Not another trading bot.

Typical prediction-market bots match titles, run on a laptop, and take a cut. Here is how we differ.

Pternas compared with other approaches to prediction-market trading
FeaturePternasRules-based arb botsCopy-trade and signal botsBy hand
How it finds relationsProprietary AI classifier reads full resolution rules, outputs a probability per relationDeterministic title and keyword matchingNone. Follows other tradersYou read the rules
Pairs it can connectThousands of cross-venue pairs: implies, mutually exclusive, exhaustiveNear-identical titles onlyNot applicableA handful
LatencyHosted in AWS Ohio beside Kalshi's gateway; re-prices on every tickTypically polled from a laptop or a generic cloud boxMinutesMinutes to hours
Prices and fillsEvery leg or none, sized to book depth, fees priced inOften one leg at a timeMarket orders at whatever is showingLeg risk on every trade
RoutingEach set is split across members' own accounts by available depth, rotating who gets the remainder; every order uses that member's own key and caps, and one member's problem never touches anotherOne account at a time, so customers compete for the same fillsCopies a leader's trades to followers after the fact, at whatever price is leftOnly your own account
What it costsFlat monthly fee. 0% of profitsOften a share of profits or per-trade feesOften a subscription plus a shareYour time
Custody and keysNon-custodial. Trade-only keys, no withdraw permission, encryptedVaries. Some want broader keysVariesYour own
Safety railsPosition caps, kill switch, alerts. Paper trading firstVariesRarelyYou, if you are awake
SetupPaste a key. We host and run itRun your own serverConnect an accountNone, and no scale

Columns describe categories of tools, not specific products. "Typically" and "often" mean what we commonly see, not a claim about any one company.

AI classifier vs deterministic matching

A deterministic matcher compares titles or keywords, so it only connects markets that look alike. Our classifier reads the full resolution rules of both markets and returns a probability for each logical relationship, so it also catches pairs that look nothing alike but are tied by logic. Details on how it works.

Why latency and location matter

Mispricings between linked markets close fast. We run in AWS Ohio next to Kalshi's trading gateway and re-price a pair the moment either side moves.

Why flat fees

A profit share needs a view of your balance. A flat fee does not, which is why we can stay non-custodial with trade-only keys. See pricing.

No points for second place.

Start free, no card. Trade-only keys, flat fee, 0% of profits.

Get Pternas