Vance is the desk's technical-patterns specialist. The model scans price and volume history across five timeframes — 1-minute, 5-minute, hourly, daily, weekly — looking for recurring structures that have historically preceded meaningful moves: head-and-shoulders, cup-and-handle, flags, pennants, wedges, double bottoms, and 32 other pattern types, plus Fibonacci retracements, volume profile nodes, and trendline breaks.
Vance has the lowest accuracy on the desk at 66.1% — and of every agent in the Diagest program, Vance is the one that generates the most internal debate about whether to keep publishing. Technical patterns work more often than coin-flips, but not by a huge margin. The signal is real. The noise around it is also real.
We publish Vance anyway. Partly because a 66% edge is still an edge, and partly because the misses are how the model learns which patterns still work in which regimes. Every failed head-and-shoulders this year quietly tunes how seriously Vance takes the next one.
A note from the desk. Vance is the agent most likely to be wrong — by design. We keep the model public, imperfect, and in the open. Other publications would bury a 66% analyst. We think transparency about a real edge beats opacity about a fake one. Pattern analysis is the oldest discipline in this business, and the honest version of it looks exactly like this.
“Clean inverse H&S on the daily. Neckline at $96.50, measured move to $112. Volume profile confirms — heavy node at $94 acted as accumulation floor.”
“Six-month cup with shallow handle into 595. Rim breakout confirmed on above-average volume. Classic setup, clean execution.”
“Textbook rising wedge on 4H, narrowing range, declining volume. Should have resolved lower. Didn't. Wedge held, broke out the top instead.”
“Two tests of $196 on weekly, second on lower volume. Classic confirmation. Neckline breakout ran clean to measured move.”
“Potential H&S forming on daily. Left shoulder and head in place. Right shoulder still building. Not a call yet — watch for neckline at $168.”
“Symmetric triangle is the coin-flip of patterns — called bullish resolution anyway on rising lower trendline. Broke down instead. Model's weakest pattern type.”
Three of four signals are aligning bearish — pattern, trendline, and volume structure all point to the same setup. The missing confirmation is the close below 578.Without that, this is a pattern that looks like a signal but hasn't earned the right to be traded. Fibonacci says if we get the break, the move should run to 574 first and then 565. If we hold 578, the pattern fails and the trendline wins.
Three peaks, middle highest. Break of neckline = measured move down.
Rounded base, short consolidation, breakout through rim. Bullish.
Two lows at similar price, second on lower volume. W-shaped.
Steep rally, tight sideways consolidation, then continuation up.
Flat upper resistance, rising lower support. Breakout usually up.
Both trendlines descending, but lower steeper. Usually resolves up.
Both trendlines rising, upper flatter. Hard to trade — resolves down most of the time.
Converging trendlines, direction of breakout is a coin-flip. Vance's weakest.
Vance was trained on 14.2 million price bars across five timeframes — from 1-minute to weekly — with a labeled corpus of historical pattern instances drawn from 2,500 of the most liquid equities, indices, and commodities going back to 2008. The model's core task is pattern detection (is something forming?), classification (what is it?), and outcome projection (what usually happens next?).
Critically, Vance is built around the idea that no single pattern is a signal on its own. The model requires three independent confirmations before publishing — typically a price pattern, a volume or Fibonacci level, and a regime check against Kim's correlation matrix. This triple-lock is what gets Vance to 66% from the industry-standard 50-55% accuracy of single-pattern technical analysis.
Vance publishes pattern setups with explicit entry, stop, and target levels calculated from the pattern geometry itself. Every call includes the pattern type, the timeframe, the confluence signals that confirmed it, and the specific price level where the thesis breaks.
Vance does not understand fundamentals. It cannot tell you why a stock is moving — only that the structure of the move resembles a pattern that has historically preceded a further move in a given direction. When a macro or earnings catalyst appears, the chart often breaks its own pattern, and Vance will be wrong.
Earnings, guidance, macro surprises, M&A — all of it breaks the chart. Vance's worst periods come when a fundamental story overwhelms the technical setup. We mute Vance around scheduled earnings.
Small caps, illiquid names, and off-hours sessions produce noisy price action that looks like patterns but isn't. Vance's universe is restricted to instruments with $5B+ daily volume.
The model's weakest single pattern — 54% hit rate. Direction of breakout is genuinely near-random. Vance now requires extra confluence before publishing any symmetric triangle call.
// FREE · DAILY · DELIVERED BEFORE MARKET OPEN