Data / Sentiment

Sentiment data

An aggregated measure of how positively or negatively an asset is being discussed. Useful as a comparative and historical signal, meaningless without the activity volume it was computed from.

Read this first

Sentiment is an analytical measurement of published opinion. It is not a forecast, not a recommendation, and not evidence about future price. A high sentiment reading tells you the conversation was positive over a window — it tells you nothing about whether that conversation was right.

What this data is

For each asset in the hourly snapshot, sentiment expresses the balance of positive against negative in the public posts that mentioned it during the trailing window, on a 0 to 100 scale. It sits on exactly the same timestamps as the market and social rows for that asset, which is what allows sentiment and price to be plotted against each other honestly.

Sentiment classification

Classification happens upstream, at the level of individual posts, before anything reaches Moonlytics. Each collected post carrying an asset topic is assigned a polarity; the polarity distribution over the window is then collapsed into a single figure per asset.

Moonlytics stores the aggregate, not the underlying posts. That is a deliberate boundary: post-level content carries platform terms and copyright constraints that an aggregate measure does not.

Posts

public postasset topic attached

Classificationupstream, per post

polarity per postweighted by engagement

Aggregation24 hour trailing window

distribution over windowcollapse to 0–100

Storage

one value per asset per houraligned to market row

Time windows

The stored value describes a trailing 24 hour window, refreshed hourly. That combination has a specific consequence worth understanding: consecutive hourly values share 23 hours of input, so the series is smooth by construction. A sharp move in the sentiment series therefore represents a genuinely large shift in the last hour of posts, not noise.

Because every value is retained, longer windows can be computed downstream — daily means, weekly baselines, or z-scores against an asset's own history. The layer stores the atom; the aggregation is the consumer's choice.

Levels of aggregation

Asset-level sentiment

The stored primitive. One value, one asset, one hour. Everything else is derived from it.

Sector sentiment

Assets carry provider categories — Layer 1, DeFi, memecoin, AI and so on — which allows sentiment to be averaged across a sector. Because category membership comes from the metadata provider and assets frequently belong to several categories at once, sector sentiment is a derived view rather than a stored series, and it is only as clean as the category assignment behind it.

Market sentiment

A market-wide reading can be formed by weighting asset sentiment by activity or by capitalisation. Moonboard's market mood surface does exactly this and blends it with realised price action, because a sentiment reading alone overstates its own importance. The blend is documented on moonboard.io.

Update frequency

Hourly, on the same schedule as market and social data. Retained since November 2025.

Available fields

sentimentintegerAsset-level sentiment for the trailing window, on a 0–100 scale. Higher means the classified conversation skewed more positive. Nullable when the asset had too little activity to classify.

observedAttimestampUTC snapshot timestamp, aligned to the top of the hour.

assetreferenceThe canonical asset the measure belongs to.

sentiment alongside the activity it came from

{
  "asset": { "id": 1, "symbol": "BTC", "name": "Bitcoin" },
  "observedAt": "2026-08-31T09: 00: 00Z",
  "sentiment": 81,
  "socialVolume24h": 41208
}

Sentiment is never served without the volume it was computed from. A reading of 92 on eleven posts and a reading of 92 on forty thousand are not the same fact, and separating them from each other invites misreading.

Limitations

Thin samples are unstable. Assets with low social volume produce sentiment values that swing on a handful of posts. Always read sentiment against socialVolume24h.
Irony and jargon defeat classifiers. Crypto discourse is dense with sarcasm, in-group idiom and deliberately inverted language. Misclassification is systematic in that register, not random.
Manipulation is possible. Coordinated posting moves both volume and polarity. The measure does not attest that the conversation was organic.
Nullable. Assets with insufficient classified activity carry no value for that hour. Null means unknown.
Not comparable across providers. A 0–100 sentiment figure from a different vendor is a different measurement. Do not mix the series.

Which applications use it

Sentiment feeds the sentiment-divergence and market mood views on moonboard.ai, where it is deliberately shown next to price rather than instead of it.

Social data — the activity volume sentiment is computed over.
News data — event flow that often precedes a sentiment move.
Historical data — the full retained sentiment series.