Trending Value is the top-performing strategy in James O'Shaughnessy's 50-year backtests. The logic: value alone does not work because many cheap stocks are cheap for good reason. But cheap stocks with momentum, stocks that are undervalued and the market has started to notice, consistently outperform. The screen runs in two steps. First, rank the entire market by Value Composite Two, the average percentile rank across six valuation metrics. Second, take the cheapest decile and sort by six-month price return.
This guide walks through the strategy step by step: computing the composite, applying the momentum sort, comparing it to single-metric screens, adding quality filters, and backtesting. For the screener overview and comparison table, see the Trending Value Screener.
Step 1: Compute the Value Composite Two
Value Composite Two ranks every stock on six valuation measures. For each metric, every stock gets a percentile rank from 0 (cheapest) to 100 (most expensive). The composite is the average of the six ranks. A company in the bottom decile (average rank below 10) is cheap on most measures simultaneously.
The six metrics: price-to-book, trailing P/E, price-to-sales, price-to-cash-flow, and EV/EBITDA are all ranked lower-is-cheaper. Shareholder yield (dividends plus net buybacks as a percentage of market cap) is ranked higher-is-better. Stocks with negative earnings or negative free cash flow are excluded because their P/E and P/CF are undefined.
"Compute Value Composite Two for all US common stocks above $200 million market cap. Percentile-rank each stock on six metrics: price-to-book (lower = cheaper), trailing P/E (lower = cheaper), price-to-sales (lower = cheaper), price-to-cash-flow (lower = cheaper), EV/EBITDA (lower = cheaper), and shareholder yield (dividends + net buybacks from the last 4 quarters / market cap, higher = better). Exclude stocks with negative P/E or negative cash flow. Average the six percentile ranks. Show the 25 cheapest stocks with their composite score and each individual metric."
The composite avoids single-metric traps. A stock with a P/E of 5 might look cheap, but if its P/B is 3x, its EV/EBITDA is 15x, and it pays no dividend and does no buybacks, it is not broadly cheap. The composite catches the difference. Expect the cheapest decile to be concentrated in financials, energy, and industrials, with occasional consumer discretionary and materials names.
Step 2: Apply the momentum sort
The composite gives you the cheapest stocks. The momentum sort gives you the cheapest stocks that are moving. Take the bottom decile (or quintile for a smaller universe like large caps) and sort by six-month price return, highest first. The top 25 to 50 stocks are the Trending Value portfolio.
"From the Value Composite Two results (average percentile rank across P/B, trailing P/E, P/S, P/CF, EV/EBITDA lower-is-cheaper + shareholder yield higher-is-better), take the cheapest decile of US stocks above $200 million. Sort by 6-month price return, highest first. Show the top 25 with VC2 score, each metric, sector, and 6-month return percentage."
This is where the strategy earns its returns. O'Shaughnessy found that the cheapest decile alone returned about 17% annually over his 50-year sample. Adding the momentum sort pushed that to over 21%. The momentum filter removes stocks that are cheap because they are collapsing, which is the single biggest source of underperformance in pure value strategies.
Step 3: Compare VC2 to single-metric value screens
The composite exists because no single metric works in isolation. P/E ignores asset values. P/B ignores earnings quality. EV/EBITDA misses capital return programs. Each metric captures one dimension of cheapness and misses the others. By averaging all six, the composite identifies stocks that are cheap from multiple angles.
"For the top 25 Trending Value stocks (cheapest decile by VC2, average percentile of P/B, P/E, P/S, P/CF, EV/EBITDA lower-is-cheaper + shareholder yield higher-is-better, sorted by 6-month return), also show each stock's individual percentile rank on each of the six metrics. Flag any stock that is in the cheapest quartile on at least 5 of the 6 metrics."
Stocks flagged as cheap on 5 or 6 metrics are the strongest composite value plays. Those cheap on only 2 or 3 qualified because they were extremely cheap on those metrics, which may signal a sector-specific distortion rather than broad undervaluation. Both can be valid, but the distinction matters for your thesis.
Step 4: Add quality filters to the Trending Value output
O'Shaughnessy's original screen does not filter by quality beyond requiring positive earnings and cash flow. You can layer additional checks. A Trending Value stock with a Piotroski F-Score of 7 or above is both cheap and financially improving. One flagged by the Beneish M-Score may have unreliable financials despite looking cheap.
"Run the Trending Value screen (cheapest decile by VC2, average percentile of P/B, trailing P/E, P/S, P/CF, EV/EBITDA lower-is-cheaper + shareholder yield higher-is-better, sorted by 6-month return) for US stocks above $500 million. For each stock in the output, also show the Piotroski F-Score. Filter to F-Score 6 or higher. This is value + momentum + quality."
Adding Piotroski to Trending Value filters out about half the cheapest decile, stocks where the financial health does not support the valuation thesis. What remains are cheap, improving companies with momentum. This three-factor combination was not in O'Shaughnessy's original book but follows the same logic: layer a quality signal on top of value to remove the worst outcomes.
Step 5: Backtest Trending Value historically
The final step: does the screen actually work on data it was not designed on? Shibui has daily valuation metrics back to 1993 and quarterly financials back to 1990, so you can run the Trending Value screen at any historical date and measure what happened next.
"Run the Trending Value screen (cheapest decile by VC2 using P/B, trailing P/E, P/S, P/CF, EV/EBITDA lower-is-cheaper + shareholder yield higher-is-better, sorted by 6-month return) as of January 2, 2020, for US stocks above $200 million. For each stock in the top 25, compute the 12-month forward return. What was the median return? How many beat the S&P 500?"
Important caveats. Survivorship bias applies: companies that went bankrupt or were delisted between 2020 and 2021 are not in the dataset, so the worst outcomes are missing. The backtest does not account for trading costs, bid-ask spreads (which are wide for small caps), or the difficulty of actually executing 25 trades on rebalance day. O'Shaughnessy's original backtests used a different data source and methodology. Shibui backtests and O'Shaughnessy backtests should not be compared directly.
What Shibui cannot do
Shibui cannot replicate O'Shaughnessy's exact backtests. His data covered 1926 to 2009 using the Compustat Point-in-Time database, which includes delisted companies. Shibui's data starts in 1990 and only covers currently listed companies. The strategy's historical track record cannot be verified from this data alone. Use Shibui to run the screen forward, track it over time, and compare results to your own benchmarks. For the screener overview and comparison table, see the Trending Value Screener.