Skip to main content
Back to Docs
EV Scanner

Reading Signals: Every Column Explained

What Market %, AI Est., Edge, and Signal columns mean

The Scanner table is the most-used view in Predite. This doc explains every column, every badge, and every filter so you can read signals fluently.

The Core Columns

When you open the Scanner, you'll see a table with these columns by default:

Market. The market name (e.g., "Will the Fed cut rates in June 2026?"). Click to open the full market detail page. The venue is not shown next to the name — it has its own column, as a text badge reading POLY or KLSH.

Category. Politics, Economics, Sports, Crypto, etc. Use the dropdown filter at the top to narrow by category.

Market %. The current market price expressed as a probability. A share trading at $0.40 shows 40%. This is the "market's belief" you're betting against.

AI Est. Our 3-model AI consensus probability. If we estimate 55% but the market is at 40%, there's a 15-point gap that may be edge.

Edge. The difference between AI Est. and Market %, in percentage points (pp). Positive edge = market underpriced (recommendation: buy YES). Negative edge = market overpriced (recommendation: buy NO). Magnitude matters — see the "what counts as a strong signal" section below.

Signal. Our actionable recommendation:

  • 🟢 YES — buy YES shares
  • 🔴 NO — buy NO shares
  • ⚪ HOLD — no clear edge, ignore

Conf. Confidence level of the AI estimate (50-95%). Higher = more certain. Below 60% = use caution; the model is unsure.

Volume. The market's cumulative traded volume in USD — its whole life, not the last 24 hours. It is a decent read on how much attention a market has had overall, and a poor read on whether anything is happening right now: a market that traded $2M last year and nothing this week still shows $2M.

Liquidity is not a column here. The scanner table doesn't carry it, so use Volume as your rough proxy and check the actual book on the market's Order Book tab — or run the size through the Slippage Simulator — before sizing anything on a small edge.

What Counts as a Strong Signal

Rough rules of thumb:

  • Edge ≥ 5pp: Worth investigating. Open the market, read the resolution criteria, decide if you have a view.
  • Edge ≥ 10pp: Strong signal. Likely worth a small position if you have time to monitor.
  • Edge ≥ 15pp: Rare. Either a real mispricing or a sign that the resolution criteria are tricky (markets are usually efficient enough that 15pp gaps don't last). Verify carefully.
  • Edge ≥ 20pp: Almost always a sign that something is off — the AI doesn't know about a recent news event, the market is illiquid, or the resolution criteria are ambiguous. Don't blindly take these.

Combine edge with confidence:

  • High edge + high confidence (>75%) = strongest signal
  • High edge + low confidence (<60%) = AI thinks there's edge but doesn't trust its estimate — verify manually
  • Low edge + high confidence = consistent +EV at small scale — fine for bots, less compelling for manual trading
  • Low edge + low confidence = noise. Skip.

Filters and Sort

Above the table, the filter bar lets you narrow signals:

  • Platform: All, POLY or KLSH
  • Direction: All, YES or NO
  • Edge > 5pp: a single on/off chip at a fixed threshold — not a slider, and off by default
  • Category: all categories or one

Those four are the entire filter bar. There is no confidence filter, no volume filter and no resolution-window filter; screen for those by reading the columns.

The table sorts on click: every column header is a button, first click descending, second ascending, third back to the default. That default is absolute edge, descending, computed on the server. Note the *absolute*: a strong NO at −18pp sorts above a weak YES at +6pp, so the top of the list is "biggest disagreement with the market", not "best YES" — and sorting by Edge keeps that convention.

Badges and Pills

Markets can have additional badges next to the name:

  • POLY / KLSH — which venue the market lives on
  • Category — the market's category
  • YES / NO / HOLD — the signal itself
  • Edge in pp — teal when positive, red when negative

That is the complete set. There are no hot, whale, news, low-liquidity, arbitrage or closing-soon badges — for those you want the Whale Tracker, Arbitrage and Resolution Tracker pages, which is where that information actually lives.

Reading the Edge Visually

Edges are color-coded for fast scanning:

  • Green: Positive edge → buy YES
  • Red: Negative edge (means AI says market is too HIGH) → buy NO
  • Gray: Edge under your "min edge" threshold → noise

The colour is binary, not a gradient: teal for any positive edge, red for any negative one. A +0.4pp edge and a +20pp edge are exactly the same shade, so read the number rather than the colour.

Common Mistakes Reading Signals

  • Treating "Signal" column as gospel. It's a suggestion based on a model. Always read the resolution criteria. The model can miss obvious things (e.g., a market about "Will X happen by date Y" where Y already passed — model says edge 30%, actually it's resolved).
  • Ignoring confidence. Edge of 15pp with confidence 50% is much weaker than 8pp with 90% confidence.
  • Trading every signal. With 50+ markets active, you'll see signals all day. Pick the ones with the strongest combination of edge + confidence + liquidity, not every single +EV blink.
  • Ignoring volume. A 15pp edge on a $500-volume market is nice in theory but you may not be able to enter $200 without moving the price 5pp against you.

Related Docs