Coin Detection Review · 2026-04-23

Can a language model actually trade crypto news?

We ingest every Tree News headline in real time, route it through a Groq-hosted classifier, and measure whether the model's directional calls align with sub-minute price action on HyperLiquid perpetuals. This is the most recent end-to-end review.

Raw headlines
40,254
pulled from Tree News
Detected
28,134
69.9% of pool
Classified
27,859
Groq openai/gpt-oss-20b
Directional calls
824
2.96% tradeable
The question

Is there edge at sub-minute horizons?

Crypto news is fast and information-dense. Most of the alpha from a material headline — an exploit, an ETF flow, a regulatory action — is absorbed by price within seconds. If an LLM-driven classifier can decide buy / sell / skipfaster than a human trader can even finish reading the headline, there may be a tradeable edge worth building infrastructure around.

This review stress-tests that premise. We compare the classifier's calls against real HyperLiquid 1-minute and 5-minute price moves on 20 tracked coins, and we deliberately sample all four corners of the confusion matrix — including the uncomfortable ones.

Headline findings

What the numbers suggest.

Signal quality
84.3%
of directional calls moved ≥ 5bp in 5m

Of 824 directional Groq calls on tracked coins, 695 were followed by meaningful price action. 129 went nowhere.

Detection gap
12.5%
of undetected headlines mention a tracked ticker

Our coin detector missed roughly 1,515 headlines that explicitly named a coin we trade. Most were TRUMP. This is an addressable detector bug, not an LLM problem.

Novelty filter
97%
is_novel=True in both Cat 3 and Cat 4

The 'delayed / stale event' hypothesis doesn't hold — flat-reaction signals are not dominated by old news. Classifier is firing on fresh events in both cases.

Every play, every reasoning, every return.

Open the plays browser to inspect all 200 sampled headlines, filtered by category, sorted by PnL, with classifier reasoning and 1m / 5m / 30m returns.

Open the plays browser