Study 11: We Built a Value-Screened Options Strategy. The Screen Made It Worse.
We ran bull put spreads on stocks with positive margin of safety and growing earnings. Sharpe 1.980, 89% win rate. Then we removed the value screen. Sharpe went to 2.782. Here's exactly why.
The Hypothesis
Sell put spreads only on stocks that pass a value screen — positive margin of safety, growing earnings. Value stocks tend to outperform over time. If they spend more time going up than down, short puts on those stocks should outperform short puts on random stocks.
It’s a clean idea. It also turned out to be wrong.
The Strategy
Universe: 56 liquid large-caps — AAPL, MSFT, GOOGL, JPM, GS, BRK-B, UNH, and similar. Options bid-ask $0.01–0.05, friction ~$5–10/contract.
Entry screen (both conditions required):
- MOS positive: earnings yield (1/PE) exceeds the long-run T-bill rate (4.5%). The stock earns more than bonds.
- EPS growing: trailing twelve-month EPS is higher than twelve months prior.
Structure: Sell a 5% OTM put, buy a 10% OTM put, same expiry (~30 DTE). Net credit received. Max loss = spread width minus credit.
Exit: 50% profit target — close when the spread’s mark-to-model value drops to half the entry credit. Otherwise hold to expiry.
Look-ahead prevention: Fundamental data uses SEC filing dates, not quarter-end dates. A company reporting Q3 results on Oct 15 is not visible to the algorithm until Oct 15. The gate is one auditable line:
sub = pit_fundamentals[
(pit_fundamentals["ticker"] == ticker) &
(pit_fundamentals["timestamp"] <= sim_date) # timestamp = SEC fillingDate
]Mean lag between quarter-end and filing date: 34.4 days. This is the difference between a clean backtest and one that cheats.
Results: 2018–2024, 4,509 Trades
| Metric | Value |
|---|---|
| Sharpe ratio | 1.980 |
| Win rate | 89.2% |
| Mean return per trade | +3.95% on max risk |
| Total net P&L | $103,157 |
| Profit-target exits | 88.8% of trades |
| Avg hold (profit target) | 13.8 days |
| Avg hold (expiry) | 33.8 days |
The 50% profit target explains the win rate. Profit-target exits win 100% of the time — you crystallize the gain before anything can go wrong. Expiry-held positions win 3.4% of the time, with an average loss of −$325/contract.
Exit breakdown:
| Exit Reason | Trades | Win Rate | Mean Net P&L |
|---|---|---|---|
| Profit target (50%) | 4,005 | 100% | +$66.65 |
| Held to expiry | 504 | 3.4% | −$324.97 |
This asymmetry — lots of small wins, occasional large losses — is the defining characteristic of short options strategies. The tail is left-skewed (−2.44) and fat (excess kurtosis 5.54). When a position blows through both strikes, the loss is close to the full spread width. It happens 11.2% of the time, mostly in rate-shock environments.
By year:
| Year | Sharpe | Win Rate | Net P&L |
|---|---|---|---|
| 2018 | −2.744 | 80.6% | −$5,596 |
| 2019 | +6.312 | 96.3% | +$23,128 |
| 2020 | +3.007 | 87.3% | +$10,706 |
| 2021 | +4.398 | 92.7% | +$37,108 |
| 2022 | −0.992 | 77.7% | −$20,444 |
| 2023 | +2.167 | 90.4% | +$14,165 |
| 2024 | +4.391 | 92.5% | +$44,091 |
Two losing years: 2018 and 2022. Both were Fed tightening cycles. The screen was passing financial stocks — they have low P/E, so earnings yield is high, so they always pass the MOS gate. Rising rates compress P/E multiples. The screen was loading the portfolio into exactly the stocks most exposed to rate repricing.
The Placebo Test
We ran the same backtest on the same 56 tickers with the value screen removed. Every qualifying spread, every day, no fundamental filter.
| Screened | Unscreened | |
|---|---|---|
| Trades | 4,509 | 52,133 |
| Win rate | 89.2% | 89.6% |
| Sharpe | 1.980 | 2.782 |
| Mean return | 3.95% | 5.82% |
| Total net P&L | $103,157 | $868,365 |
| Profit exits | 88.8% | 89.2% |
The unscreened version is better on every metric. The value screen rejected 94% of eligible trades — and the trades it rejected were the better ones.
Year-by-year comparison:
| Year | Screened | Unscreened | Edge |
|---|---|---|---|
| 2018 | −2.744 | −0.066 | −2.678 |
| 2019 | +6.312 | +5.749 | +0.563 |
| 2020 | +3.007 | +4.649 | −1.642 |
| 2021 | +4.398 | +3.641 | +0.757 |
| 2022 | −0.992 | −0.027 | −0.966 |
| 2023 | +2.167 | +4.806 | −2.639 |
| 2024 | +4.391 | +4.266 | +0.126 |
The screen outperformed in exactly two years: 2019 and 2021. In 2018, 2022, and 2023 the gap was substantial and went the wrong way. The screen added concentration, not edge.
Why the Screen Hurt
What the screen concentrated into: The top screened tickers by trade volume were MS, BRK-B, JPM, GS, WFC, C, BAC — banks and brokerages. They dominate because their P/E ratios are structurally low, which means their earnings yields are structurally high, which means they almost always pass the MOS gate.
What the screen kept out: High-P/E quality compounders — MCD, HD, MSFT, DHR, ACN, LLY. These don’t pass an earnings yield screen because the market prices them at premium multiples. But they also have stable, mean-reverting price action that is ideal for selling put spreads. The algorithm excluded them.
The rate sensitivity mechanism: The MOS screen requires PE < 1/0.045 ≈ 22x. When rates rise, that bar shifts — what was cheap at a 4.5% T-bill rate is not cheap at 5.5%. Financial stocks (which dominated the screened universe) are most directly repriced in a rate increase because their book value and earnings are rate-sensitive. In 2018 and 2022, the screen produced a portfolio of financial stocks in a rising-rate environment. The losses followed directly from that.
The unscreened version held all 56 tickers simultaneously. It was diversified across sectors and business models, so rate repricing hit a smaller fraction of the portfolio.
What the Data Actually Supports
The strategy works. Sharpe 2.782, win rate 89.6%, consistent across both halves of the 2018–2024 period. The edge is:
1. Short-dated OTM bull put spreads on liquid large-caps. The defined-risk structure and liquid options market keep friction manageable. The short put captures implied volatility premium. The long put caps the loss.
2. The 50% profit target. This is the most important decision rule in the entire strategy. Taking profits at 50% max gain cuts the average holding period to 13.8 days. You’re getting most of the available premium in less than half the time, which dramatically improves return on margin and reduces the number of positions that get caught in sudden drawdowns.
3. No fundamental screen. The value screen does not add alpha in this context. It concentrates the portfolio into a specific factor (value/low-PE stocks) that happens to underperform in rate-rising environments, which are precisely the environments most dangerous for put sellers.
What the MOS screen was doing — selecting businesses that pass a quality threshold — is a valid equity investing framework. It doesn’t translate to an options premium strategy where the entry timing is daily, holding period is 14–34 days, and the primary risk is short-term price movement, not long-term business quality.
Known Limitations
Survivorship bias: The 56-ticker universe is today’s large-cap survivors. Companies that delisted or were removed from the S&P 500 between 2018–2024 are excluded. Estimated win-rate inflation: 3–8%.
IV proxy: We use HV30 × 1.20 as an implied volatility estimate. Real options will trade at different IVs depending on put skew, earnings cycles, and current vol regime. In high-skew environments (VIX CRISIS), the short put is worth more than the model estimates.
Concentration risk in the unscreened version: 52,133 trades on 56 tickers over 7 years means ~74 active positions per day on average. This is only practical with a capital base large enough to hold 30–50 concurrent spreads. At $385 max risk per contract, running 30 concurrent positions requires ~$11,550 in committed margin.
Beta exposure: Correlation of per-trade returns with same-day SPY: 0.346. The strategy is long equity market beta. In a broad selloff, positions go wrong simultaneously.
The Practical Summary
Run bull put spreads on liquid large-cap ETF options or individual large-caps. No fundamental screen. 5% OTM short put, 10% OTM long put, ~30 DTE. Take 50% of the available credit and get out. Repeat.
The structure does the work. The exit rule does the work. A value screen, applied as an entry filter for a 14-day holding period, adds sector risk without adding return.
Backtest methodology: 2018–2024, 56 liquid large-cap tickers, synthetic options pricing via Black-Scholes, HV30 × 1.20 as IV proxy, 5%/10% OTM bull put spreads, 30-DTE target expiry, 50% profit target, filing-date-gated fundamental data (FMP, mean lag 34.4 days). Commission $1.30/leg × 4 legs = $5.20/contract. Max 30 concurrent positions. Placebo backtest: identical mechanics, no fundamental filter, same universe and period.