spread_capture

This page covers spread_capture, one of the five built-in strategies, on Kalshi, Polymarket and Polymarket US: how it quotes, each parameter, venue differences, backtests and an example. Rules all five share, such as order sizes and how to write params, are on the Built-in Strategies page. Every number here is an illustrative example, not a suggested setting.

spread_capture quotes both sides of a market around the mid. Each loop, for each market:

1. If your position is already at risk.max_position or more, it skips quoting. Quotes already resting stay where they are.
2. It reads the best YES bid and YES ask. On Kalshi the ask is 1 minus the best NO bid.
3. If the book is one-sided, or the ask minus the bid is less than spread_floor, it cancels its open quotes and posts none.
4. Otherwise it cancels its open quotes and posts order_count levels. Level i buys YES at mid − i × spread_floor/2 and sells YES at mid + i × spread_floor/2.

Note: Each refresh cancels only the Bot's own quotes. Your manual orders and other Bots' orders in the market stay resting. On Kalshi and Polymarket US they still count toward Max open contracts.

How the Bot knows which orders are its own:

• Kalshi: the Bot recognizes its orders by this deployment's client order id. When you redeploy a Bot in place, the deploy asks you to adopt the market's resting orders; the previous deployment's quotes you adopt are refreshed like the new ones, so its old ladder doesn't sit at stale prices. An adopted manual order is still never cancelled.
• Polymarket and Polymarket US: orders carry no client id, so the Bot remembers the id of every order the exchange accepts from it. That memory survives restarts and redeploys, but not stopping the Bot, so quotes a stopped Bot left behind stay resting until you cancel them.

spread_floor

Quotes are centered on the mid, not on the edge of the book. When the spread is wider than spread_floor, the first level sits inside the current spread. Prices are rounded to whole cents on Kalshi, to the market's tick on Polymarket (buys round down, sells round up), and to 2 decimals on Polymarket US.

A buy quote below price_floor is raised to the floor, and a sell quote above price_ceiling is lowered to the ceiling. If the floor sits above the ask, the raised buy can trade immediately unless post_only: true.

order_count

Each level is one YES buy and one YES sell. Orders go out one at a time. On Kalshi and Polymarket US, a quote refused by a deployment risk limit ends that loop's ladder.

• The open-contract cap, the lower of max_position and Max open contracts, counts every resting quote, so at 3 a Kalshi or Polymarket US ladder stops after about 2 levels. Raise both max_position and Max open contracts to about 2 × order_count.
• On Kalshi the ladder sends up to 2 × order_count orders per market per loop, and the order budget is 20 per 60 seconds. With loop.interval: 10 and order_count: 3 that is 36 orders a minute, so some loops post no quotes.
• On regular Polymarket each buy quote is the market's minimum size (often 5 shares), whatever max_position is, so a fill can take you past max_position.

spread_capture by venue

On Kalshi, YES sell quotes are never blocked by risk.max_entries_per_market or by the Max open contracts check, because both apply only to buys. Resting sell quotes still count toward the open-contracts total that the next buy is checked against.

spread_capture backtests

Kalshi spread_capture backtests use a resting-order replay, which differs from live in a few ways:

• Quotes are always post-only.
• Each level is modelled as a 1-contract YES bid and a 1-contract NO bid, and max_position caps YES and NO holdings separately.
• When the mid is outside price_floor to price_ceiling, the backtest skips quoting; the live Bot clamps its prices instead.

Regular Polymarket rejects spread_capture backtests, and Polymarket US has no backtests. See Backtest.

spread_capture example

Illustrative only: a one-level ladder on a single Kalshi game market. It quotes only when the spread is about 4 cents or more (an exactly 4-cent spread can miss by rounding), and stops refreshing once 3 contracts are held. With a mid of 0.50 it posts a buy at 0.48 and a sell at 0.52. The one-level ladder fits the deployment risk limits Studio pre-fills. The ticker is an example; saving a strategy doesn't check that the market exists.

version: 1
platform: kalshi
strategy: spread_capture
market:
  ticker: "KXNBAGAME-26NOV04LALBOS-LAL"
risk:
  max_position: 3
  price_floor: 0.10
  price_ceiling: 0.90
loop:
  interval: 15
params:
  spread_floor: 0.04
  order_count: 1
  post_only: true