Tuning guides
The Component vocabulary page explains what each named component is for. These pages go one level deeper: what each individual knob does, which way raising it moves the strategy, and who is allowed to turn it. They are the reference you keep open on a second screen while you tune a draft on the Tune canvas.
Parachute is software you run on your own Kalshi account — you build and tune the algorithm; your bot executes it against your own funds. Nothing here is financial advice or a prediction of results, and every strategy can lose money. See Money safety.
How to read these guides
Each component gets a short recap, then a dial table. The table always has the same shape:
| Column | What it means |
|---|---|
| Dial | The knob's label as it appears on the canvas, and its underlying config key in mono |
| Turning it up… | The direction of effect — what raising the value does to the strategy. Turning it down does the opposite. |
| Range | The values the field accepts (from the strategy schema), the unit, and the "off" value where one exists |
Two conventions run through every table:
- A knob's effect is described as a direction, not a recommendation. We tell you that raising the Min Score makes entries rarer and more selective — we do not tell you the right number. The right number depends on your risk appetite and, above all, on evidence from a backtest. Where the schema itself carries a suggested starting value (a few fields do), we repeat it and label it as a suggestion.
- "Off" is a real setting. Many filters and exits are disabled by a sentinel value — a
0threshold, a99¢price cap, an empty toggle. The tables call these out so you don't arm something by accident, or leave something dormant when you meant to switch it on.
Robustness over knife-edges
The single most useful habit when tuning: prefer a value that works across a range of nearby settings over a single number that only looks good on one historical window. A dial that has to be exactly 0.037 to work is almost always overfit — it has memorized noise. Adding more filters and turning more knobs is not automatically better; each active dial is another way to fit the past instead of the future.
Who is allowed to turn each knob
Every component belongs to one of three ownership tiers, and the tier decides who can change it once a strategy is published and sold:
- Platform substrate — fixed safety infrastructure Parachute owns. Not a tuning surface at all (the Execution backbone, the Daily-Loss Breaker).
- Builder — the Model and the Filters. These define what the strategy is. You tune them freely while building; when you publish, they lock as one unit, so someone who buys your strategy runs them but cannot re-tune them.
- Buyer dial — sizing, the exit brackets, and the hedge overlays. These are meant to be adjusted for an individual account and risk appetite, even on a strategy you bought.
Each guide labels its component's tier up front.
The guides
| Guide | Components covered | |---|---| | Model & signal | Fair-Value calculators (Gaussian / Jump-Diffusion), ITM Probability Tilt, Momentum Bias | | Filters | Confidence Bands, Trend Filter, Net-Edge Filter, Dynamic Edge Floor, Spread Filter, Max Entry Price, Overpriced-Edge Guard, Volatility-Regime Filter, Moneyness Filter, Entry Window, Fast-Move Filter, Counterparty Filter, Volatility Override | | Sizing | Fractional Kelly, position and exposure caps, the fixed-cash ticket-budget shape | | Exits | Exit Bracket (take-profit / stop-loss), Tail Hedge, Hedge Ladder |
A reminder about structural changes
Turning a builder knob (Model, any Filter) or adding/removing a component changes what the strategy is, so it deploys through the evidence gate — a linked backtest, or an audited "deploy untested" override. Turning a buyer dial (sizing, an exit bracket) does not. Every guide notes where a change is structural.