Skip to main content
← Back to Docs
✨AI Tools

AI Market Match

Paste a headline, find which markets it moves and how

## What AI Market Match does

AI Market Match turns a piece of news into a ranked list of tradeable markets. You paste a headline, a tweet, a Fed statement, an earnings line, or a full paragraph of context, and the AI reads it against the current universe of active prediction markets, then tells you three things for each market it considers affected:

- **Which markets** the news touches (by question and market ID, with a direct link into Predite's market detail page and out to Polymarket).

  • •**Expected direction** — `YES` if the news pushes the probability of the YES outcome up, `NO` if it pushes it down.
  • •**Confidence** — a 0–100% score for how strongly the model believes the connection and the direction.

The point is speed. When a headline drops, the edge is gone in minutes. Manually scanning hundreds of markets to figure out which ones moved — or are about to — is exactly the kind of work that loses you the trade. AI Market Match collapses that search into one paste-and-go step so you can spend your time on the part that actually matters: verifying the thesis and sizing the bet.

It lives at **/dashboard/ai-match** and is built for fast, repeatable triage during live news flow. It is not a signal service and it is not a recommendation to trade — it is a search-and-direction layer on top of the live market list. The output is a starting point you confirm, not an answer you act on blindly.

### Who it's for

- **Event traders** reacting to macro prints (CPI, FOMC, jobs), elections, court rulings, sports, and crypto headlines.

  • •**Anyone with a thesis but no ticker** — you know the news matters, you just don't know which of the hundreds of live markets it maps to.
  • •**Researchers** building a watchlist around an unfolding story.

## How it works under the hood

Knowing the mechanics helps you write better inputs and read the output correctly.

1. When you submit, Predite pulls the **top active markets ranked by volume** (the most liquid, most-traded markets at that moment). 2. It sends your text plus that market list to the model, which is asked to act as a prediction-market analyst. 3. The model returns **up to 5 matches**, sorted by confidence (highest first), each with a direction and a one-to-two sentence reason. 4. Predite re-attaches live data to each match — the **current YES price**, market volume, and a slug for the outbound Polymarket link — and renders the ranked cards.

Two consequences follow directly from this design, and both matter:

- **The model only sees liquid, active markets.** If the news affects a thin, obscure, or brand-new market that isn't in the top-volume set, the AI literally cannot return it. For deep-cut or long-tail markets, use the **EV Scanner** and **AI Probability** tools directly on the market itself.

  • •**It returns at most five matches.** This is a triage tool, not an exhaustive index. The five you get are the model's highest-confidence picks, not the only markets the news could conceivably touch.

## Using it step by step

1. Go to **/dashboard/ai-match**. 2. Paste your text into the headline box. Minimum **10 characters**, maximum **2,000** — enough room for a full paragraph of context, not just a one-liner. A live character counter shows how much room you have left. 3. Click **Find Markets**. Results usually return within a few seconds. 4. Read the ranked cards top to bottom. Card `#1` is the model's highest-confidence match. 5. For any card worth pursuing, click **View market** to open the full Predite market detail page (price history, EV, AI probability, order book depth), or click **Trade on Polymarket** to jump straight to the venue. 6. Before you place anything, **read the resolution criteria** on the market page. This is the single most important step — see the gotchas below.

### Writing inputs that get good matches

The quality of the output is mostly a function of the quality of the input. A bare three-word headline gives the model very little to reason from.

- **Add the specifics.** "Fed cuts rates 50 bps in surprise move, Powell signals two more cuts in 2026" beats "Fed news." Numbers, names, dates, and magnitudes all sharpen the match.

  • •**Include the surprise, not just the event.** Markets price expectations. "CPI comes in at 3.1% vs 3.4% expected" tells the model the *direction of the surprise*, which is what actually moves probability. "CPI released today" tells it nothing tradeable.
  • •**One story per query.** If you paste two unrelated headlines, the model splits its attention and confidence drops across the board. Run them as separate queries — you have plenty of quota.
  • •**Paste a paragraph when you have one.** The model handles up to 2,000 characters. A few sentences of context routinely produce better direction calls than a clipped headline.
  • •**Use plain English.** Keep tickers and proper nouns (Polymarket, Powell, BTC, SCOTUS), but you don't need to format anything. Natural language works best.

### Reading the results

Each result card shows:

- **Rank** (`#1`, `#2`, …) — ordered by confidence, highest first.

  • •**Direction badge** — a teal **YES** or red **NO**. This is which side the news favors, not which side is currently cheap.
  • •**Confidence %** — colour-coded so you can scan it at a glance: - **Green (≥ 70%)** — the model is confident in both the link and the direction. - **Amber (40–69%)** — plausible but uncertain. Treat as a lead to research, not a conviction call. - **Grey (< 40%)** — weak. The model is flagging a possible connection it isn't sure about. Usually not actionable on its own.
  • •**YES now** — the live YES price in cents. Pair this with the direction to judge whether there's still room. A `YES` call on a market already trading at 96¢ has very little upside left even if the model is right; the move may already be priced in.
  • •**Volume** — rough liquidity, so you know whether you can actually get filled at size.
  • •**Reason** — the model's one-to-two sentence rationale. Read this critically. If the reason misreads the resolution question or relies on a fact you can't confirm, distrust the direction no matter how high the confidence.

A practical way to triage: scan the **direction + confidence + the gap between the YES price and where you think it should be**. The best opportunities are high-confidence calls where the live price hasn't moved yet. If price already reflects the news, the edge is gone — confidence tells you the model agrees with the market, not that there's money to be made.

## The rate limit

AI Market Match is capped at **20 queries per hour**, per user. The limit is enforced as a rolling window (it tracks the last 60 minutes, not a clock-hour that resets on the hour), so heavy bursts during a busy news window will throttle before light, spread-out usage does. Hit the cap and you'll get a clear "rate limit reached" message; wait for the window to roll forward and you're back in.

Each query runs a fresh model call against the live market list, which is why there's a ceiling rather than unlimited use. Twenty per hour is generous for genuine event-driven triage — if you're brushing the limit, you're likely re-running near-identical headlines. A few tips to stay well under it:

- **Don't spam tiny variations** of the same headline. Write one good, specific input instead of five vague ones.

  • •**Batch your context into a single paste** rather than firing the headline, then the subhead, then a quote as three separate queries.
  • •**Note that the limit can be adjusted per plan.** The 20/hour figure is the standard cap; your effective limit is whatever is configured for your plan tier, and the in-app rate-limit message always states the exact number that applies to your account.

## Gotchas and caveats

The matches are heuristics. Verify before you trade. This is the most important thing to internalize. The AI is pattern-matching news against market questions, not running a calibrated forecasting model with a track record. It can:

- **Get the direction backwards.** Resolution wording is full of double negatives, "before [date]" clauses, and "by which" thresholds that flip a YES into a NO. A headline that sounds bullish for an event can resolve the market the *other* way depending on exactly how the question is phrased.

  • •**Miss the real driver.** It reasons from the text you give it. If the market actually hinges on a detail your headline omits, the AI won't account for it.
  • •**Surface a stale or already-priced move.** The model doesn't know what the market did in the last five minutes. A high-confidence YES on something already trading at 94¢ is information, not an edge.
Always read the resolution criteria yourself. Before placing anything, open the market and confirm: What exactly resolves this YES? By what date? Who decides, and on what source? A surprising share of "the AI got it wrong" cases are actually "the resolution criteria didn't say what the headline implied." The model's reason field is a hint, never a substitute for reading the actual market terms.
Empty results are a real answer. If the model returns no matches, it's telling you the news doesn't clearly map to any of the current top-volume markets. That's often correct — not every headline is tradeable. Try adding context or rephrasing once; if it's still empty, the market for your thesis may simply not exist yet, or may be too thin to surface.
Confidence is the model's, not yours. A 90% score means the model is confident, not that the trade is 90% to win. Do your own sizing — pass the trade through the Kelly Calculator rather than betting big just because the badge is green.

## How it fits the rest of Predite

AI Market Match is the front door of a workflow, not a destination. The natural path:

1. **AI Market Match** → find the affected markets and a directional read. 2. **Market detail / EV Scanner / AI Probability** → confirm the edge. Check the AI-estimated fair probability against the live price, read the order book, and look at price history to see whether the move already happened. 3. **Kelly Calculator** → size the position based on your real edge and bankroll, not gut feel. 4. **Paper Trading** → if you're testing a new event-trading approach, log the idea on paper first and review whether your AI-Match-sourced trades actually print over time in the **Trade Journal**. 5. **Live Trading** → execute for real. Live order execution via the Polymarket CLOB requires the **Bot plan**; the **TWAP Orders** tool is the right way to enter size on a thinner market without moving the price against yourself.

It also pairs naturally with the **News Calendar** (know *when* market-moving events are scheduled, then run AI Market Match the moment they land) and **Notifications** (get pinged on the events you care about so you're at the keyboard when the headline drops).

### Plan requirements

- **AI Market Match** is included on the **Pro** ($59/mo) and **Bot** ($99/mo) plans. It is **not** available on **Starter** ($29/mo) — Starter is limited to the core EV Scanner. If you're on Starter and hit the upgrade prompt, moving to Pro unlocks it along with the rest of the AI toolset (Chat, Summaries, Whale Tracker, and more).

  • •**Live order execution** — actually placing the trades AI Market Match surfaces — requires the **Bot** plan. On Pro you can research, paper trade, and journal everything; turning a match into a real position on-venue is a Bot-tier capability.

Used well, AI Market Match is the fastest way to go from "this news matters" to "here are the three markets to look at and which way they should move" — leaving you the time to do the verification that actually makes the trade profitable.

## Related Docs

- [EV Scanner](/docs/ev-scanner)

  • •[AI Probability](/docs/ai-probability)
  • •[Kelly Calculator](/docs/kelly-calculator)
  • •[News Calendar](/docs/news-calendar)
← All Documentation
AI Market Match | Predite