Dalton is the desk's sentiment and positioning specialist. The model ingests 17 separate positioning feeds — CFTC Commitment of Traders, 13F filings, AAII retail surveys, Investors Intelligence newsletter survey, equity put-call ratio, SKEW index, retail flow data, prime broker net leverage, ETF creation-redemption activity, and social-media sentiment scraped from the major platforms — and synthesizes them into a live map of where positioning is extreme.
The model is explicitly contrarian by design. Not permanently bearish, not permanently bullish — but calibrated to treat extreme consensus as a fading signal. When 85% of AAII respondents are bullish and COT speculators hit multi-year long extremes in the same week, that is historically a better signal that the rally is over than any earnings print or Fed decision.
Dalton has the second-lowest accuracy on the desk at 64.5%. That's structural: contrarian calls take longer to resolve and are most often wrong in the short term, because the crowd is usually right — until the moment it isn't. The model's value is concentrated in the ~15% of calls where positioning genuinely flips a trend. Those calls tend to be large.
“Spec net long on yen carry hit a 7-year extreme. AAII dollar bullishness at 84%. Capitulation signature. Flagged 10 days before BoJ intervention.”
“AAII bull-bear spread hit +42 while equity put-call ratio collapsed to a 2-year low. Everyone in the pool. Called a correction; got one.”
“COT spec short crude at 18-month extreme. Energy ETF outflows 4 consecutive weeks. Pessimism peaking. Classic contrarian buy.”
“Similar bull-bear extreme as later hit. Fired too early — positioning was extreme but momentum stayed positive for five more weeks. Stopped out, re-entered later.”
“Retail call buying in mega-cap tech hit 9-month peak last week. 0DTE call volume up 3x in NVDA. Not capitulation yet — but the setup is forming.”
“Gold COT net long at multi-year low while spot held up. Positioning washout. Correct on direction, slower to develop than modeled horizon assumed.”
Dalton is built around a simple, uncomfortable idea: most of the time, the crowd is right. Markets trend. Consensus is usually an accurate read of the near future. Contrarian-for-contrarian's-sake is a losing strategy. The model spends most of its life watching positioning drift and doing nothing.
What Dalton is looking for is the minority of moments when consensus becomes so one-sided that the market runs out of incremental buyers (or sellers). When 85% of survey respondents are bullish and institutional cash levels are at record lows, there's nobody left to bid. When COT speculators are 2σ short and real money is defensive on the same name, there's nobody left to sell. Those are the reversal points. Dalton publishes there.
Dalton publishes positioning extremes, crowded-trade warnings, and capitulation signals. Every call identifies the specific feed(s) triggering, the historical base rate of resolution, and the horizon over which mean-reversion typically plays out.
Dalton cannot time reversals precisely. Extreme positioning can persist for weeks or months — a fact the track record reflects. The model is designed to identify where the reversal will happen, not when. Contrarian calls issued a month too early look identical to contrarian calls issued a month too late, until hindsight sorts them out.
Positioning can stay extreme for months. Selling Nvidia at the first sign of a crowded trade has been a losing strategy for two years. Dalton fires early; the miss rate reflects this.
AAII membership demographics shift. Investors Intelligence audience changes. Survey-based signals should be compared against their own history, not the absolute numbers. The model does this — but the drift is real.
Sometimes consensus is right because the world has actually changed. Selling every AAII bullish extreme since 2009 would have been a bad strategy. Dalton requires multi-feed confluence to publish; false signals persist anyway.
// FREE · DAILY · DELIVERED BEFORE MARKET OPEN