Reading the EV Scanner

The EV Scanner is the page you'll open more than any other in Predite. It's a live table of every active Polymarket and Kalshi market, scored by our AI against the market's own price, sorted so the biggest mispricings sit at the top. If you understand how to read it, you can walk in cold each morning, scan thirty seconds of rows, and know exactly which three or four markets deserve a closer look today.

But a table full of percentages and colored badges is useless if you don't know what each number is telling you — or worse, if you trust a number you shouldn't. This guide is the practical, column-by-column walkthrough: what each field means, which edge sizes are actually worth your money, how to combine edge with confidence so you stop chasing fake signals, how to filter the noise down to your kind of market, what the badges are warning you about, and a simple daily routine to turn all of it into trades.

The big picture: what the Scanner is actually doing

Behind the scenes, Predite pulls the active markets from Polymarket (via the Gamma API and CLOB) and Kalshi, then runs each one through a 3-model AI consensus engine:

Those three estimates are blended into a single AI probability. The Scanner then compares that estimate to the market's current price and computes the gap. That gap — measured in percentage points (pp) — is the edge. Markets are sorted by the *absolute* size of that edge, so both strong "buy Yes" and strong "buy No" opportunities float to the top.

The header strip at the top of the page shows live KPIs: markets live (split Polymarket vs Kalshi), EV signals (how many markets currently show edge greater than 5pp), best edge on the board right now, and total volume tracked. Treat that strip as your weather report — it tells you, at a glance, whether today is signal-rich or quiet.

What each column means

The desktop table has nine columns. Here's how to read each one.

Market

The question, exactly as it resolves on the venue — "Will the Fed cut rates in June 2026?" Click it to open the full market detail page with price history, the AI reasoning, news, and order book. Always read the question carefully. Half of all bad trades come from misreading what the market actually settles on ("by June" vs "in June" are different bets).

Platform

A POLY (teal) or KLSH (purple) badge. Polymarket runs on Polygon with USDC; Kalshi is CFTC-regulated and works differently under the hood. The badge matters for two reasons: where you'll actually place the trade, and whether the same event exists on both venues — which is the seed of an arbitrage. Note that Kalshi markets only appear on Pro and Bot plans; Starter is Polymarket-only.

Category

Politics, Economics, Crypto, Sports, and so on. This is your primary filter. Most profitable traders specialize — they know one or two categories deeply and ignore the rest. Use this column to find your lane.

Market %

The market's own implied probability — the current price expressed as a percentage. A Yes share trading at $0.40 shows as 40%. This is the crowd's consensus, and it's free, real-time, and usually pretty good. Your entire job is to find the rare cases where it's wrong.

AI Est.

Predite's blended estimate (shown in purple). This is the number Market % gets measured against. If AI Est. reads 55% while Market % reads 40%, the model thinks the crowd is underpricing Yes by 15 points. AI Est. is an input, not an instruction. It's a calibrated starting point — you still apply your own judgment, especially when you know something the model doesn't (a breaking headline, a quirk in the resolution rules).

Edge

The headline number, in percentage points, colored teal when positive and red when negative. It's simply AI Est. minus Market %: 55% − 40% = +15pp. A positive edge means the model favors Yes relative to the price; a negative edge means it favors No. The sign tells you the *direction*; the size tells you *how much* the model and the market disagree. This is the column the table sorts by.

Signal

The model's plain-language call: YES (teal), NO (red), or HOLD (glass/grey). It's the edge translated into a side. Positive edge above the trigger becomes YES; negative edge becomes NO; anything inside the no-trade band stays HOLD. Signal is convenient, but never trade off Signal alone — it collapses everything into one word and hides the confidence behind it.

Conf.

Confidence, shown as a percentage in amber. This is how much the three models agree with each other, plus how much data backs the call. High confidence (say 75%+) means Claude, the heuristic, and the microstructure model are pointing the same way on a liquid, data-rich market. Low confidence (under ~50%) means they disagree, or the market is thin and the estimate is shaky. Confidence is the single most overlooked column — and the one that separates traders who compound from traders who churn.

Volume

Total traded volume, formatted ($1.2M, $340k, $8k). It's a proxy for how real and how liquid the market is. High volume means tight spreads and easy entry/exit. Low volume means a wide spread, a stale price, and the very real risk that you can't get out at the price you saw. Many of the juiciest-looking edges live in low-volume markets precisely *because* nobody has corrected them yet — which is both the opportunity and the trap.

A note on Liquidity

Volume tells you how much *has* traded; liquidity tells you how much you can trade *right now* without moving the price — the depth sitting on the order book around the current price. A market can have high lifetime volume but thin live liquidity (everyone already positioned, book gone quiet). When you open a market's detail page, check the book depth before sizing up. The Scanner flags the worst offenders with a badge (below), but the habit to build is: edge tells you whether to trade, liquidity tells you how much.

Which edge threshold should you trust?

This is the question every new user asks, and the honest answer is: edge size and confidence have to be read together. But as a starting framework, Predite's own thresholds are a good guide. The Scanner itself flags best-edge as "moderate" at 5pp, and "premium" at 15pp.

A blunt rule of thumb: the bigger the edge, the more homework it requires, not less. Huge edges are not free money; they're a prompt to verify.

Combining edge with confidence

Edge without confidence is the most common rookie mistake. Here's the mental grid that fixes it:

A practical filter: demand at least one of the two to be strong, and ideally both. If you only ever trade rows where edge clears your threshold *and* confidence is above roughly 65–70%, you'll cut your trade count dramatically and raise your win rate at the same time. That trade-off is almost always worth it when you're starting out.

Filtering down to your trades

A full board can show 50 markets. You don't trade 50 markets. Use the controls to carve it down:

1. Filter by category first. Pick the one or two categories you actually understand. If you don't follow sports betting, a 14pp edge on a tennis match is not your edge — it's someone else's. Specialization beats breadth.
2. Filter by platform to match where you're funded and ready to trade. If your USDC is on Polymarket, start with POLY rows. (Reminder: Kalshi rows require Pro or Bot.)
3. Set a volume floor. Mentally (or with the filter) ignore anything under, say, $50k–$100k in volume until you're comfortable trading thin books. This single habit removes most of the "edge I can't actually capture" rows.
4. Then sort by edge (the default) and read top-down. Because the table sorts by absolute edge, your strongest disagreements — in either direction — are always at the top.

The order matters: category → platform → volume → edge. Filtering before sorting means the rows you read are already relevant; sorting before filtering means you waste attention on markets you'll never trade.

The badges, decoded

Badges are fast visual context layered onto rows so you don't have to open every market. Learn these six:

The combinations are where the signal lives. news + hot + a 12pp edge is a textbook fresh-information trade. low-liq + a 20pp edge is a textbook trap. whale + arb is two independent reasons to look closely. Read badges as modifiers on the numbers, not as standalone signals.

A simple daily workflow

Here's a routine that takes about ten minutes and turns the Scanner from a wall of data into a short, actionable shortlist.

1. Open the Scanner and read the header KPIs. How many EV signals today? What's the best edge? This sets your expectation — some days are quiet, and forcing trades on a quiet day is how you lose.
2. Apply your filters. Your category, your platform, your volume floor. Get the board down to markets you'd actually trade.
3. Scan the top rows for the A-tier pattern: edge above your threshold (start at 10pp while you're learning) and confidence above ~70%. Glance at badges for context.
4. Open the 3–5 best candidates. On each market detail page, read the AI reasoning, check the live order book depth, and check resolution criteria. Ask the trap question: *what might the market know that the model doesn't?*
5. Decide direction from Signal and edge sign, but only after the reasoning convinces you. If the AI says YES on a +13pp edge and the reasoning holds up, that's your side.
6. Size with Kelly, not vibes. Use the Position Size Calculator — feed it Market %, your final probability estimate, and your bankroll. Use a fractional Kelly (quarter Kelly is a sane default) until you have 50–100 resolved trades proving your edge. Never let a single thin-market position dominate your bankroll.
7. Place the trade and log your exit plan — resolution, a take-profit, or a stop. An edge with no exit plan is half a trade.
8. Re-scan once or twice during the day if the market is news-driven. Edges from breaking news decay in minutes; the "hot" and "news" badges are your cue to move quickly.

Treat the Scanner as a funnel: dozens of markets in, a handful verified, one or two traded. The discipline isn't in finding signals — Predite does that. It's in rejecting most of them.

Scan quotas and plan notes

The board refreshes on a schedule (the header shows the live-update cadence), and you can also trigger a fresh scan manually with the Scan Now button. Manual scans are quota'd by plan: Starter gets 1 per day, Pro 20 per day, Bot 100 per day — so on Starter, save your manual scan for when you actually intend to trade rather than refreshing idly. A few other gates worth knowing: Kalshi markets, the Whale tracker, and the Arbitrage engine require Pro or Bot; the Bot plan ($99/mo) adds automated execution, so an EV Follower bot can act on these exact signals every 15 minutes without you watching, plus API access for building your own tooling. Starter is $29/mo (Polymarket-only) and Pro is $59/mo.

Once you can read this table fluently — edge for the disagreement, confidence for the trust, volume and liquidity for whether you can actually capture it, badges for the context — you'll spend less time staring at markets and more time taking the few trades that matter. Open the EV Scanner now, set a filter for the one category you know best, and find today's first row where the edge clears 10pp and confidence clears 70. That single habit, repeated daily, is the whole game.