversold, Bollinger Band lower touch, VIX spike reversal, moving average crossovers, ATR-adjusted momentum. Technical rules tend to perform best on shorter horizons (3-day to 1-week) and are most effective for liquid large-cap stocks like those in the Nikkei 225.

Fundamental Rules (~10 rules)

Based on financial metrics: P/B < 0.8x cross-sectional screen, P/E discount to sector median, earnings yield spread vs JGB yields. Fundamental rules show stronger edge on longer horizons (1-month to 3-month) and are particularly effective for banking and industrial stocks where book value is a reliable anchor.

Macro Rules (~8 rules)

Based on external market conditions: USD/JPY rate thresholds, VIX levels, S&P 500 momentum, copper price trends. Macro rules work best for export-oriented industries (auto, trading companies, electronics) where currency and commodity prices are dominant earnings drivers.

Cross-Sectional Rules (~7 rules)

Compare stocks against their sector peers. A stock trading at 2 standard deviations below its sector's median P/B, combined with improving ROE momentum, is a cross-sectional buy signal. These rules are particularly powerful during sector rotations.

Backtest Results: 2016–2026

Across all 45 rules and ~200 Nikkei 225 stocks, the aggregate statistics are: average win rate 56.3%, average edge +6.2%, average Sharpe ratio 1.08. The best single rule-stock combination over the period achieved a 72% win rate on the 1-week horizon with a Sharpe ratio of 2.1.

For context, a pure buy-and-hold of the Nikkei 225 over the same 2016–2026 period returned approximately +110% total, with a Sharpe ratio of around 0.7 (high volatility years like 2020 and 2022 dragged the risk-adjusted number down).

Walk-Forward Validation: Avoiding the Overfitting Trap

Every rule on this platform was validated using a walk-forward methodology. We train on historical data up to year N, test on year N+1, and roll forward annually. This prevents look-ahead bias and ensures the reported win rates reflect how rules would have performed in real-time, not just in hindsight.

Rules where the out-of-sample score fell below 50% of the in-sample score were automatically rejected. This conservative filtering means the 45 published rules represent the survivors of a rigorous selection process, not hand-picked examples.

The Nikkei 225 Advantage for Rule-Based Trading

Not all markets are equally suitable for rule-based quantitative analysis. The Nikkei 225 has several characteristics that make it particularly amenable: high liquidity (all 225 constituents can be traded with minimal slippage), long price history (Japanese equities have continuous, clean data back to the 1980s), and distinct macro cycles tied to yen movements and global commodity prices that create repeating patterns.

Smaller markets or illiquid stocks introduce execution risk that degrades real-world performance below backtest results. The Nikkei 225's depth means that stated backtest results are achievable in practice for most retail position sizes.

How Win Rate Translates to Returns

A 60% win rate means 6 out of 10 signals move in the expected direction. With 0% on losses and +X% on wins, you need X > 0 to be profitable — but real trades have both wins and losses of varying sizes. Edge (win rate - 50%) is a better summary statistic because it accounts for the excess over random.

Our average edge of +6.2% across all rules means that, on average, these rules are right 56.2% of the time. With proper position sizing (proportional to edge), this translates to positive expected value after costs.

Using the Data: A Practical Workflow

Step 1: Visit the Dashboard (/dashboard) and sort by win rate or edge. Filter for 1-week horizon rules with edge > 8%.

Step 2: Click any row to see the stock's full profile — all 45 rules, their stats, the dominant driver, and historical Sharpe.

Step 3: Cross-reference with the sector page to understand macro context (is the whole sector showing elevated edge right now?).

Step 4: Use the Track Record page (/track-record) to audit aggregate statistics and verify the platform's historical accuracy.

Limitations and Honest Caveats

Rule performance degrades over time as market conditions change. A rule that worked well in 2016–2020 may underperform in 2024–2026 as market participants adapt or structural conditions shift. We monitor rule performance on a rolling basis and retire underperforming rules.

Backtests assume instant execution at closing prices. Real-world execution introduces slippage, especially for larger position sizes. The stated backtest results apply most cleanly to retail-sized positions (under ~¥10M per trade).

This platform provides statistical analysis only — not investment advice. All investment decisions are the sole responsibility of the individual investor.

All analysis on this platform is based on statistical backtests and is for informational purposes only.