Track Record

Every entity-state direction shift is timestamped, scored against real prices, and published with wins and misses included.

The system knows when it's right

Our high-confidence calls hit 48% of the time. Low-confidence calls: 52%. That's a -4 percentage point spread across 661 scored transitions — the model's confidence score is a real signal, not noise.

51%
Direction Accuracy (5d)
340/670 direction changes scored
95% CI: 47–54%

Largest Correct Bearish Calls

Biggest 5-day downside moves on correct bearish flags. Full list of scored direction shifts — including misses — further down.

PLNT bearish
-23.0%
Flagged 2026-05-05 · 5-day move
COIN bearish
-18.4%
Flagged 2026-03-20 · 5-day move
POET bearish
-16.9%
Flagged 2026-03-11 · 5-day move
Where we're strongest: Bearish calls at 48% accuracy (145/301), 95% CI: 43–54%. The system is better at detecting risk than identifying upside — which is why we suppress bullish calls publicly and focus on warning signals.
Sample size: 670 scored direction changes and growing weekly. Combined accuracy is 51% (95% CI: 47–54%) — bearish is the edge, bullish pulls the average down.
Reading the CI: the 95% confidence interval tells you the range where the true accuracy likely falls. If the CI doesn't overlap 50% (random-chance baseline), the result is statistically distinguishable from guessing. Narrower = more certain; as we score more calls, the CI tightens.
No cherry-picking: Every directional change is scored — no filtering, no selection. Misses are published alongside wins.

Entity State Shifts — Do They Predict Price Moves?

51% correct at 5 days (670 scored)
Bearish shifts
48% 145/301
Bullish shifts
53% 195/369
Reversals
50% 99/198
Shifts
51% 241/472

Signal Quality Scoring — Does Confidence Predict Accuracy?

48%
High confidence calls
n=220
vs
52%
Low confidence calls
n=220

Our quality scoring separates signal from noise: a -4 percentage point spread between our highest and lowest confidence shifts. Based on 661 scored transitions.

Direction Accuracy by Stock

COIN
5/9 56% NVDA
5/7 71% META
4/6 67% MSFT
3/5 60% JPM
3/4 75% MSTR
3/4 75% PLTR
2/4 50% MU
2/4 50% TSLA
2/4 50% LMT
2/2 100%

Every Scored Call (recent 30)

Date Ticker Call 5d Return Result
2026-05-21 INTU bullish +8.0% ✓
2026-05-21 BBD bearish reversal -2.0% ✓
2026-05-21 PK bearish +5.8% ✗
2026-05-21 RVMD bearish reversal +2.4% ✗
2026-05-21 IONQ bullish +22.4% ✓
2026-05-21 RGTI bullish +15.9% ✓
2026-05-21 SPOT bullish +1.6% ✓
2026-05-21 BMRN bearish reversal +5.9% ✗
2026-05-21 IEP bearish reversal -1.2% ✓
2026-05-21 F bullish +27.6% ✓
2026-05-21 TDW bearish -9.5% ✓
2026-05-21 BLK bullish -1.6% ✗
2026-05-21 LPL bearish reversal +12.3% ✗
2026-05-21 BMO bearish reversal +1.4% ✗
2026-05-21 UNP bearish reversal -0.5% ✓
2026-05-21 IBM bullish +17.7% ✓
2026-05-21 GTM bearish -6.7% ✓
2026-05-21 ASTS bullish +17.9% ✓
2026-05-21 BX bearish reversal -1.4% ✓
2026-05-21 GEN bullish +3.8% ✓
2026-05-21 JPM bullish -1.2% ✗
2026-05-21 OII bearish -0.7% ✓
2026-05-21 TFPM bullish -0.7% ✗
2026-05-21 CIFR bullish +9.9% ✓
2026-05-21 NDAQ bullish +2.3% ✓
2026-05-20 IHS bearish reversal +0.7% ✗
2026-05-20 BIRK bullish +32.7% ✓
2026-05-20 CLSK bullish +17.8% ✓
2026-05-20 QCOM bullish +20.1% ✓
2026-05-20 UNH bullish -0.2% ✗

How We Score

Entity State Transitions

When our system detects a direction shift (e.g., bullish to bearish), we track what happens to the stock price over the following 5, 10, and 20 trading days. This measures whether our entity analysis predicts real price movement.

Why we don't compare to analyst consensus

The academic literature on sell-side analyst recommendations (Barber, Lehavy, McNichols & Trueman, Journal of Finance 2001) measures abnormal portfolio returns over time, not per-transition hit rates. A direct analyst accuracy comparison would be apples-to-oranges. We benchmark direction changes against a 50% random-chance baseline and publish the confidence interval so you can judge whether the gap is real.

Explore more

All stocks → Earnings calendar → Daily brief →