Attention DensityLISTENING TO THE TAPE…
Attention DensityAttention DensityAttention Densitylivesso-v1 · 52 measures · additive-only, versioned contractsso-v1:b7532835ede0b705— cite this to pin the exact version you validated against| Measure | Formula | Inputs |
|---|---|---|
| attention_density scoring | WSA / (market_cap / 1e9) | weighted_social_activity, market_cap |
| float_attention_density scoring | WSA / (float_market_cap / 1e9) | weighted_social_activity, float_market_cap |
| liquidity_pressure scoring | WSA / (avg_dollar_volume / 1e9) (the crowding ratio) | weighted_social_activity, avg_dollar_volume |
| weighted_social_activity pipeline.wsa | Σ source×author_quality×engagement×relevance×novelty×time_decay×spam_penalty | social_posts, social_authors, ticker_mentions |
| attention_density_percentile pipeline.percentiles | rank(attention_density within market_cap_bucket | sector) | attention_density, market_cap_bucket, sector |
| social_pressure_index pipeline.spi | weighted Σ(density_z, velocity_z, accel_z, liquidity, float, source_quality, novelty) − pump_penalty (mode-tuned) | attention_density_percentile, velocity_zscore, acceleration_zscore, pump_risk_score |
| attention_share pipeline.rotation | WSA / Σ WSA (zero-sum); rotation = Δshare | weighted_social_activity (universe) |
| attention_candle pipeline.candles | OHLC of WSA per window bucket | weighted_social_activity (samples) |
| Measure | Formula | Inputs |
|---|---|---|
| velocity_zscore pipeline.velocity | (WSA − baseline_mean) / baseline_std | weighted_social_activity, baseline |
| acceleration_zscore pipeline.velocity | Δ velocity_zscore | velocity_zscore (series) |
| velocity_robust_z pipeline.robust | 0.6745·(WSA − median)/MAD — outlier-resistant companion to velocity_zscore | weighted_social_activity, baseline |
| attention_anomaly_band pipeline.anomaly_bands | classify WSA vs robust seasonal band (median ± 2σ) → above/within/below + signed attention_sigmas_out | weighted_social_activity, baseline |
| attention_freshness pipeline.freshness | minutes_since_latest, median post age, fresh_share (share within 60 min) | post timestamps, cycle clock |
| attention_ignition pipeline.ignition | current_WSA / median(baseline) → ratio + magnitude band (ignition≥10×/surge≥3×/elevated≥1.5×/normal) | weighted_social_activity, baseline |
| seasonal_z pipeline.seasonal_store | shrunk seasonal z (WSA vs ticker×dow×hour cell, shrunk toward the ticker parent) + basis (cell/shrunk) | weighted_social_activity, seasonal cells (archive) |
| seasonal_log_z pipeline.seasonal_log | shrunk seasonal z in LOG space: (log1p(WSA) − mean_log)/std_log vs the ticker×dow×hour cell (log1p Welford), shrunk toward the ticker parent + basis — symmetric for multiplicative moves, robust to a single viral outlier | weighted_social_activity, seasonal cells (log1p, archive) |
| attention_dose pipeline.attention_dose | accumulated excess attention above baseline over the window: dose = Σ max(0, WSA−B), normalized_dose = dose/B, + dwell_share/peak_excess/spike_share and a spike/sustained/mixed/quiet profile (magnitude × duration in one number) | weighted_social_activity (series), baseline (archive history mean) |
| lifecycle_state pipeline.lifecycle | state machine over (percentile, velocity, acceleration) | attention_density_percentile, velocity_zscore, acceleration_zscore |
| attention_half_life pipeline.halflife | ln(2)/λ from post-peak exp-decay fit | weighted_social_activity (series) |
| session_phase pipeline.session_phase | US-equity session (America/New_York, DST-aware): weekend/premarket/open/midday/close/afterhours/overnight | window_end |
| attention_share_rotation pipeline.rotation | Δ(share) period-over-period; top inflows/outflows | attention_share (t, t-1) |
| attention_jerk pipeline.jerk | Δ acceleration_zscore (3rd derivative); regime_break = |jerk| ≥ 2 | acceleration_zscore (series) |
| attention_persistence pipeline.persistence | lag-1 autocorrelation of the WSA series; durable ≥ 0.5, fading ≥ 0, else choppy | weighted_social_activity (series) |
| attention_trend_quality pipeline.trend_quality | R² of a linear fit to the WSA series (clean ramp vs choppy) + direction | weighted_social_activity (series) |
| Measure | Formula | Inputs |
|---|---|---|
| narrative_velocity pipeline.narrative | z-score(theme_activity vs theme_baseline) | theme_members WSA, theme_baseline |
| attention_beta pipeline.beta | cov(ticker Δ, market Δ) / var(market Δ) | weighted_social_activity (ticker, market) |
| idiosyncratic_attention pipeline.idiosyncratic | current residual Δticker − β·Δmarket (market tide removed) + its z + is_name_driven | weighted_social_activity (ticker, market series) |
| attention_beta_corr_alpha pipeline.beta | see attention_beta; also correlation + alpha (name-specific drift) | weighted_social_activity (ticker, market) |
| narrative_breadth pipeline.breadth | share of theme constituents lit (velocity_z ≥ 1) + top_wsa_share; broad ≥ 0.5 vs single-name (one name ≳ half the WSA) | theme constituents (velocity_zscore, wsa) |
| Measure | Formula | Inputs |
|---|---|---|
| source_quality_score pipeline.source_quality | weighted mean of unique-author quality | social_authors.quality_score |
| pump_risk_score scoring | 0.25·spam_cluster+0.20·low_float+0.20·low_liquidity+0.15·low_source_quality+0.10·promotional_language+0.10·price_extended | spam_score, float_market_cap, avg_dollar_volume, source_quality_score, recent_dilution |
| organic_score pipeline.organic | 1 − (0.35·dup+0.25·burst+0.20·new_account+0.20·author_concentration) | post texts, posted_at, account_age_days, author_id |
| new_returning_attention pipeline.audience | |current_authors − prior_authors| / |current_authors| (fresh ratio) | current-window author set, prior-window author set |
| organic_score_painting pipeline.organic | manufactured_probability = weighted(dup, burst, new_account, concentration) | post texts, posted_at, account_age_days, author_id |
| attention_tiers pipeline.attention_tiers | WSA share per credibility tier (proven≥0.65/neutral/unproven/unrated); smart_money_share = proven share | per-author weighted_social_activity, author credibility |
| painting_attribution pipeline.organic | per-tell weighted contributions (Σ == manufactured_probability) + dominant_factor + banded label + reasons | dup_ratio, burst_ratio, new_account_ratio, author_concentration |
| quality_adjusted_density pipeline.quality_adjusted | attention_density × organic_score (organic clamped to [0,1]); organic = 1 − manufactured_probability when not given directly | attention_density, organic_score |
| attention_concentration pipeline.concentration | gini + hhi (Σ share²) of per-author WSA; effective_authors = 1/hhi | per-author weighted_social_activity |
| cross_source_corroboration pipeline.corroboration | material sources (share ≥ 0.15) across platforms; effective_sources = 1/Σ(source share²); is_corroborated ≥ 2 material | per-source weighted_social_activity |
| attention_integrity pipeline.integrity | weight-renormalized mean of organic, dispersion (1−gini), corroboration (eff_sources/3), persistence (clamp autocorr); None < 2 components | organic_score, attention_concentration, cross_source_corroboration, attention_persistence |
| Measure | Formula | Inputs |
|---|---|---|
| author_lead_credibility pipeline.reputation | decay-weighted Σ earliness·tanh(forward_return/scale) / Σ recency; earliness = lead_days/(lead_days+2); None < 5 positive-lead calls | author calls (lead_days, forward_return, days_ago) |
| reputation_weighted_density pipeline.reputation_weighted | attention_density of Σ author_wsa·(0.5+credibility); None until any contributing author has a known credibility (credibility_coverage) | per-author weighted_social_activity, author credibility, market_cap |
| author_ring pipeline.author_network | mean pairwise co-occurrence lift among a spike's authors; suspicious ≥ 3× (coordination ring) | spike author sets (archive) |
| independent_voices pipeline.independent_voices | Sybil-resistant breadth: effective_voices = n/(1+(n−1)·ρ̄) where ρ̄ = mean pairwise Jaccard co-appearance; + independence_ratio + coordinated flag | spike author sets (archive) |
| Measure | Formula | Inputs |
|---|---|---|
| divergence pipeline.divergence | classify(velocity_z, price_return) → ATTENTION_LEADS/REVERSAL_RISK/NOT_SOCIAL/… | velocity_zscore, price_return |
| volume_confirmation pipeline.volume_confirmation | cross attention_z × relative-volume_z → confirmed/froth/silent_volume/quiet | attention z (velocity), relative-volume z |
| swing_divergence pipeline.swing_divergence | trend(attention) vs trend(price) over a window → bearish/bullish divergence / confirmation; strength = corr(attention, price) | attention series, price series |
| contagion_lead_lag pipeline.contagion | argmax_k corr(leader[t], follower[t+k]), k≥0 | WSA leader series, WSA follower series |
| Measure | Formula | Inputs |
|---|---|---|
| social_surprise pipeline.surprise | actual_WSA − expected_WSA(baseline_mean, size_floor) | weighted_social_activity, baseline, market_cap |
| event_attention_surprise pipeline.event_surprise | (realized − mean(prior same-event attention)) / std(prior) | realized event attention, prior-event attention history |
| Measure | Formula | Inputs |
|---|---|---|
| stratified_hit_rate pipeline.hitrate | hits/n per lifecycle_state & divergence_signature; Wilson lower bound | forward outcome (hit bool), lifecycle_state, divergence |
| Measure | Formula | Inputs |
|---|---|---|
| breakout_conviction pipeline.conviction | stack abnormal × volume_confirmed × trustworthy × low_pump_risk → high/medium/low/unknown tier + reasons | attention_anomaly_band, volume_confirmation, attention_integrity, pump_risk_score |