The standard a study has to meet
Five commitments, and the last is the one that costs
something.
- A drawdown beside every return.Never a growth
figure on its own, and never in a footnote.
- Costs stated, and stated as zero.No commission
or slippage is modelled, so every edge figure is a ceiling. Where it
matters, the cost that would erase the edge is given too.
- A benchmark that fits, or none.A distribution
measured on one strategy can't score another, so Restrike refuses the
comparison, and most positions in the tracker are graded against your
own rule.
- Generated figures, not typed ones.Every backtest
figure is generated from the data at build time, apart from the
recommended sizing band, which is a judgement. A figure typed into a
sentence is checked against the data.
- Corrections stay published.The pricing
calibration was wrong and was fixed in public. A cost claim was reasoned
out instead of measured, missed by three points, and its retraction is
still on the page that carried it.
How to apply the same tests to anybody else's
backtest, including the one about frequency that most timing claims
fail.
The work itself
One study, one strategy family.
Everything measured here is written against an S&P 500 index
fund: monthly at-the-money short puts, the same position run three
ways, covered calls, both together, and how every contract in the sample
was priced.
- Selling options against an S&P 500 index fundHow often a cycle loses and how much premium survives, how large a position a fall leaves standing, whether rolling or laddering helps, what a covered call costs, and how every contract was priced.
- The betas this tool usesNot part of the study above. Every cushion figure divides by a blended beta: where yours comes from, why it's measured rather than looked up, and what the blend leaves out.
What has not been studied here
The list matters as much as the one above it.
Every other strategy Restrike tracks. Wheels, spreads, straddles
and strangles, earnings trades, anything you devised: the tracker scores
them on what they did in your account, and there's no thirty-year study
behind any of them. That's why it won't compare one against a
benchmark.
Any tenor other than about a month. The pricing calibration was
solved against a published one-month index, and there's no equivalent at
45 or 60 days. Longer-tenor results exist and are deliberately
unpublished.
Individual stocks. A short put on one company carries gap risk an
index doesn't, and nothing here has measured it.
Tax. The rules are described on the study page with sources, and
nothing computes a figure for anybody. A professional review was
declined, so applying the rules to your situation isn't something this
site does.
Anything since the export. The runs end at a fixed date, and the
sample holds one 2008 and one 2020. Two crises in thirty years is not a
large sample of crises, however many cycles it contains.