Every signal our system fires in real time during the session is graded by a machine against one set of rules fixed before the day begins, and published the next morning, including the misses.
posted Sep 22, 2026 at 07:00 EDT / fully automated
One segment per graded trading day, newest first. The strip runs in sequence and stays unbroken, so a losing day stays on the tape as the next segment.
Cumulative output of the grader since the tape started. Regenerated nightly by the job that posts this page; no figure here is typed by a person.
The same tape, split by the share price of the stock each signal named. Every band swings off the same centerline, so a losing band is drawn exactly like a winning one: same line, same scale, opposite direction.
3 of 4 price bands show a net paper loss. They sit on the same centerline as the winner, drawn at the same scale. What did not work is posted with everything else.
During the session the system reads the market minute by minute and fires a signal in real time when it predicts a stock is about to move +5%. A signal is one such call, on one stock, on one day. This page reports outcomes only: no model internals, no ranking, no live feed. Nothing here is actionable today; the newest gradeable day is yesterday.
Every signal is graded by one rule set, the policy, fixed before the day is graded and never changed: a +5% profit target, a two-hour window (120 minutes of market time), exit at the end of the trading day if the target is not reached, no stop-loss, and entries only after 10:30 ET. A signal that reaches +5% inside the window is a hit; one that runs out of time is a miss.
The dollar figures are paper results: what the policy would have made or lost if each signal had been traded at the same fixed dollar size every time. No real money traded these figures, and they are not fund performance. That is why every dollar figure on this page carries the label 'at fixed trade sizes' (the industry term is notional sizing).
The system fires its signals in real time during the session; nothing is published in real time. Each day's trading is graded the following morning and posted here by an automated job, one trading day after the fact, which is what 'T+1' means wherever it appears on this page. There is no same-day or live signal here. Member access to the system's signals is a separate product, delivered after a stated delay, under rules fixed in advance.
On August 19, 2026 we found a defect in the price data behind these grades. The grader had been reading price columns that a pipeline step meant for model inputs had clipped (winsorized: extreme values pulled in to a limit), so some paper trades registered a +5% target touch that did not happen in the market. On August 21, 2026, 27 sessions (July 14, 2026 to August 20, 2026) were regraded offline against the vendor's raw minute bars (per-minute price records straight from our market data provider): the same signals, the same rules, the same sizes; only the price data changed. Every session since is graded on vendor minute bars by the nightly job. No manual edits since August 21, 2026.
Grading source per session: vendor minute bars for 8 of 41 sessions (from September 10, 2026); each of those morning reports names the bar file it was graded on, with the file's checksum (a bar-source field); 33 sessions predate the field: 27 are certified by the August 21, 2026 regrade manifest and 6 were graded on vendor minute bars by the nightly job before the field existed.
The reasoning behind this record is published as letters, by Verbatim. subscribe →