Studies 138-157 — twenty varied teardowns: ML (Random Forest, AI ETF), microstructure (Amihud, turnover), events (split drift, dividend capture), allocation/global (Permanent Portfolio, home bias, country & FX momentum), quirky seasonality (lunar, daylight-saving, SAD), long-run valuation (stocks-for-the-long-run, inflation hedge), fundamentals (NOA, leverage, asset turnover), and position sizing (martingale, Kelly)

Twenty deliberately VARIED ideas across ten families (zero new technical-indicator
studies). Data widened: yfinance + the Shiller dataset (GitHub) + the EDGAR
fundamentals cache; FRED/CBOE/AAII remain unreachable in this sandbox.

- 138 Random-Forest    — walk-forward RF; a shuffled-label control matches it (both just learn "be long"). NONE/MIRAGE
- 139 AI-Powered-ETF   — AIEQ (IBM Watson) trails SPY ~5pp/yr, alpha t=-1.28. WEAK/MIRAGE
- 140 Amihud-Illiquidity — t=17 illiquidity "premium" is a survivorship-bias ghost. MIXED/MIRAGE
- 141 Turnover-Anomaly — the effect inverts on the survivor S&P 500. MIXED/MIRAGE
- 142 Split-Drift      — split stocks trail their own baseline post-effective-date. NONE/MIRAGE
- 143 Dividend-Capture — the ex-date drop offsets the dividend; costs+tax sink it. NONE/MIRAGE
- 144 Permanent-Portfolio — Browne 25/25/25/25: real low-drawdown all-weather mix, but forfeits return. REAL/FRAGILE
- 145 Home-Bias        — rising global correlation has thinned the diversification lunch. NONE/MIRAGE
- 146 Country-Momentum — a real but thin, cost-sensitive country-ETF momentum. WEAK/FRAGILE
- 147 FX-Momentum      — a documented but decayed FX factor, gone after costs. WEAK/MIRAGE
- 148 Lunar-Effect     — no full-moon return effect survives. NONE/MIRAGE
- 149 Daylight-Saving  — the clock-change slump doesn't hold. NONE/MIRAGE
- 150 SAD-Effect       — the seasonal-daylight return pattern is noise here. NONE/MIRAGE
- 151 Stocks-For-Long-Run — Siegel's real equity premium holds in every long window, but the unit is the decade. REAL/FRAGILE
- 152 Inflation-Hedge  — stocks are a poor SHORT-run inflation hedge (Fama-Schwert). MIXED/MIRAGE
- 153 Net-Operating-Assets — a faint balance-sheet-bloat signal, muted on large caps. WEAK/FRAGILE
- 154 Leverage-Anomaly — no leverage premium on the survivor panel. NONE/MIRAGE
- 155 Asset-Turnover   — efficiency doesn't forecast returns here. NONE/MIRAGE
- 156 Martingale       — averaging down doesn't beat buy-and-hold and courts ruin. NONE/MIRAGE
- 157 Kelly-Sizing     — the maths is right but the drawdowns/estimation risk make it impractical. WEAK/FRAGILE

Bench now 157 studies (156 stamped): 28 Real / 68 Weak / 60 None; 3 Investable /
42 Fragile / 111 Mirage. Two new real-but-fragile signals (Permanent Portfolio,
Stocks-for-the-Long-Run) join the podium; ML is now 0-for-6. Published: 20 README
rows, bench.md score+families+podium synced, bench_map.png regenerated, tests.yml
wired. Gate 157/157 green, all study tests passing, all 40 notebooks executed clean.

Author: Guillain-RDCDE

.github/workflows/tests.yml

@@ -447,6 +447,66 @@ jobs:
       - name: Study 137 — Mansfield-RS test-suite
         run: pytest -q studies/137-mansfield-rs/tests
 
+      - name: Study 138 — Random-Forest test-suite
+        run: pytest -q studies/138-random-forest/tests
+
+      - name: Study 139 — AI-Powered-ETF test-suite
+        run: pytest -q studies/139-ai-powered-etf/tests
+
+      - name: Study 140 — Amihud-Illiquidity test-suite
+        run: pytest -q studies/140-amihud-illiquidity/tests
+
+      - name: Study 141 — Turnover-Anomaly test-suite
+        run: pytest -q studies/141-turnover-anomaly/tests
+
+      - name: Study 142 — Split-Drift test-suite
+        run: pytest -q studies/142-split-drift/tests
+
+      - name: Study 143 — Dividend-Capture test-suite
+        run: pytest -q studies/143-dividend-capture/tests
+
+      - name: Study 144 — Permanent-Portfolio test-suite
+        run: pytest -q studies/144-permanent-portfolio/tests
+
+      - name: Study 145 — Home-Bias test-suite
+        run: pytest -q studies/145-home-bias/tests
+
+      - name: Study 146 — Country-Momentum test-suite
+        run: pytest -q studies/146-country-momentum/tests
+
+      - name: Study 147 — FX-Momentum test-suite
+        run: pytest -q studies/147-fx-momentum/tests
+
+      - name: Study 148 — Lunar-Effect test-suite
+        run: pytest -q studies/148-lunar-effect/tests
+
+      - name: Study 149 — Daylight-Saving test-suite
+        run: pytest -q studies/149-daylight-saving/tests
+
+      - name: Study 150 — SAD-Effect test-suite
+        run: pytest -q studies/150-sad-effect/tests
+
+      - name: Study 151 — Stocks-For-Long-Run test-suite
+        run: pytest -q studies/151-stocks-for-long-run/tests
+
+      - name: Study 152 — Inflation-Hedge test-suite
+        run: pytest -q studies/152-inflation-hedge/tests
+
+      - name: Study 153 — Net-Operating-Assets test-suite
+        run: pytest -q studies/153-net-operating-assets/tests
+
+      - name: Study 154 — Leverage-Anomaly test-suite
+        run: pytest -q studies/154-leverage-anomaly/tests
+
+      - name: Study 155 — Asset-Turnover test-suite
+        run: pytest -q studies/155-asset-turnover/tests
+
+      - name: Study 156 — Martingale test-suite
+        run: pytest -q studies/156-martingale/tests
+
+      - name: Study 157 — Kelly-Sizing test-suite
+        run: pytest -q studies/157-kelly-sizing/tests
+
       - name: Smoke-run the offline demos
         run: |
           python studies/01-overnight-anomaly/examples/run_synthetic_demo.py

README.md

@@ -45,7 +45,7 @@ Almost everything ends up bottom-right; two chips are green — and neither one
 ![The bench map — every study placed on a Signal × Tradability grid](docs/bench_map.png)
 
 The counts, the mortality by family of idea, and the five lessons the bench keeps
-teaching are in **[What 137 teardowns taught us](docs/bench.md)**.
+teaching are in **[What 157 teardowns taught us](docs/bench.md)**.
 
 ---
 
@@ -190,6 +190,26 @@ teaching are in **[What 137 teardowns taught us](docs/bench.md)**.
 | **[135](studies/135-fomc-cycle/)** | **FOMC-Cycle** | Do stock returns really hide in the even weeks of the Fed cycle? | ![Weak](https://img.shields.io/badge/Weak-dab617?style=flat-square) | ![Mirage](https://img.shields.io/badge/Mirage-c0392b?style=flat-square) |
 | **[136](studies/136-mark-twain/)** | **Mark-Twain** | Is October really the most dangerous month to own stocks? | ![None](https://img.shields.io/badge/None-c0392b?style=flat-square) | ![Mirage](https://img.shields.io/badge/Mirage-c0392b?style=flat-square) |
 | **[137](studies/137-mansfield-rs/)** | **Mansfield-RS** | Does Weinstein's Stage-2 relative-strength screen beat the market? | ![None](https://img.shields.io/badge/None-c0392b?style=flat-square) | ![Mirage](https://img.shields.io/badge/Mirage-c0392b?style=flat-square) |
+| **[138](studies/138-random-forest/)** | **Random-Forest** | Does a walk-forward Random Forest on price features beat a shuffled-label control? | ![None](https://img.shields.io/badge/None-c0392b?style=flat-square) | ![Mirage](https://img.shields.io/badge/Mirage-c0392b?style=flat-square) |
+| **[139](studies/139-ai-powered-etf/)** | **AI-Powered-ETF** | Can IBM Watson's AI ETF beat a dirt-cheap S&P 500 index fund? | ![Weak](https://img.shields.io/badge/Weak-dab617?style=flat-square) | ![Mirage](https://img.shields.io/badge/Mirage-c0392b?style=flat-square) |
+| **[140](studies/140-amihud-illiquidity/)** | **Amihud-Illiquidity** | Does sorting on price-impact illiquidity earn a premium, or just survivorship? | ![Mixed](https://img.shields.io/badge/Mixed-dab617?style=flat-square) | ![Mirage](https://img.shields.io/badge/Mirage-c0392b?style=flat-square) |
+| **[141](studies/141-turnover-anomaly/)** | **Turnover-Anomaly** | Do low-turnover stocks deliver the Datar-Naik-Radcliffe liquidity premium? | ![Mixed](https://img.shields.io/badge/Mixed-dab617?style=flat-square) | ![Mirage](https://img.shields.io/badge/Mirage-c0392b?style=flat-square) |
+| **[142](studies/142-split-drift/)** | **Split-Drift** | Do stocks drift up after a split takes effect? | ![None](https://img.shields.io/badge/None-c0392b?style=flat-square) | ![Mirage](https://img.shields.io/badge/Mirage-c0392b?style=flat-square) |
+| **[143](studies/143-dividend-capture/)** | **Dividend-Capture** | Can you pocket a dividend by buying just before the ex-date? | ![None](https://img.shields.io/badge/None-c0392b?style=flat-square) | ![Mirage](https://img.shields.io/badge/Mirage-c0392b?style=flat-square) |
+| **[144](studies/144-permanent-portfolio/)** | **Permanent-Portfolio** | Does Harry Browne's four-way mix really survive every regime? | ![Real](https://img.shields.io/badge/Real-2ea44f?style=flat-square) | ![Fragile](https://img.shields.io/badge/Fragile-dab617?style=flat-square) |
+| **[145](studies/145-home-bias/)** | **Home-Bias** | Does diversifying abroad still improve a US investor's portfolio? | ![None](https://img.shields.io/badge/None-c0392b?style=flat-square) | ![Mirage](https://img.shields.io/badge/Mirage-c0392b?style=flat-square) |
+| **[146](studies/146-country-momentum/)** | **Country-Momentum** | Does rotating into the strongest-trending country ETFs pay? | ![Weak](https://img.shields.io/badge/Weak-dab617?style=flat-square) | ![Fragile](https://img.shields.io/badge/Fragile-dab617?style=flat-square) |
+| **[147](studies/147-fx-momentum/)** | **FX-Momentum** | Does currency momentum still pay after costs? | ![Weak](https://img.shields.io/badge/Weak-dab617?style=flat-square) | ![Mirage](https://img.shields.io/badge/Mirage-c0392b?style=flat-square) |
+| **[148](studies/148-lunar-effect/)** | **Lunar-Effect** | Do stocks really return less around the full moon? | ![None](https://img.shields.io/badge/None-c0392b?style=flat-square) | ![Mirage](https://img.shields.io/badge/Mirage-c0392b?style=flat-square) |
+| **[149](studies/149-daylight-saving/)** | **Daylight-Saving** | Does the market really slump after the clocks change? | ![None](https://img.shields.io/badge/None-c0392b?style=flat-square) | ![Mirage](https://img.shields.io/badge/Mirage-c0392b?style=flat-square) |
+| **[150](studies/150-sad-effect/)** | **SAD-Effect** | Do shorter autumn days really depress stock returns? | ![None](https://img.shields.io/badge/None-c0392b?style=flat-square) | ![Mirage](https://img.shields.io/badge/Mirage-c0392b?style=flat-square) |
+| **[151](studies/151-stocks-for-long-run/)** | **Stocks-For-Long-Run** | Do stocks really always win over the long run? | ![Real](https://img.shields.io/badge/Real-2ea44f?style=flat-square) | ![Fragile](https://img.shields.io/badge/Fragile-dab617?style=flat-square) |
+| **[152](studies/152-inflation-hedge/)** | **Inflation-Hedge** | Are stocks really an inflation hedge? | ![Mixed](https://img.shields.io/badge/Mixed-dab617?style=flat-square) | ![Mirage](https://img.shields.io/badge/Mirage-c0392b?style=flat-square) |
+| **[153](studies/153-net-operating-assets/)** | **Net-Operating-Assets** | Does a bloated balance sheet predict low returns? | ![Weak](https://img.shields.io/badge/Weak-dab617?style=flat-square) | ![Fragile](https://img.shields.io/badge/Fragile-dab617?style=flat-square) |
+| **[154](studies/154-leverage-anomaly/)** | **Leverage-Anomaly** | Do high-leverage firms earn the higher return theory promises? | ![None](https://img.shields.io/badge/None-c0392b?style=flat-square) | ![Mirage](https://img.shields.io/badge/Mirage-c0392b?style=flat-square) |
+| **[155](studies/155-asset-turnover/)** | **Asset-Turnover** | Do capital-efficient (high asset-turnover) firms beat the market? | ![None](https://img.shields.io/badge/None-c0392b?style=flat-square) | ![Mirage](https://img.shields.io/badge/Mirage-c0392b?style=flat-square) |
+| **[156](studies/156-martingale/)** | **Martingale** | Does averaging down ("martingale") beat just buying and holding? | ![None](https://img.shields.io/badge/None-c0392b?style=flat-square) | ![Mirage](https://img.shields.io/badge/Mirage-c0392b?style=flat-square) |
+| **[157](studies/157-kelly-sizing/)** | **Kelly-Sizing** | Does Kelly-optimal position sizing beat fixed sizing in practice? | ![Weak](https://img.shields.io/badge/Weak-dab617?style=flat-square) | ![Fragile](https://img.shields.io/badge/Fragile-dab617?style=flat-square) |
 
 > **Click any study** for the full teardown — two narrative notebooks (one for the curious,
 > one for the quant), reproducible code, and every number behind the two stamps.