The full ledger · updated every session · misses stay
Every call the desk publishes gets graded by the same pipeline that published it — the morning regime call, every signal in the feed, the levels themselves. Nothing here is hand-edited, and nothing gets quietly removed. Check our work for a few weeks before you trust us with anything.
A call that simply said "expansion" every single day would still post a high hit rate during a volatile stretch — without predicting anything. So the only number we treat as real is skill: our hit rate minus the base rate a constant call would have scored. Skill near zero means no edge yet, however good the raw hit rate looks.
We mark the scorecard PROVISIONAL until it clears a practical pre-launch bar: at least 20 graded sessions, at least 5 expansion and 5 compression calls, and positive skill versus the majority-outcome baseline. “Published” means the track record is reportable; it is not a claim of certainty. Until then we show you the honest in-progress numbers instead of a headline we haven't earned.
In short: PROVISIONAL is us refusing to flatter ourselves with a number that doesn't yet mean anything. When it flips to VERIFIED, the edge is real and measured — not assumed.
| Date | Call | Realized regime | Result |
|---|
Grading rule: at the next session close we compute realized RTH volatility; EXPANSION if it exceeds the 20-day median, COMPRESSION otherwise — then compare to the morning call. Objective and falsifiable.
The signal-outcome ledger goes live with the curated feed — the first resolved signals will appear here automatically, and none will ever be removed.
| Strategy | SIM outcomes | Win rate | Gross points | Avg / outcome |
|---|
| Date | Strategy | Dir | Result | Points |
|---|
| Alert type | Firings | Graded | Status |
|---|
V2 waits for a volatility-scaled move and two continuous minutes of closes before declaring either a hold or break. Both outcomes use the same evidence and a 60-minute horizon. Controls preserve each session's zone width and opening distance. The probability forecast is walk-forward: a day's outcomes cannot train that day's call. Accuracy, Brier calibration, unresolved touches, later reversals, exclusions, and clustered uncertainty remain in the machine-readable ledger. The record stays provisional until it clears the pre-set practical bar.
The whole page is generated by the same pipeline that publishes the calls. If you want the machine-readable version: the ledger is served at api.convergepointresearch.com/v1/scorecard.
Research and decision support, not execution advice or a trade recommendation. Futures and options involve substantial risk.
