The Lab

Every trading day at 9:25 ET the Lab forecasts where SPY will be at 10:00, noon, 2:00 and the close, then learns from what actually happened. The bar it has to clear is the simplest honest forecast: the latest pre-market price holds.

Skill at the close +1.5% ± 1.0% · promising, not proven

Edge hit rate 63% direction beyond the latest price, 63 leaned days

Range coverage 70% target 68%

Days learned 135 first 15 are warmup and not counted

Call

Tue, Sep 15, 2026 Graded All inputs in

Close target $757.12, range $753.32 to $760.95, 21% chance of closing above the prior close ($760.88). Closed at $757.39.

Checkpoint	P(up)	Target	Range	Actual	In range	vs naive
10:00	14%	$758.36	$756.82 – $759.91	$759.86	Yes	Beat (0.316 vs 0.662)
Noon	21%	$757.85	$754.87 – $760.85	$757.51	Yes	Beat (0.229 vs 0.250)
2:00	21%	$757.56	$754.28 – $760.86	$757.17	Yes	Beat (0.211 vs 0.237)
Close	21%	$757.12	$753.32 – $760.95	$757.39	Yes	Beat (0.213 vs 0.240)

What moved this callWeight times today's reading, for the lead Kalman expert, in sigmas.

pushes downpushes up

Distance to gamma flip
Put skew
Put/call ratio (log)
Net dealer gamma
Baseline lean
IWM minus QQQ yesterday
Distance to max pain
Implied vs realized vol

Track record

Is it learning? Skill vs "latest price holds"Cumulative CRPS skill at the close, with a 2-standard-error band. Above zero means better than naive.

-10.0%-5.0%+0.0%+5.0%+10.0%

naive+1.5%

May 22Jul 20Sep 14Nov 5

Is the range honest?Share of the last 20 closes inside the displayed range. The Lab steers this toward 68%.

40%60%80%100%

68% target70%

May 22Jul 20Sep 14Nov 5

Who the Lab trustsEnsemble weight per expert at the close. Experts earn weight by forecasting better than the rest.

Kalman: everything, slow memory 20%
Kalman: price action 12%
Gaps keep going 12%
Analog days (10 nearest) 10%
Kalman: options positioning 9%
Everyone else 36%

0%50%100%

Apr 27Jul 1Sep 2Nov 5

CalibrationWhen the Lab said X% chance of an up close, how often it was

0%50%100%

0%25%50%75%100%

Leaderboard

Average CRPS at the close in sigma units (lower is better), and each expert's edge over "latest price holds". Weight is how much the ensemble listens to it today.

Expert	CRPS	vs naive	Weight	Verdict
Kalman: everything, slow memory	0.5147	+0.0114 ± 0.0100	19.7%	~252 more days to confirm
Gaps keep going	0.5188	+0.0073 ± 0.0077	12.1%	~416 more days to confirm
Kalman: price action	0.5211	+0.0050 ± 0.0099	12.5%	~1744 more days to confirm
Kalman: options positioning	0.5219	+0.0042 ± 0.0076	9.4%	~1468 more days to confirm
Analog days (10 nearest)	0.5247	+0.0013 ± 0.0077	10.1%	~16048 more days to confirm
Yesterday's move reverses	0.5258	+0.0003 ± 0.0050	7.0%	~156503 more days to confirm
Latest price holds (random walk)	0.5261	—	7.4%	the benchmark
Kalman: Jev text reads	0.5299	-0.0038 ± 0.0051	5.6%	no edge so far
Dealer hedging drift	0.5318	-0.0057 ± 0.0044	5.1%	no edge so far
Yesterday's direction continues	0.5331	-0.0071 ± 0.0051	5.5%	no edge so far
Kalman: everything, fast memory	0.5362	-0.0101 ± 0.0148	3.5%	no edge so far
Gaps fade	0.5501	-0.0241 ± 0.0084	2.1%	no edge so far

By checkpoint

Checkpoint	Skill	From range	From edge	Coverage	Edge hit rate
10:00	+9.4% ± 2.2%	+2.2%	+7.2%	72%	76% (74)
12:00	+1.4% ± 0.9%	+0.5%	+0.9%	68%	68% (41)
14:00	+2.0% ± 0.8%	+1.5%	+0.4%	70%	54% (35)
Close	+1.5% ± 1.0%	+0.5%	+1.0%	70%	63% (63)

How it learns

Senses. Options positioning from the 9:25 snapshot, yesterday's moves in stocks, bonds, credit, the dollar and VIX, and Jev's reads of the morning headlines and WSB.
Experts. Simple fixed hypotheses ("gaps fade"), Kalman filters whose weights drift slowly with the market, and an analog expert that asks what happened on the most similar past days.
Ensemble. Experts earn weight by forecasting better (Hedge), and every expert keeps a small share so one that fell out of favor can come back when the market changes.
Range. The range widens after a close lands outside it and narrows after one lands inside, so over time about 68% of closes fall inside it.