How to Use Wikipedia Attention Data as a Trading Signal

Turn Alphanume's Wikipedia Views z-score into an attention-spike signal you can filter and backtest.

Page-view counts vary enormously across companies, so a raw count is almost useless as a cross-sectional signal. What you want is a normalized measure of unusual attention, and that is exactly what the z-score in the Alphanume Wikipedia Views dataset provides.

The endpoint

GET https://api.alphanume.com/v1/wikipedia-views

This is a Pro endpoint. Authenticate with ?api_key=alp_your_key or an X-API-Key header. The useful filters are ticker, date, and the z-score filters zscore_30d_gte, zscore_30d_lte, zscore_30d_gt, zscore_30d_lt, and zscore_30d_eq.

Find attention spikes

import requests

r = requests.get(
    "https://api.alphanume.com/v1/wikipedia-views",
    params={"zscore_30d_gte": 3, "api_key": "alp_your_key"},
)
for row in r.json()["data"]:
    print(row["date"], row["ticker"], round(row["zscore_30d"], 1))

Filtering on zscore_30d_gte=3 surfaces names where today's page views are at least three standard deviations above their 30-day norm, a strong attention spike. Because the z-score is already normalized by each name's own history, a quiet company and a mega-cap are directly comparable.

Use it as a factor or a filter

• As a factor: rank your universe by zscore_30d and study forward returns of high-attention versus low-attention names.
• As a filter: require an attention spike to confirm an event-driven setup before you act.
• As context: pull the views and avg_30d alongside the z-score to see whether a spike is large in absolute terms or just unusual for that name.

Backtest it cleanly

The dataset is cursor-paginated and ordered by date, so you can pull a long history, join it to your price data by ticker and date, and test attention-regime strategies. Loop on next_cursor while has_more is true to page through large pulls.

Access and limits

Wikipedia Views is Pro-only, with a 600 requests per minute rate limit and HTTP 429 on overage. Read the API documentation for the full schema, review Alphanume datasets, or grab a free Alphanume API key to start with the free catalog.

Spike detection without the noise

The z-score is what makes attention tradable. Filtering on zscore_30d_gte=3 surfaces names whose page views are at least three standard deviations above their 30-day norm, a genuine attention spike rather than a name that simply gets a lot of traffic every day. The normalization removes the scale problem that makes raw counts unusable.

From there the data supports both factor and filter approaches: rank the universe by z-score to study forward returns of high-attention names, or require a spike to confirm an event-driven setup before acting. The dataset is cursor-paginated and ordered by date, so you can pull a long history and join it to price data by ticker and date for a clean backtest.

Frequently asked questions

What is the Wikipedia Views endpoint?

GET https://api.alphanume.com/v1/wikipedia-views. It is a Pro endpoint; authenticate with ?api_key=alp_your_key or an X-API-Key header.

How do I find attention spikes?

Filter on the z-score, for example zscore_30d_gte=3, to surface names whose page views are at least three standard deviations above their 30-day norm.

Which z-score filters exist?

zscore_30d_gte, zscore_30d_lte, zscore_30d_gt, zscore_30d_lt, and zscore_30d_eq, alongside ticker and date.

Can I use it as a factor?

Yes. Rank your universe by zscore_30d and study forward returns of high-attention versus low-attention names, or require a spike to confirm an event-driven setup.

How do I page through history?

Loop on next_cursor while has_more is true to pull a long history, then join it to price data by ticker and date for backtesting.