Tonight's 20:35 at Chepstow was a seven-horse maiden. By the time we'd looked at the dashboard, five of the seven were carrying the `⚠OR` chip — meaning the BHA handicapper had certified each of them as operating significantly below their career peak, with 12 or more runs of evidence. Believitanducan down 50 points. St Faz down 52. The others ranging from −11 to −24.
That left two clean horses.
Weston Court — 56 career runs, 4 wins. A horse that knows how to cross the line first, just inconsistently. No ceiling chip. Market's choice.
Marisitta — 46 career runs, 0 wins. Never won in her life. Best going is soft; tonight was good-to-firm. Best distance is 1m; tonight was shorter. The dashboard would have shown her `G:Sf` in red and `D:` in amber. On every metric we'd built, she was the wrong horse on the wrong ground at the wrong trip.
She won at 18/1.
The five ⚠OR horses finished where the data said they would — behind. The signal did its job. What the signal can't do is choose between two horses it has no objection to. That becomes a coin flip, and tonight the coin landed on the one with 46 previous attempts and nothing to show for it.
This is the core use case for these signals: elimination, not selection. You don't need to know who will win. You need to know who probably won't. In a seven-horse race where five are flagged, the field is effectively two horses — and at 18/1, the one that wins pays handsomely regardless of which you pick.
We'd spent part of today building a statistical profile of horses like her — 20+ runs without a win. The numbers:
Marisitta has now joined the 90.3% that eventually break through — which is precisely why laying them blindly doesn't work. The edge is in the price, not the outcome. At 18/1, the market was being honest about her. At 6/4, it wouldn't be. The historical pattern says a horse like this typically returns to losing. The win doesn't reset anything structurally. We'll track her next few runs.
The badge we built tonight — `✗W 46` in dark grey — is named the Marisitta badge. It appears in the Form column for any horse with 20+ career runs and zero wins. The tooltip tells the full story, and ends with the note that the badge is named after a horse who proved the exception on the night it was built.
Six new signals and badges, all live in the dashboard Form column:
⚠DB Draw Burden — last run prominent from a disfavoured draw stall on a biased course. A/E = 0.483 next run (n=817). The horse spent fuel getting to the front that the market doesn't account for next time.
⚠FI Fade Index — led or ran 2nd at halfway in 3+ commentary runs, faded to the back half of the field in 75%+ of those occasions. A/E = 0.455 next 3 runs (n=395). A structural stamina ceiling, not bad luck.
⚠OR OR Stagnation — current rating 10+ points below career peak, 12+ runs. A/E = 0.627 (n=11,852). The BHA has certified the ceiling. Shows as `⚠OR 114→90` so you can judge severity at a glance.
D: Best Distance — horse's optimal trip vs today's. Green = on ideal trip (A/E=0.804). Red = wrong trip by 1.5f+ (A/E=0.508). Stable every year 2020–2026.
G: Best Going — horse's best going band (Sf/Gd/Fm/AW), coloured by how today's going compares. Best-going A/E=0.794 vs wrong-going A/E=0.465. Soft horse on fast: A/E=0.397.
✗W Marisitta — 20+ career runs, zero wins. Named tonight.
All of this started with a different horse. Our NAP the previous evening — Control Room at Windsor — crossed to the rail from a wide draw, took the lead, and then couldn't sustain it in the final furlong. It wasn't that Control Room was outclassed. It was that the energy spent getting across to the rail left nothing for the finish.
That observation became the Draw Burden signal. Then the Fade Index. Then the question of how you measure a horse's absolute ability ceiling — which led to the OR stagnation work, the RPR plateau analysis, and eventually to Marisitta and the never-won breakdown.
One race, watched closely, generated a full day of research. That's the feedback loop the dashboard is built on.
The Fade Index and Draw Burden signals cover the 2026 flat season in commentary data (10,389 runs, 34 courses). The jumps season — October 2025 to March 2026 — is the next download. Once transcribed, the `weakened` jumps signal (currently A/E=0.979 on only 43 instances) should reach a meaningful sample.
And we'll be watching Marisitta's next race.
Dashboard live at dre.cogtrix.eu. Picks via email after daily pipeline run.