Home/Learn/Backtests/Quality at a fair price

SledgeKey Backtests

We ran a quality at a fair price screen honestly. It gave up 39 percent of the upside to avoid 7 percent of the downside.

Published September 2, 2026 · By The SledgeKey Team · Window September 4, 2016 to September 2, 2026

Buying good businesses at sensible prices is the most repeated advice in modern value investing, and it is easy to state as a screen: high return on equity, healthy operating margins, a price-to-earnings ratio that is not stretched, and free cash flow that is actually positive. We ran exactly that screen once a year for ten years, on the universe as it stood on each rebalance date, and let it hold whatever it found. It returned 12.46 percent a year. The S&P 500 returned 15.15 percent.

The gap of 2.69 points a year is the least interesting part of the result. What the ten years actually show is a portfolio that behaved defensively in the wrong direction. It participated in 93 percent of the index's down moves and only 61 percent of its up moves. A quality tilt is usually bought as insurance, and insurance that expensive is worth looking at closely.

The exact screen

Every threshold below was set in the SledgeKey screener and run without further adjustment. The screen matched 327 companies on the most recent rebalance date, so the twenty-name portfolio was never short of candidates.

Two of those rows deserve a note. The positive free cash flow requirement is expressed through the screener's Quality filter, which also imposes a $50M market capitalization floor; that floor sits well below the $1B minimum already applied, so it changes nothing here. The liquidity floor removes securities that do not trade in meaningful size on a typical day, measured as a median rather than an average so that a single block trade cannot make a dormant listing look tradeable.

The result against the index

The screen was genuinely calmer. Its worst peak-to-trough fall was 29.1 percent against the index's 33.7 percent, and its annual volatility was two points lower. Anyone who bought a quality tilt for a smoother ride received one. The question is what that smoothness cost, and the capture ratios answer it more clearly than the headline return does.

Where the defense actually showed up

Capture ratios describe how much of the benchmark's movement a portfolio inherits, separated by direction. An upside capture of 61.25 means the screen rose about 61 percent as much as the index did in the index's up months. A downside capture of 93.07 means it fell about 93 percent as much in the down months. The portfolio therefore surrendered close to two fifths of the gains while avoiding well under a tenth of the losses.

That asymmetry is the substantive finding here. The 2.69 point annual gap is a consequence of it rather than a separate fact about the screen. The chart shows the same thing without any statistics. The two lines run close together through 2017, drift apart during 2018 and converge again over 2019, and the screen falls appreciably less in the crash of early 2020, which is the moment a quality tilt earns its reputation. The separation that lasts opens in 2021. The screen ended that year almost ten percentage points further behind the index in cumulative terms than it began it, and a second leg during 2024 widened the shortfall again. By the close of the window the screen trailed the index by 21 percent cumulatively, and roughly two thirds of that shortfall was already in place before 2024 began.

A plausible reading is that a P/E ceiling of 25 excluded the companies that drove most of the index's return in the second half of the window. The screen was not wrong about those businesses being expensive. It was simply unwilling to own them, and over this particular decade that unwillingness was costly. A different decade could reverse the conclusion, which is the whole reason for stating the window in every sentence that quotes a number.

What this backtest does not prove

One ten-year window is a single observation. The dispersion statistics on this page, including volatility, Sharpe, Sortino, capture ratios and tracking error, move as the window moves at either end, so they should be read as a description of this decade rather than as constants belonging to the strategy. The stable figures are total return, annualized return, final value and the daily-basis maximum drawdown.

The run assumes annual rebalancing at a fixed date, equal weights, a 10 percent position cap and 10 basis points of cost per trade. It does not model taxes, slippage beyond that flat cost, or the behavior of a real person watching a portfolio trail its benchmark for two consecutive years. Companies that delisted during the window are kept in the universe and booked out at their last traded price, which is what keeps the result honest and is the reason it is lower than the same screen would show on a survivor-only universe.

The screen itself is a reasonable expression of a popular idea, and it is not the only one. Raising the P/E ceiling, ranking on quality rather than filtering on it, or rebalancing more often would each produce a different portfolio. Those are separate backtests, and the configuration table above exists so that anyone can reproduce this one before arguing with it.