Compare

Chart Library vs TrendSpider — pattern recognition vs cohort outcome intelligence.

TrendSpider draws shapes on charts and tells you what those shapes are: head-and-shoulders, ascending triangle, double bottom, etc. It’s a pattern recognition product.

Chart Library answers a different question: what did the cohort of historical analogs to this exact pattern do over the next 1, 5, and 10 days, with full forward-return distribution and calibrated probability bands? It’s a cohort outcome intelligence engine.

These are complementary, not competing. TrendSpider names the pattern; Chart Library tells you what those patterns historically did.

What each product gives you

TrendSpider

Automatic shape recognition (named patterns)
Multi-timeframe charting with overlays and indicators
Backtesting strategies with TA rules
Watchlists, alerts, screener
Best for: discretionary traders who want labeled patterns and chart-side automation

Chart Library

Cohort intelligence: 300 historical analogs to any (symbol, date, timeframe) anchor, with full forward-return distribution
Conformal-calibrated probability bands (80% nominal coverage hits 80% empirical)
Feature attribution showing what separated cohort winners from losers (regime, sector, news context)
API + MCP server for AI agents — designed to be a primitive consumed by Claude / Cursor / custom agents
Best for: AI agent builders, quants, sophisticated traders who want methodology-honest distributions instead of labeled shapes

The substantive difference

TrendSpider says: “NVDA chart matches a bull flag pattern.”

Chart Library says: “This NVDA chart pattern’s 300 closest historical analogs had a 5-day median return of -1.3%, p10/p90 of -11.3%/+6.8%, win rate 44%, and the strongest positive feature was tight credit spreads (currently tight).”

The first is a label. The second is a probability distribution + explanation.

TrendSpider’s named pattern doesn’t tell you whether this particular bull flag is one of the 60% that resolved up or the 40% that didn’t. It also doesn’t tell you which regimes the historical bull flags worked in. Chart Library does.

Why they're complementary

For a discretionary trader: use TrendSpider to identify the pattern visually, then use Chart Library to look up what historical analogs of that pattern did next. The named pattern is a useful shorthand; the cohort distribution is the probability information you need to size a position.

For an AI agent builder: TrendSpider doesn’t expose itself as an MCP server, so your agent can’t directly call “name this pattern.” Chart Library’s MCP server is built for agent consumption. If your agent needs pattern shape labeling, do that in your own logic; if it needs cohort outcome intelligence, call Chart Library.

The methodology divide

TrendSpider is built around classical technical analysis. Named chart patterns (head-and-shoulders, double bottoms, etc.) come from the TA tradition. Whether they have predictive power empirically is a long-running debate; many studies find marginal effects.

Chart Library doesn’t use named patterns. We use self-supervised embeddings of chart shape — the model learns its own similarity function from the data, without committing to any particular pattern taxonomy. The cohort is the 300 shapes most similar in the embedding space, regardless of whether those shapes correspond to any named TA pattern.

This is a deeper methodology difference than it appears. Named patterns are categorical labels imposed on a continuous shape space; the embedding lets the data tell us what counts as “similar.” In practice it captures more nuance and avoids the lookahead problem (named patterns are often only identifiable in retrospect).

When to use TrendSpider

You’re a discretionary trader who wants automated chart annotation
You need integrated charting + alerts + screener in one app
Your strategy is built on classical TA patterns specifically
You want a polished GUI for chart analysis

When to use Chart Library

You’re building an AI agent that reasons about stocks
You want calibrated forward-return distributions, not labels
You’re a quant doing methodology-honest backtests
You want feature attribution + regime stratification
You’re building software that consumes pattern intelligence as an API

Pricing

TrendSpider: $39-149/month depending on plan. GUI-focused product.

Chart Library: free Sandbox (200 calls/day); $29/mo Builder; $99/mo Scale; Enterprise custom. API + MCP only; the web app is a thin demo on top of the API.

Frequently asked questions

Does Chart Library identify named patterns like TrendSpider?: We have a pattern detection layer that recognizes 4 specific patterns (breakout, failed breakout, bull flag, ascending wedge), but it's not the primary product. Named patterns are mostly useful as a UX shorthand; the cohort intelligence is the load-bearing primitive.
Can I get TrendSpider's pattern labels via Chart Library's API?: We expose the pattern detection results in the /patterns/{symbol} endpoint. The labels are: breakout, failed breakout, bull flag, ascending wedge. If you need the broader TrendSpider taxonomy (head-and-shoulders, cup-and-handle, etc.) we don't have that today.
Do TrendSpider's named patterns actually predict returns?: Mixed evidence. Academic studies typically find marginal effects after correcting for multiple-testing and look-ahead bias. The cohort approach sidesteps this debate by not committing to any taxonomy — let the data tell us what 'similar' means via the embedding space.
Can I use both products in one trading workflow?: Yes, and this is the most common usage pattern for sophisticated traders: TrendSpider for visual chart context and named patterns; Chart Library API for cohort outcome intelligence on the patterns you care about.

Try it