The short-dated CHEAP and RICH cards: do they beat trading every name in the list?
When the options cards mark a same-day, next-day or weekly contract as cheap (implied volatility at or below realized) or rich (at least 1.10 times realized), does buying the cheap contract or selling a 1:4 credit spread on the rich one beat the same trade on every name in the list, after costs?
No. None of the six pre-registered trials passed, and every lower bound is below zero. Buying the list's cheap contract did no better than buying the same kind of contract on every name in the list. Same-day expiries: -3.9% a trade against -8.1% for all names, a +4.2-point difference with a one-sided lower bound of -0.6. Next-day: +5.1% against +5.2%, -0.1 point (-9.2). Weekly: +17.0% against +19.5%, -2.5 points (-15.3). The weekly gain is a 2023-24 rally that the all-names trade shared, not the ratio's: the median weekly buy lost its whole premium. Selling the list's rich 1:4 credit spread won 68% to 72% of trades, as a 1:4 spread is built to, but its average was no better than selling on every name: same-day -1.6% against -0.5% (-1.1 points, -19.5), next-day +4.6% against -2.6% (+7.3 points on 221 trades, -21.8), weekly -3.5% against -4.6% (+1.1 points, -10.3). Both positive rich differences rest on one session, 5 August 2024: without that session the next-day rich mean is -3.5% and the weekly rich mean is -5.2%. Most rich spreads used the nearest out-of-the-money fallback (90 of 117, 187 of 221, 1,108 of 1,365). The test covers 33 of the 35 names in the list from January 2023 to August 2024; the other two had no listed options in the window. The holdout stays sealed.
Backtested by QS Research
Pre-registration
Six trials were fixed before any quote was read, on the 35 names of the app's momentum list: buying a cheap contract and selling a rich 1:4 credit spread, each on same-day, next-day and weekly expiries (B0, S0, B1, S1, BW, SW), judged at a family of six with a one-sided 0.05/12 bound. At 10:00 ET each expiry is read from its calls and puts: at-the-money implied volatility is divided by the lower of one-year and three-month realized volatility for cheap (at or below 1.00) and by the higher for rich (at or above 1.10), with earnings reaction days removed. The side rule buys the cheaper 25-delta wing and sells the richer one when their implied volatilities differ by more than 2 points; otherwise it buys the at-the-money call and sells the 25-delta with the higher implied volatility. A cheap contract is bought at the ask plus $0.0065 a share and held to expiry; a rich contract is sold at the bid with the nearest further-out-of-the-money strike paying at least 1:4 of credit to maximum loss bought at its ask, both legs paying the commission, and when that strike has no such wing the nearest out-of-the-money strike on the same side that has one is sold instead. Trades settle at intrinsic on the expiry's official close, and a trade with a split in its life is left out. Every priced trade in the tenor counts in the baseline, whatever its ratio. A trial passes only if its mean return beats that baseline, the lower bound of a weekly-block bootstrap of the difference (10,000 resamples of entry weeks) is above zero, and its own mean is above zero. Entries run from 3 January 2023 to 30 August 2024, and a name enters a tenor only from the first session that tenor's expiry was listed. The 120 sessions after 9 November 2026 are a sealed holdout; a descriptive re-look window from September 2024 was listed in the pre-registration but not run.
Registered 2026-11-09Amendments
- 2026-11-20 Recorded before any quote was read: how the written rules are read where they leave a choice (the quote-sanity rule only, every priced trade in the baseline, no upper limit on the credit spread's wing), the listing start dates for each name and tenor, and the content hashes of the selection. No window, name, threshold, trial, fill, cost or pass rule changed. approved: Henry's rulings, 2026-11-20, recorded before the fetch
- 2026-11-20 Pinned the option quotes and expiry listings as one snapshot before the counted run, and recorded the fetch counts and that the rebuilt selection matched the committed files byte for byte. No rule changed. approved: Recorded before the counted run, before any quote value was read
- 2026-11-21 After the counted run: corrected the pre-registration's prior-look note. A companion study on the same 2023-24 window, run on 2026-11-12, had read the 10:00 quotes for SPY, QQQ and IWM for an at-the-money straddle test and found no profitable rule; this study's rules were approved on 2026-11-09, before that run. No window, name, threshold, trial, fill, cost or pass rule changed. approved: Recorded at the independent review, 2026-11-21; it changes no rule
A holdout sample is sealed: its results stay unread until the study is final.
Trials · 6 trials · 0 passed · judged at N = 6 trials; 2965 trials run in total
| Trial | Verdict | |
|---|---|---|
| B0 · same-day expiry · CHEAP (ratio ≤ 1.00), buy | FAIL | counted |
| S0 · same-day expiry · RICH (ratio ≥ 1.10), 1:4 credit spread | FAIL | counted |
| B1 · next-day expiry · CHEAP (ratio ≤ 1.00), buy | FAIL | counted |
| S1 · next-day expiry · RICH (ratio ≥ 1.10), 1:4 credit spread | FAIL | counted |
| BW · weekly expiry · CHEAP (ratio ≤ 1.00), buy | FAIL | counted |
| SW · weekly expiry · RICH (ratio ≥ 1.10), 1:4 credit spread | FAIL | counted |
Limitations
- The 35 names are a list picked in November 2026 from names still listed, so the study carries survivorship bias. The app's live universe is the 400 most-traded names, which cannot be rebuilt for these years, and names that later delisted are missing.
- Two of the 35 names (IBIT and SNDK) had no listed option expiry in the window and contribute no trade; 33 names do.
- Only Friday expiries (or the Thursday when Friday is a market holiday) were listed for 26 of the 27 names outside the six funds with listed options, so their same-day and next-day trials trade one day a week: a Friday entry, or a Thursday entry into the Friday expiry; VST lists only 31 Friday expiries, so it trades far less often. The funds list more expiries, so the pooled result mixes a daily sample with a weekly one.
- Fills are the first usable quote between 10:00 and 10:05 ET: buys pay the ask and sales receive the bid, with $0.0065 a share on each leg. Trades are held to expiry and settle at intrinsic on the official close; there is no early exit, no interest, no assignment or early-exercise friction and no position-size limit.
- Returns are on premium. A long option that expires worthless loses 100% of its premium, so the averages are carried by a few large winners: the median trade in all three cheap trials lost its whole premium.
- A rich spread's return is profit over the maximum loss, and the credit has no upper limit, so a spread sold for close to its width can return many times its risk when it wins.
- The rich averages are tail-driven. The next-day rich mean (+4.6% a trade) is -3.5% without its single best spread, which returned 17.9 times its risk on 5 August 2024; the weekly rich mean is -3.5% and -5.2% without that session. Five of the weekly trial's best trades and the next-day trial's best trade are all 5 August 2024, a market-wide volatility spike. The win rate is high because a 1:4 spread is built that way; the averages are not broad.
- A rich spread also needs a 1:4 wing. Sessions with none are left out of the rich comparison but stay in the cheap one: 1,601 of the 3,449 priced same-day name-sessions (46%), 901 of 3,708 next-day (24%) and 1,691 of 13,356 weekly (13%) are dropped, and they are the lowest-premium sessions, so the rich comparison is a selected, higher-premium half of the same-day sample.
- Trades whose life spans an earnings release are kept (603 of the 1,365 weekly rich trades); earnings dates are removed only from the realized-volatility read.
- A companion study on the same 2023-24 sessions, run eight days before this one, had already read the 10:00 quotes for SPY, QQQ and IWM for its own at-the-money straddle test, and found no profitable rule either. Those three names are 31% of the same-day cheap trades, 54% of the same-day rich trades, 29% of the next-day cheap trades and 48% of the next-day rich trades (9% of each weekly trial); excluding them does not change the verdict: the point differences are no larger (+2.7, -4.3, -1.0, +3.5, -3.0 and +0.7 points) and the next-day rich mean turns negative. The six rules here were approved three days before that study ran, so none was chosen on its result; the pre-registration's note that it had not yet run was true when written and stale at the run.
- The at-the-money strike is the nearest quoted strike; when that one strike lacks a call or a put, the session is refused rather than moving to the next strike that has both, which is narrower than the pre-registration's wording. 614 name-sessions are refused that way (357 same-day, 97 next-day, 160 weekly); no ratio, trade or return was computed for them.
- The bounds are wide. With 87 weekly blocks, the same-day cheap trial's +4.2-point edge has a -0.6-point lower bound, so a small edge is not ruled out; the study shows these six rules did not clear the pre-registered bar, not that no version of them could.
- The descriptive re-look window (September 2024 to October 2026) was not run: no quotes were fetched for it. The 120 sessions after 9 November 2026 stay sealed.
- While the pricing code was being built, a few quotes from one counted-window session were printed to check the data format; no ratio, trade or return was computed before the run.
- The result is tied to one pinned snapshot of option quotes, expiry listings, daily closes and earnings dates; a later or different snapshot was not tested.
The numbers: every trial's checks, metrics, curve and yearly results, and the provenance
2 published versions · current cd143659e2a96339, published 2026-11-21