Crypto Express 3000App

Forecast

Desk follow-ups after Kwon's morning brief: one symbol, one courier, technical read with a link back to the wrap.

+ How our 24h forecasts score

198 resolved follow-ups · last updated 2026-07-28

We publish once per day after Kwon's brief. Each symbol forecast is graded after 24 hours: where price closed, whether key levels were touched during the day, and how well the stated probabilities held up.

Close inside expected range

73%

144 of 198 forecasts

How often the final price landed inside the model’s typical 10–90% band — the range where most similar past setups ended up.

73%

Directional calls

51%

100 of 198 forecasts · 66 with a clear signal

When the desk filed a directional call (not “no edge”), did the 24h closing price move the same way as the highest-probability scenario?

51%

Key levels touched

52%

96 of 184 forecasts

Did price reach the Up or Down target from the forecast during the 24h window — even if it pulled back before the close?

52%

Probability calibration

Building

Brier 0.60 · lower is better

Still calibrating — sample or regime may be noisy.

25%
+ What do these numbers mean?
Directional calls
Many follow-ups are tagged “no edge” — a levels watch, not a trade call. This row only counts posts where the desk published a directional signal.
Key levels touched
Example: upside target near $67k — price trades to $67.5k during the day but closes at $64k. The closing-direction call may miss; this metric still counts the level as touched.
Close inside expected range
The Bayesian block includes a 10–90% price band from historical analogs. We check whether the 24h close stayed inside that band.
Probability calibration
Compares the three stated probabilities (Up / Down / Flat) to what actually happened. Useful over weeks of data, not a single day.

Showing 1–6 of 20 briefes (100 desk follow-ups)