Skip to content
bellwether

Lab & evidence

Backtests versus the track record, and what does and does not count as proof.

9 articles

The Lab is a workbench: stress tests on the live book, queued backtests, replays. It changes no setting and it proves less than it looks like it proves.

  1. Backtests versus the track recordWhat the Lab's four tabs do, how queued backtests and grids of settings carry their own reasons to doubt them, and why the live record on Performance is the number that counts.Everyone on the book7 min read
  2. What counts as proofAn in-sample backtest is not proof. What is: out-of-sample and walk-forward results, deflated Sharpe, costs and survivorship taken seriously, a paper record over time, scored calls, and a readiness scorecard.Everyone on the book4 min read
  3. Where the Lab's prices come fromThe stress tests, replays and projections are computed from twenty years of real daily prices, a fixed price history that ships with each release; the corner label says which prices and where they end, and a copy of Bellwether without them says so instead of guessing.Everyone on the book3 min read
  4. Evidence before moneyTwo honest halves (a deterministic backtest report and the forward-only live track record), plus what each does and does not prove, and how to export, print or email them.Everyone on the book7 min read
  5. Reading the Evidence tabThe Evidence tab of Performance separates what is proven about the desk today (the machinery) from what only time can prove (the live record), with the sample and date behind every figure and a plain list of what none of it proves.Everyone on the book9 min read
  6. Sharpe ratio: what it is, why we chase it, and why one good year proves nothingThe desk's scoreboard and goal (return over cash per unit of bounce): how Bellwether measures it with an honest interval, what a Sharpe of 1.0, 1.5 or 2.0 actually feels like in a year, and why it is never one of the signals.Everyone on the book9 min read
  7. Luck or skill? Reading the section on PerformanceFour readings the platform computes every night to say, honestly, whether this book's returns can be told from luck yet, how each one is made, and why the answer after a handful of trading days is always "too early".Everyone on the book10 min read
  8. What the track record does and does not proveEvery book keeps one sealed row per trading day in a tamper-evident chain, and a tear sheet computed from those rows alone. What that record can back up for an outsider, what it cannot, and why it starts on paper today.Everyone on the book6 min read
  9. How to read \"vs SPY\": the book against the index, after AI costWhat the row under the Overview figures and the Against SPY section on Performance compare, why the gap is in points, what "after AI cost" and "before tax" mean, why a young gap prints in plain type, which close it is dated by, and what the Discovery tile counts.Everyone on the book7 min read