moonboard.ioResearch & Data
Methodology

Sentiment Divergence Edge (SDE)

The gap between how positive the conversation about an asset is and how calm its price has been, ranked across the universe.

Scale
percentile Relative scale — the value states rank within the scanned universe, not magnitude.
Formula
sentiment − (1 − min(volatility, 1)) × 100 → percentile
Inputs
  • sentiment
  • volatility
What the second term actually is

"Divergence" invites the assumption that SDE compares sentiment against price direction — the classic contrarian setup where mood and returns disagree. It does not. The second term is derived from realised volatility, which has no sign. SDE therefore measures enthusiasm relative to turbulence, not relative to direction, and it cannot distinguish an asset that rallied 20 % from one that fell 20 %. Everything below follows from that.

Purpose

Two quantities that usually track each other are placed side by side. Enthusiastic conversation normally accompanies price movement — communities get loud when something happens. SDE isolates the cases where the two are out of step, in either direction:

  • Positive gap — the conversation is far more upbeat than recent price behaviour has been eventful. Enthusiasm without much movement to justify it.
  • Negative gap — the price has been turbulent while sentiment stayed muted. Movement the conversation has not caught up with.

Both are observations about a mismatch. Neither is a signal about what happens next.

Inputs

Two fields: sentiment (0–100) and volatility (a fraction). No social counts, no volume, no market cap. This makes SDE computable for assets with almost no social coverage — but see the limitations, because that is a mixed blessing.

Formula

stability = (1 − min(volatility, 1)) × 100 divergence = sentiment − stability SDE = percentile of divergence across the universe

min(volatility, 1) caps the input at 100 % volatility, so stability is bounded to 0–100 and cannot go negative on an extreme reading. The raw divergence therefore spans −100 to +100, and the percentile step maps that onto 0–100.

Working through the sign

Because stability is the inverse of volatility, the composition is easy to get backwards. Written out:

SentimentVolatilityStabilityDivergenceProfile
850.0991−6Upbeat but calm
850.4060+25Upbeat and turbulent
350.4060−25Turbulent, conversation unenthused
350.0595−60Quiet on both counts

A high SDE therefore requires both high sentiment and high volatility. It is not a measure of positivity alone: a strongly positive, stable asset lands in the middle of the distribution, not the top.

Interpretation

SDEReading
~20Sentiment is low relative to how much the price has been moving, or both are subdued. Little enthusiasm attached to whatever is happening.
~50Sentiment and turbulence are roughly in the usual relationship for this universe.
~70Conversation running warmer than recent price behaviour alone would suggest.
~90Strongly positive sentiment alongside substantial volatility. In the application this region is presented as the overheated end of the scale.
On "contrarian" readings

A high SDE is sometimes described as a signal to fade and a low SDE as a signal to accumulate. Neither interpretation follows from the construction, and neither is validated here. SDE is a description of a current mismatch between two measured quantities. Whether such mismatches resolve in a particular direction is an empirical question — and one nobody should treat as settled without a study that accounts for look-ahead bias, survivorship and the shifting composition of the universe. See research notes for how that question is framed.

Worked example

Asset A: sentiment 78, volatility 0.31 stability = (1 − 0.31) × 100 = 69 divergence = 78 − 69 = +9 Asset B: sentiment 78, volatility 0.04 stability = 96 divergence = 78 − 96 = −18 Same sentiment. Asset A ranks well above Asset B on SDE, purely because its price has been more turbulent.

Two assets with identical conversation quality land far apart. If your intuition says they should rank together, that intuition is describing a sentiment metric — and SDE is not one. The sentiment field on its own is available directly in the application.

Limitations

  • Volatility is unsigned. A crash and a rally of the same magnitude produce the same SDE. This is the metric's single largest constraint.
  • Not a price-momentum comparison, despite what the name suggests to most readers.
  • Two different scales are subtracted directly. Sentiment is a model output on a 0–100 classification scale; stability is a transformed physical measurement. Their difference is arithmetically defined but has no natural unit — which is part of why only its rank is published.
  • Sentiment quality varies with coverage. An asset with a handful of posts produces a sentiment figure computed from a handful of posts, and SDE gives it the same standing as one derived from thousands.
  • The volatility cap flattens extremes. Above 100 % volatility every asset receives a stability of 0, so the most turbulent assets become indistinguishable on that term.
  • Universe-dependent. The percentile moves when the scanned set changes.

Data quality

Both inputs must be present. A missing sentiment classification is read as zero, which produces a strongly negative divergence and places the asset near the bottom of the ranking — an artefact of missing data, not a finding about the asset. Thinly covered assets are the ones most affected.

Related