Backtest methodology
Everything on the strategy pages is hypothetical. It describes what a published rule would have done against the disclosure record — not trades anyone placed. Methodology version v1.0.0, prices as of 2026-08-27.
The rules
What a strategy is
A strategy is a deterministic rule over disclosure events. There are two, and neither is ours to tune:
- Per-politician replication. Enter on each of that member’s disclosed purchases that names a ticker. A disclosed sale closes every open position in that ticker.
- Cluster buy. Enter when two or more insiders at the same company disclose open-market purchases within a week of each other.
The conventions
How a position is priced
- Entry — the first close after the disclosure becomes public. Not the transaction date, which can be up to two and a half years earlier, and not the disclosure date’s own close. A follower reading a filing on the day it appears cannot transact at that day’s close, so the entry prices at the next session’s close. Where a filing lands on a non-trading day the entry is simply the next session, with no extra penalty.
- Exit — four paths, all deterministic. A disclosed sale closes the position at the first close after that sale becomes public, on the same one-session lag as the entry. A position with no disclosed sale is marked at the latest close in our price record. If a symbol’s price history ends before that date the position is closed at its last available close and labelled forced. And where a sale is disclosed on the same day as the purchase — filers routinely disclose both ends of a round trip at once — the position never opens, because nobody could have held something already closed by the time it became public.
- Assets — exchange-priced equities and funds only. We replicate ETF, REIT, Stock disclosures. Options, corporate and municipal bonds, preferred shares and mutual funds are excluded from entries and never substituted with a proxy, an index leg or a cash placeholder. They still appear in the trade log, stating that the rule did not act on them — a disclosure we cannot replicate is reported, not hidden.
- Sizing — one unit of capital per entry, and no rebalancing. Each position gets an equal notional amount at its own entry price; nothing is sold to fund a later one, and weights are never reset afterwards. Disclosed amounts are brackets, never position sizes, so they are read for display only and never guessed into a weight.
- The series is time-weighted. Capital entering or leaving the book never creates a return. Money that leaves on an exit stops contributing from that day; the engine holds no cash and models no interest on it.
- Benchmarks — SPY and RSP, both shown. SPY is the cap-weighted S&P 500; RSP is the same 500 companies weighted equally. Both are measured over the strategy’s own window — from its first entry to the as-of — because comparing two different windows is not a comparison. A rule that equal-weights its own entries shown only against a cap-weighted index confuses which companies it picked with how it weighted them; showing both separates the two.
- A missing price is an error, never a guess. If a symbol has no close on a day a position is open — while that symbol is still trading on either side of the gap — the run fails loudly and the rule is reported as one we could not price. Prices are never carried forward and never interpolated. A stated absence is the honest answer; a spanned gap would be an invented price wearing a real one’s clothes.
- A delisted holding is carried to its last available close, and the page names the date. When a company is acquired, delisted or renamed, its price record stops and never resumes. The position is closed at that final close, the figure is carried to it, and both the summary and the trade log state which holding stopped and on what date — the same sentence a disclosure table uses for the same situation. The figure is never hidden and never set to zero: a zero would be a claim that the position went nowhere, and an omission would quietly shorten the record. What the figure then measures is the period up to that date rather than up to today, which is why the date is printed beside it.
- Clusters enter on their last filing. A cluster is not fully public until the last of its constituent Form 4s is filed, so that is the date its entry prices from — not the transaction date the cluster is anchored to.
The limits
What these figures do not include
- No fees, commissions, spreads, taxes or slippage are modelled. Every figure is gross of all of them. Real execution costs money and these figures do not.
- Prices are split-adjusted but not dividend-adjusted. Neither the strategy nor the benchmark reinvests dividends, so both are understated — and a high-yield holding is understated more than the benchmark is.
- The history is short, and each page says how short. Entries can only begin once a disclosure exists to act on, and the disclosure record — not the price record — is what limits this. Most rules here hold a few weeks; some hold a single day. Every strategy page states its own span next to its figures, and none of them is long enough to be a track record.
- Disclosed amounts are ranges. Filers report brackets like “$1,001 – $15,000”, so no position size here reflects what anyone actually held.
- Disclosure is late by law. Congressional filings arrive up to 45 days after the transaction, and often later. The entry date reflects when the public could act, which is frequently long after the trade.
- No shorting. A disclosed sale closes positions; it never opens a short one.
- Sources and cadence. Disclosures come from the House Clerk and Senate eFD records and the vendor feed that mirrors them; insider purchases from SEC Form 4 filings; prices from end-of-day closes. All are ingested nightly.
Required disclosure
Hypothetical performance
Hypothetical performance results have inherent limitations. They do not represent actual trading and may not reflect the impact of material economic and market factors on decision-making. No representation is made that any account will or is likely to achieve profits or losses similar to those shown. Past performance, hypothetical or actual, is no guarantee of future results. Nothing on these pages is investment advice or a recommendation to follow any rule described.
The per-transaction excess-return figure used elsewhere on the site is a different measurement with its own page: methodology.