Finding Mispriced Attention Before the Market Does
Finding Mispriced Attention Before the Market Does
Information moves fast. Prices don't.
The gap between when something becomes news and when it becomes priced is the window. Most traders try to close that window by reading faster. The Attention Mispricing Watch closes it by never stopping.
It runs hourly. It scans for assets experiencing attention spikes — mentions surging, sentiment shifting, unusual volume patterns — and cross-references that attention against price movement. If attention is spiking but price hasn't moved, it flags it. If price is moving on no attention, it flags that too.
How it works
The engine has three data sources feeding it continuously:
Attention signals. Blog sentiment, RSS feed mentions, social volume. The system tracks 17 RSS feeds and a blog sentiment scanner that processes financial media in real time. When a ticker suddenly appears in 4+ sources where it was absent yesterday, that's an attention spike.
Price signals. The market snapshot pipeline captures price data every 30 minutes across tracked assets. It computes short-term changes, volatility expansion, and volume anomalies.
The gap. The magic is in the delta. Attention without price movement means the market hasn't caught up. Price without attention means something's moving that nobody's talking about. Both are signals.
Every hour, the engine scores each tracked asset on an attention/price alignment scale. Anything with a delta above threshold gets flagged and delivered to Telegram.
A real example
Last month, a small-cap mining stock appeared in three RSS feeds within two hours — a drilling result, an analyst note, and a regulatory filing. The attention score spiked from 0.2 to 0.9. Price had moved 0.3%. The engine flagged it.
By the time the market opened, the stock was up 12%.
The engine didn't predict that. It didn't recommend a trade. It just noticed that a lot of people were suddenly talking about something the market hadn't priced yet. That's the job. The decision — whether to act, how much, when — is yours.
The architecture
- Cron trigger: Every hour, on the hour
- Data sources: RSS feeds (17), blog sentiment scanner, market snapshots
- Compute: Pure Python — pandas, numpy, scipy
- Cost: $0.00 per run
- Output: Telegram message with flagged assets and delta scores
It's a simple pipeline. Fetch → Score → Filter → Deliver. No LLM. No API calls. Just math running on a timer.
The hardest part wasn't building it. It was calibrating the thresholds. Too sensitive and everything gets flagged. Too conservative and nothing does. I ran it silently for two weeks, logging every flag against actual price movements, before setting the current thresholds. The engine learned its own calibration from its own data.
What this means
Most financial media is backward-looking. It tells you what happened. The Attention Mispricing Watch tells you what's happening right now that hasn't happened yet in price.
It's not a trading strategy. It's an information advantage. And it costs less than a single market data subscription.