Reading Kalshi order flow

What the venue publishes, what can honestly be inferred from it, and why nobody — this site included — can show you a Kalshi wallet.

Search for a Kalshi whale tracker and you will find products advertising one. There is nothing to track. Kalshi's public trade API serves markets, order books, an anonymous trade tape and open interest. It serves no per-user data of any kind. A print carries a ticker, a size, a price, a timestamp, which side crossed the spread, and a block flag. It carries no identity, and nobody can attach one to it.

That is architecture, not a gap better scraping closes. Polymarket settles on a public chain, so any wallet's positions and settled record are readable — which is where every wallet-level tool here gets its data. Kalshi settles on an internal ledger, so its half of this product is aggregate flow, labelled as aggregate.

What is actually public

Market data needs no authentication. Per market: last traded price, executable YES and NO asks and bids with top-of-book sizes, volume, open interest, close time. Per event: nested markets and a mutually exclusive flag. Plus a taxonomy of roughly twelve thousand series, and the tape — up to a thousand recent prints per request.

Open interest is the most misread of those. It rises only when a new position is created, so rising open interest beside lopsided taker flow is positioning while rising volume on flat open interest is churn. But every contract has a long and a short: it is symmetric, and never says which side committed.

The catalog is not the board

Tickers are hierarchical — series, then event, then outcome, so KXITFWMATCH-26JUL19RADKON-RAD splits into a recurring family, one occasion, one leg. Splitting on the first hyphen is what rolls loose prints up into series.

The raw market listing is close to useless for finding what trades. Measured live, it opens with more than sixty thousand auto-generated multi-variant parlay legs: an eight-thousand-row crawl returned exactly two series, zero rows with twenty-four-hour volume, zero with open interest. The open-events feed carries none of them and exhausts at roughly nine thousand events and seventy-four thousand markets — still a catalog rather than a board, because most rungs of a strike ladder never print. So the traded board is derived from the tape: group prints by series, drop any series under three prints, rank by notional-weighted flow, keep the top eighteen.

Taker side, blocks, and the Flow Score

Every print records which side was the taker. The taker crossed the spread; the maker was already resting there — the closest thing to intent an anonymous tape contains. Notional is booked at the price the taker paid: one minus the YES price when the taker bought NO.

One-sided flow keeps markets with at least six recent prints where seventy-eight per cent or more landed on the same taker side. The Flow Score rates one print from one to ninety-nine: thirty-five base, up to forty for its size percentile among that market's own prints, fifteen for a block, ten either way for one-sidedness. It scores the print, never the trader — two prints from the same unknown account are scored independently, which is why it can never become a track record.

Mutually exclusive is not negRisk

Kalshi flags multi-outcome events mutually_exclusive: at most one listed outcome resolves YES. It does not promise the list is exhaustive. Polymarket's negRisk is stronger — exactly one leg resolves YES — and the gap decides what can be locked. Under negRisk both full-set directions are provable: one YES of every leg collects a dollar, one NO of every leg collects n − 1. On Kalshi only all-NO is a lock, paying at least n − 1 and n if nothing listed wins; all-YES could pay nothing, so it is never surfaced there. The flag is also the inclusion test — an event without it is skipped, not inferred from its title.

The fee curve peaks mid-price

Kalshi's published per-contract trading fee is 0.07 × P × (1 − P), charged to the taker: a parabola, zero at both ends, heaviest at the middle — about 1.75¢ at 50¢ against 0.63¢ at 10¢. Polymarket charges no trading fee, and this site still books two tenths of a cent per share there for gas and spread, so at mid Kalshi costs roughly eight times as much. The shape matters more than the level: venue disagreements cluster around even money, exactly where the fee is worst, so the one and two cent gaps a scanner finds most often are not capturable. It is a taker charge, so a resting order that fills pays none of it — hence the maker toggle in the sizing calculator, and hence those fills being selected against you.

Executable ask, not last trade

The last tradetells you something moved. It is not a price you can get, and pricing off it invents edges that do not exist: a longshot that last printed at a penny can be offered at two cents, doubling that leg's cost. Every lock here is priced off the executable ask of the exact contract bought — Kalshi serves the NO ask directly, and it equals one minus the YES bid, verified against the live book. When an ask is missing or not strictly between zero and one, the answer is no live book, never a fabricated lock.

Where the inference stops

• No identity, ever. Nothing here will name a Kalshi trader, because the venue publishes none.
• A block print can be one leg of a spread arranged off-book. Size is not conviction.
• One-sided flow is measured over the sampled tape, not a market's history — a quiet market looks lopsided on few prints.
• A Flow Score is relative to one market's own tape, so scores from two markets are not comparable.
• Nothing here can be forward-tested the way a wallet can: there is no account to follow between trades.
• Feeds degrade rather than blank — the last good snapshot is served on an upstream failure, and the catalog behind market prices can be thirty-five minutes old.

Where this shows up

Market Movers carries the three boards this data supports — volume leaders, block prints by notional, one-sided taker flow under a series heatmap. The Live Sharp Feed merges scored anonymous Kalshi prints into the same tape as attributed Polymarket trades, labelling the difference on every row. Overround Scanner applies the mutually-exclusive rule; Arbitrage Scanner nets both fee models out of cross-venue pairs; Market Browser, Fresh Markets, Event Calendar and Weather Edge read the same catalog crawl; Bankroll & Kelly sizes off price plus that fee curve.

Kalshi deliberately does not appear in Portfolio Tracker, which reads public Polymarket addresses — no public record of an individual Kalshi account exists for it, or any other site, to read. Tools are members-only at $49.99/mo; this page is the open documentation of their Kalshi half.

The Polymarket half is a different world — addresses are public and every settled trade is auditable, which is why wallet-level tracking exists there and cannot exist here. For the venues head to head — fees, resolution, multi-outcome rules, depth — see Polymarket vs Kalshi.

WhaleTracks is not affiliated with, endorsed by, or partnered with Kalshi. Everything here is read from Kalshi's public API; WhaleTracks holds no exchange credentials, never places trades and never holds funds. Fee schedules and API behaviour change — confirm current details with the venue before sizing on them. Informational analytics, not financial advice.