Mean Reversion Stocks

Mean reversion is a statistical approach with defined entry criteria (2 SD from mean, RSI extreme, reversal confirmation) and exit rules. Bottom fishing typically refers to buying falling stocks hoping they'll bounce - without rigorous criteria. Mean reversion has a mathematical edge based on statistical properties; bottom fishing is often hope-based. Mean reversion also applies to overbought conditions (shorting), while bottom fishing is only about buying dips.

No, mean reversion works best on SGX large-cap stocks and REITs with stable fundamentals, high liquidity, and established trading ranges. Avoid: penny and illiquid small-caps (can trend or gap for extended periods), stocks with pending events (results, M&A), stocks in structural decline, and names with thin liquidity. STI constituents and liquid mid-caps are preferred. Note that liquid single-stock options are scarce in Singapore, so leveraged or hedged exposure is usually via structured warrants, DLCs, or CFDs rather than options.

Well-executed mean reversion trades have a 50-60% success rate. This might seem modest, but with proper risk-reward (2:1 or better), you're profitable even at a 40% win rate. The edge comes from statistical probability - buying at 2 SD below mean has a 95%+ historical probability of price eventually trading back to mean. The key is managing the 5% where it doesn't revert through stop losses.

For Singapore large-caps, mean reversion typically takes 3-7 trading days. Set a time stop of 5 days for intraday-originated trades and 10 days for swing trades. If price hasn't begun reverting within 1.5x the expected half-life, the original thesis may be wrong - exit regardless of current P&L. Holding too long ties up capital and often results in an eventual stop-out anyway.

Generally no - averaging down in mean reversion is dangerous because if your thesis is wrong (the stock is trending, not reverting), you're adding to a losing position. However, a planned scaling approach can work: enter 50% at 2 SD, add 50% at 2.5 SD with the original stop still valid. This must be pre-planned with position sizing that accounts for the full position at the worst entry. Never average down reactively.

Key differences: (1) ADX below 20 suggests range-bound (mean reverting), above 25 suggests trending (continuation likely). (2) Weekly chart: if the weekly trend is down, daily oversold may just be a pause before further decline. (3) Volume pattern: mean reversion shows declining volume into the low then a spike on reversal; trend continuation shows high volume throughout the move. (4) RSI behavior: mean reversion RSI bounces from extreme quickly; trending RSI stays below 30 for an extended period (chronic oversold).

Absolutely - this is the ideal approach. Mean reversion and trend following have negative correlation (when one struggles, the other often works). Run both strategies with separate capital allocation. In trending markets (high ADX, directional volatility), trend following dominates. In ranging markets (low ADX, calm volatility), mean reversion dominates. A combined portfolio shows more consistent returns with lower drawdowns than either strategy alone.

Avoid mean reversion 1 week before and 1 week after a stock's results announcement. Most SGX companies report half-yearly, so there are fewer earnings events than in India, but each is higher-impact. The extreme move may be fundamental (revaluation based on results) rather than temporary. After results, wait for a new trading range to establish - the old 20-day mean is obsolete. If you're already in a position and results are announced unexpectedly, exit immediately - you're now in an unpredictable fundamental event, not a statistical setup.

ATR-based stops adapt to current volatility. For mean reversion: Initial stop = Entry - 2 x ATR for longs (Entry + 2 x ATR for shorts). This places the stop outside normal noise. As the trade moves favorably, trail the stop to Entry - 1.5 x ATR, then Entry - 1 x ATR. Using ATR rather than a fixed percentage prevents getting stopped out by normal volatility while protecting against adverse moves. A 14-period ATR works well for SGX stocks; remember to size positions in 100-share board lots.

Sector rotation can invalidate mean reversion. If money is flowing out of REITs (e.g. on a rate spike) into banks, an oversold REIT may continue declining despite statistical extremes. Monitor: (1) Sector relative strength - is the sector underperforming the STI? (2) Foreign institutional fund flows by sector (Singapore is a foreign-flow-driven market, without India's domestic FII/DII split). (3) Individual stock vs sector divergence. Best mean reversion: a stock oversold while its sector is stable or strong. Worst: a stock oversold inside an oversold sector (rotation away).

Walk-forward prevents overfitting: (1) Divide historical data into in-sample (IS) and out-of-sample (OOS) periods, e.g., 2 years IS, 6 months OOS. (2) Optimize parameters (entry threshold, lookback period, etc.) on the IS period. (3) Test optimized parameters on the OOS period without modification. (4) Roll forward: next period, IS = previous IS + OOS, new OOS = next 6 months. (5) Average OOS results across all windows for a realistic performance estimate. Parameters that degrade severely in OOS indicate overfitting.

Eigenvalue decomposition on the covariance matrix of stock returns identifies principal components. The last few eigenvectors (smallest eigenvalues) represent portfolios with minimum variance - these are most likely to be stationary/mean-reverting. Construct a spread portfolio using these eigenvector weights. This mathematically identifies the most mean-reverting linear combination of stocks in your universe. It's more robust than pairwise analysis for multi-stock mean reversion, though Singapore's smaller liquid universe means fewer independent components to work with. Eigenportfolios can be traded using the same Z-score framework.

Standard mean reversion assumes a normal distribution, but stock returns have fat tails (more extreme moves than normal predicts). Adjustments: (1) Use Student's t-distribution instead of normal for Z-score calculations - this requires higher thresholds for the same statistical significance. (2) Implement hard-coded maximum loss limits regardless of model signals. (3) Maintain tail risk hedges (index put warrants or short index futures) sized to cover the portfolio in 3+ sigma events. (4) Reduce position sizes when the kurtosis of recent returns exceeds the historical norm - fat tails are clustering.

Microstructure matters significantly, and more so in Singapore where liquidity outside the top names is thinner than in India, so market impact is higher: (1) Mean reversion often requires entering against prevailing order flow (buying when others are selling) - you face adverse selection and wider effective spreads. (2) Limit orders may not fill as price continues against you; market orders ensure a fill but face slippage. (3) Optimal execution: use limit orders at slight improvement to the best bid/offer, with a time limit converting to market if unfilled. (4) Avoid executing at the open, the close, or right around the midday break when microstructure is most adverse. (5) For systematic portfolios, spread execution across the day to reduce impact, and consider that contra positions must still close by T+2.

Real-time monitoring: (1) Rolling win rate - alert if it drops below 40% over 20 trades (potential regime change). (2) Rolling Sharpe ratio - alert if it drops below 0.5 annualized (strategy degradation). (3) Running ADF test on mean reversion signals - if the p-value rises above 0.1, the mean reversion property may be weakening. (4) Correlation of returns with the STI - if rising, the strategy is losing market neutrality (watch closely given the index's bank concentration). (5) Maximum drawdown - if it exceeds 1.5x the historical max DD, pause and investigate. Automated alerts enable a quick response to strategy breakdown.