Free APIs That Replace a $2k Market Data Terminal
Bloomberg costs $2k/month. Tacavar's free stack — Wikipedia pageviews and FRED net liquidity — leads the market by weeks.
Bloomberg charges $2k/month. These two free APIs lead the market by hours and weeks.
That sounds like a troll until you look at what terminals actually sell: sentiment and liquidity. Both are computable from public endpoints. At Tacavar, we replaced our paid sentiment and macro layer with Wikipedia's Pageviews API and FRED's open data — two ingestors, one Postgres table, a daily cron, and a monthly bill of zero. The kicker is timing: these feeds lead the news cycle rather than echo it, so they tend to fire before the $2k dashboard moves.
Why Sentiment and Liquidity Are the Two Signals Worth Automating
If you're going to automate market data as a founder-operator, automate the things that move price and can't be gamed by press releases. That narrows to two categories.
Sentiment — the aggregate attention of the crowd — shows up in behavior before it shows up in commentary. Liquidity — the actual plumbing of money in the system — shows up in Fed line items before it shows up in risk assets. Everything else most terminals sell is either lagging price or repackaging one of these two.
The problem with paid sources is structural: they ingest news, and news is downstream of attention and policy. If your sentiment feed is built from headlines, you're trading a derivative of a derivative. Free market sentiment data that comes straight from human behavior — and macro plumbing straight from the Fed — gives you leading indicators Bitcoin actually respects. That's the thesis Tacavar's pipeline is built on: fewer inputs, upstream of everyone else.
Wikipedia Pageviews: The Free Curiosity Feed That Beats the News Cycle
The most underused free signal on the internet is Wikipedia's Pageviews REST API. It's free, unlimited, and reliably leads news cycles by hours. When millions of people start asking what is happening with Jerome Powell, the pageview count moves before CNBC books the segment and long before a sentiment dashboard ingests the coverage.
Tacavar runs an ingestor that pulls daily pageview stats for 15 curated articles — Bitcoin, Ethereum, Jerome Powell, Donald Trump, ChatGPT, Anthropic, and a handful of others chosen for their track record of leading risk-asset narratives. Each article is a proxy for a specific cluster of attention: monetary policy, crypto itself, political risk, AI. The aggregate curiosity of billions of humans, sampled daily, for free.
Almost nobody tracks this. That's the whole point. The Wikipedia pageviews trading signal works precisely because it isn't crowded — there's no sell-side coverage of aggregate curiosity, and no vendor can package what anyone can pull with a single HTTP request.
MAD Z-Scoring: Detecting Spikes Without Chasing Viral Outliers
Raw pageviews are noisy. One viral meme or one celebrity death in an adjacent article will dwarf a genuinely interesting move on the Powell page. Naive thresholding — flag anything above 3x average — chases exactly the wrong spikes.
The fix is median absolute deviation z-scoring. Instead of mean and standard deviation, both of which get dragged around by outliers, you compute the median of each series, take the median of absolute deviations from that median, and scale deviations against it. A spike on the Trump article during election season, where the baseline is elevated and jagged, won't trip the detector. A sudden, concentrated move on the Powell article the morning before a rate decision will.
In practice, Tacavar computes MAD z-scores per article over a rolling window and flags anything above a tuned threshold; z > 4 has worked well for daily data. It's roughly ten lines of code, robust to viral outliers, and it's the difference between a signal and a novelty feed.
FRED Net Liquidity: Ten Lines of Python for the Best Crypto Macro Signal
Liquidity is where the alpha-per-effort ratio gets absurd. The formula macro desks sell for $500–$2,000/month is:
net liquidity = DFF + WALCL - RRPONTSYD*1000 - WTREGEN
Fed funds rate, plus the Fed's balance sheet, minus reverse repo, minus the Treasury General Account. Every component sits on FRED, free, with an API key you can get in under a minute. The whole thing is ten lines of Python.
Why it matters: net liquidity leads crypto price action by weeks. When reverse repo balances drain and the TGA stops absorbing, dollars flow back into the system, and Bitcoin is among the first risk assets to respond. Macro hedge funds package exactly this and market it as proprietary research. Tacavar ships it as fred.py — a single ingestor that pulls four series, computes the formula, and writes the result next to everything else we track. If you build only one macro signal in Python this year, build this one: the FRED net liquidity crypto signal is the cleanest leading indicator for Bitcoin in all of free data.
Wiring Both Into a Unified Postgres Signals Table With a Daily Cron
Two signals are only useful if they're queryable together. Tacavar's architecture is deliberately boring: one Postgres table, one row per signal per day.
Schema essentials: signal_name, value, zscore, source, and captured_at, with a unique constraint on (signal_name, captured_at) so daily cron runs are idempotent upserts rather than duplicates. The Wikipedia ingestor and fred.py each pull, transform, and upsert independently — if one source has a bad day, the other keeps writing, and you can backfill cleanly.
The cron fires once a day after FRED and Wikipedia refresh their data. Every downstream consumer — dashboards, alerting, position-sizing logic — reads from the same table. No JSON blobs, no per-source logic leaking into strategy code. When you add a third signal later, it's a new ingestor and a new signal_name; nothing else changes.
What These Signals Caught That Paid Terminals Missed
Three patterns recur in Tacavar's pipeline data.
First, the Powell pattern: pageview spikes on his Wikipedia article reliably appear hours before press conferences dominate the news cycle — often the evening before, while markets are still pricing the previous consensus. By the time a terminal's sentiment feed flips, the move has started.
Second, the liquidity runway: sustained reverse repo drawdowns show up in the net liquidity series weeks before Bitcoin's accelerations, and Treasury General Account rebuilds show up before the corrections. A weekly glance at the net liquidity z-score is worth more than any daily macro newsletter.
Third, narrative rotation: simultaneous spikes across the ChatGPT and Anthropic articles have flagged AI-sector attention shifts days before equity and altcoin narratives repriced. Paid terminals caught all three eventually. The point is ordering: free, upstream data arrived first, and in trading, arriving first is the product.
From Signal to Action: Alerts, Weighting, and Review Cadence
A signal you don't act on is a hobby. Tacavar operationalizes both feeds with three layers.
Alerting: any MAD z-score breach or a week-over-week net liquidity inflection fires a webhook to Slack with the signal name, value, z-score, and a link to the source series. No dashboards to check — the pipeline comes to you.
Weighting: sentiment spikes adjust tactical positioning and attention strategy; net liquidity sets the risk regime. When the liquidity z-score is positive and expanding, Tacavar leans risk-on; when it's contracting, the pipeline says so weeks before the charts confirm it.
Review cadence: a weekly pass over false positives — which articles fired, whether they meant anything — is what keeps the article list at 15 high-quality proxies instead of 150 noisy ones. Signals decay; cadence is the maintenance.
None of this required a terminal subscription, a vendor contract, or proprietary data. It required two free APIs, one table, and the discipline to look upstream. Explore the Tacavar signal pipeline at tacavar.com/signals and see how free data becomes daily trading intelligence.