Speed sample figures
How fast this copy of Ping is, and whether that speed cost you anything. Every decision a rule makes is timed, then checked: if this had been a real order at the asking price, would the offer still have been there when it landed?
cost you 9 of 64 decisions were lost to speed: a very fast bot could still have bought, your order would have arrived too late. 8 of the 9 settled would have won, so speed cost you about $0.41 a contract. That is too few to read much into.
Offers still there
| Speed | Your decisions | Lab study |
|---|---|---|
| Very fast bot, 30 ms | 84% | 79% |
| You, 90 ms | 70% | |
| 650 ms | 58% | 59% |
| Human, 2 s | 39% | 36% |
The lab study replayed 39 signals from one rule in October 2026, so read it as a rough guide. Your column is your own rules on your own connection.
What it was worth
one contract a decision, after taker fee| If you had caught every offer a very fast bot could reach | +2.12 |
| Only the offers you would have reached | +1.71 |
| The decisions lost to speed (8 of 9 would have won) | +0.41 |
Dollars, settled decisions only, from the exchange's results. Blame is narrow: a decision counts as lost to speed only if the offer was still there at 30 ms and gone when your order would have landed. This is an estimate: it reads the top of the book, and nothing was actually sent.
How to get faster
- Your connection is uneven
Usually 64 ms, but one sample in twenty takes 210 ms or more. That is typical of Wi-Fi. A network cable usually fixes it, and it is free. - One account, one rate limit
Kalshi's request limit is shared by everything on your account. A second bot on the same key can get your order refused at the moment it matters.
Speed decides which offers you reach. It does not make a losing rule win.