🐆 SPOTS — SCOREBOARD
Every number we publish, graded. Misses shown the same size as hits.
Lifetime record BACKFILLED
63.5% of calls hit · 20,327 calls graded
2019-08-28 → 2026-07-28 · 6 instruments · average stated 63.5%, actual 63.5% — a gap of 0.0 points.
Calibration — does a stated probability mean what it says?
said 50–60%55.4%56.3% actualn=1,830
said 60–70%65.8%66.0% actualn=7,958
said 70–80%73.6%74.1% actualn=2,309
said 80–90%86.9%87.0% actualn=2,308
said 90–100%91.5%91.6% actualn=824
Green = within 2 points of the claim. This is the table that matters: a 74% that happens 74% of the time is worth more than a 90% that happens 60% of the time.
Recent form — last 90 days
69.0% actual vs 64.0% stated · 849 calls
Lifetime numbers can hide a system that stopped working. This is the short window.
By year — was one era carrying it?
201961.3%60.4% actualn=874
202064.6%64.7% actualn=2,848
202164.7%62.8% actualn=2,885
202263.3%62.8% actualn=2,917
202362.3%60.1% actualn=2,818
202462.2%62.1% actualn=2,854
202564.3%65.9% actualn=3,238
202663.6%67.9% actualn=1,893
Per-signal health — worst drift first
Price returns to the midnight open62.9%57.5% (-5.3)n=325
NY morning breaks the London low64.3%59.8% (-4.5)n=652
Closes in the bottom third59.2%56.4% (-2.8)n=516
NY morning breaks the London high56.9%54.6% (-2.3)n=885
Closes in the top third61.9%60.7% (-1.3)n=615
Range expands vs yesterday63.7%63.0% (-0.8)n=848
Morning low is in by 10:3059.5%59.4% (-0.1)n=6,287
Breaks yesterday's low77.3%77.4% (+0.2)n=2,011
Breaks yesterday's high72.3%73.1% (+0.7)n=2,521
Outside day (takes both extremes)27.5%28.3% (+0.8)n=665
Day closes green66.5%68.1% (+1.6)n=4,833
A signal drifting below its claim gets flagged here in public rather than quietly dropped. That is the point of publishing this at all.
The last 40 calls
2026-07-28NQNY morning breaks the London high❌ miss
2026-07-28NQDay closes green✅ hit
2026-07-28NQMorning low is in by 10:30✅ hit
2026-07-28NQBreaks yesterday's high✅ hit
2026-07-28NQRange expands vs yesterday❌ miss
2026-07-28ESNY morning breaks the London high✅ hit
2026-07-28ESMorning low is in by 10:30❌ miss
2026-07-28YMDay closes green✅ hit
2026-07-28RTYBreaks yesterday's low❌ miss
2026-07-28GCDay closes green✅ hit
2026-07-28GCMorning low is in by 10:30✅ hit
2026-07-28GCDay's HIGH printed in Asia❌ miss
2026-07-28CLDay closes green✅ hit
2026-07-28CLMorning low is in by 10:30❌ miss
2026-07-28CLBreaks yesterday's high✅ hit
2026-07-27ESDay closes green❌ miss
2026-07-27ESMorning low is in by 10:30✅ hit
2026-07-27ESBreaks yesterday's low❌ miss
2026-07-27YMDay closes green✅ hit
2026-07-27YMMorning low is in by 10:30✅ hit
2026-07-27YMBreaks yesterday's low✅ hit
2026-07-27YMCloses in the top third❌ miss
2026-07-27RTYDay closes green✅ hit
2026-07-27RTYMorning low is in by 10:30✅ hit
2026-07-27RTYBreaks yesterday's low✅ hit
2026-07-27CLDay closes green✅ hit
2026-07-27CLMorning low is in by 10:30❌ miss
2026-07-27CLBreaks yesterday's high✅ hit
2026-07-24NQMorning low is in by 10:30✅ hit
2026-07-24ESMorning low is in by 10:30✅ hit
2026-07-24YMDay closes green✅ hit
2026-07-24RTYBreaks yesterday's low✅ hit
2026-07-24CLDay closes green✅ hit
2026-07-24CLMorning low is in by 10:30❌ miss
2026-07-24CLBreaks yesterday's high✅ hit
2026-07-23NQDay closes green✅ hit
2026-07-23NQMorning low is in by 10:30❌ miss
2026-07-23ESDay closes green✅ hit
2026-07-23ESMorning low is in by 10:30❌ miss
2026-07-23ESBreaks yesterday's high✅ hit
How this is computed (the part nobody else publishes)
Walk-forward. The probability quoted on any day is computed only from days before it. Grading a number that was fitted on the same data it is graded against is worthless, and it is what almost every published "win rate" actually is.
No cherry-picking. Every day that clears the evidence bar records a call — win or lose. There is no filter that can be tuned after the results are known.
Minimum evidence. A call needs at least 120 comparable prior days and an edge of at least 8 points over the base rate. Otherwise we say nothing, and silence is a legitimate output.
No tautologies. Outcomes logically implied by the conditions are never quoted as forecasts — if the overnight session already took yesterday's low, "breaks yesterday's low: 100%" is not a prediction.
Self-correcting. The stated number is adjusted by how similar past calls actually resolved, using only calls made before that day. The record does not just report the tool — it improves it.
Backfilled, and labelled. Everything above is reconstructed from history under those rules. Live calls accumulate on top and are marked separately. We are not going to pretend a track record started earlier than it did.