Component vocabulary

A strategy is assembled from a shared set of named components. This page is the conceptual reference for that vocabulary: what each component is for and when to reach for it. For the dial-by-dial guide to every knob — what each does, which way to turn it, and who may turn it — see the Tuning guides. For the canvas surface where you actually drag and tune them, see Strategy components; for the deeper editor, the Strategy reference.

Parachute uses industry-standard names on purpose, so a strategy reads the same way it would on other algorithmic-trading platforms.

The three stages

Every strategy is the same pipeline in three stages. The archetype (carry, arbitrage, mean reversion, …) is defined by which components fill the stages — not by having exotic stages.

| Stage | Its job | |---|---| | Model | Price each candidate contract and compute the edge | | Entry rules | Decide whether a candidate is worth trading (the Filters), and how big to size it | | Exit rules | Decide how and when a position is closed or hedged |

Two things that are not part of the strategy itself, by design:

  • Sizing is an Entry property — how big a position gets is decided inside the Entry stage, not as a separate strategy.
  • Account-level caps live in the cockpit, not the strategy. The daily-loss limit and the trading schedule are set in Bot Control, so they apply no matter which strategy is deployed. See Money safety.

Model stage

The Model prices each contract and turns that price into a signal.

  • Fair-Value Model — the pricing law in the model slot; it outputs the estimated settle probability. Today's models are Jump-Diffusion (prices in jump and tail risk — the live default) and Gaussian (GBM) (a classic log-normal law with no jump term). Choosing a different model is a structural change and goes through the evidence gate.
  • Edge — the gap between the model's fair value and the market price. This is the raw signal every downstream rule acts on. (Shown on the canvas as Edge vs Market.)
  • Momentum Tilt — an optional nudge to fair value toward recent price momentum. Off by default. (Shown on the canvas as Momentum Bias.) Reach for it only if you want the model to lean with a short-term move.
  • ITM Probability Tilt — an optional nudge that raises the estimated settle probability for contracts already deep in-the-money near settlement. It is a model knob, and it is not the Moneyness Filter below — the two are deliberately kept distinct and never displayed as a bare "ITM."

Entry rules

Filters

Filters are the admission chain: each one can reject a candidate, and a trade is only taken if it clears every filter that is switched on. Adding or removing a filter is a structural change (it changes what the strategy is), so it goes through the evidence gate. Use them to make a strategy selective — most of a strategy's durable edge lives here, not in the raw signal.

  • Confidence Bands — admits trades by probability band, each band carrying its own edge tolerance. The always-on backbone filter. (Formerly "Probability Bands.")
  • Trend Filter — blocks trades that fight the current market move. Reach for it when your signal should align with, not lean against, momentum. (Formerly "Directional Guard.")
  • Fast-Move Filter — blocks a new entry that would trade against a fast recent price move. Like Confidence Bands it is a fixed filter — always in the chain, not add/removable — but it ships dormant. It watches a trailing window of price and, when that window's high-low range crosses a threshold, rejects entries pointed against the move; an entry in the same direction as the move is never affected. Two dials control it: a range threshold — the fraction of price movement that trips the filter, set to 0 to disable it (and it ships disabled) — and a window — how many minutes of recent price the range is measured over. It applies to normal-mode entries.
  • Net-Edge Filter — scores whether the edge is actually capturable after costs — the bid-ask spread, thin books, and time pressure — rather than just large on paper. Not a second signal; a cost check. Use it so you only trade edges that survive real execution. (Formerly "Micro-Edge.")
  • Dynamic Edge Floor — raises the required edge when spreads are wide and relaxes it when books are deep, so the bar adapts to conditions instead of being one fixed number.
  • Volatility-Regime Filter — slows or pauses trading in abnormal volume/volatility regimes. Reach for it when a strategy is safe in calm conditions but exposed through turbulent ones. (Formerly "Volume Regime Guard.")
  • Spread Filter — rejects a market when the bid-ask spread is too wide to trade into cleanly.
  • Max Entry Price — a price ceiling; never pay above it for a contract.
  • Moneyness Filter — restricts entries to contracts at or beyond a moneyness threshold (how far past the strike the price already is). This is an instrument-selection filter and is the defining component of the Deep-ITM strategy — see Deep-ITM tuning. Distinct from the Model-stage ITM Probability Tilt.
  • Overpriced-Edge Guard — rejects edges that are "too good to be true." An implausibly large edge is usually a stale quote or a miscalibrated model, not a bigger opportunity, so this caps edge from above. Counterintuitive but load-bearing: raising it does not mean bigger profits. See Maximum edge ceiling. (This is a Parachute-specific guard with no direct equivalent on other platforms.)

Sizing

  • Fractional Kelly — sizes each position as a fraction of the theoretically-optimal Kelly bet, under position and portfolio caps. A smaller Kelly fraction is more conservative. (Shown on the canvas as Kelly Sizing.) Sizing sits under Max Position Size and Max Portfolio Exposure caps.

Exit rules

How a position is closed or protected once it is open.

  • Exit Bracket — groups Take Profit and Stop Loss (with a Re-entry Cooldown) as a single set of buyer dials on the Exit stage, instead of loose unrelated knobs. They specify the take-profit and stop-loss levels for a position as a buyer-tunable dial. Set to zero, they are off (the default).
  • Tail Hedge — after a confident entry, opens a small, cheap position on the opposite side as insurance against a large adverse move. Because Kalshi does not net positions, this hedge is a real, separately-settling contract with its own cost — see Concepts.
  • Hedge Ladder — a laddered set of opposite-side hedge rungs that cap the paired cost of a position. Note it is a hedge, not a "buy more of the same side" averaging ladder — the intent is the opposite. It is the distinguishing exit component of the price-time arbitrage strategy.

Who owns which knobs

Not every part of a strategy is editable by everyone. Parachute separates components into three ownership tiers:

  • Platform substrate — non-composable safety controls that Parachute owns and that always apply, independent of the strategy. The Daily-Loss Breaker is the clearest example: it is an out-of-loop control set in the cockpit, and no strategy can switch it off. See Money safety.
  • Builder components — the Model and the Filters that define what a strategy is. When a strategy is published, these are locked as a unit (all-or-nothing in the current version) so that what a buyer runs is the algorithm the builder designed.
  • Buyer dials — the knobs a consumer is meant to adjust for their own account without changing the strategy's identity: the trading schedule, position sizing, and the Exit Bracket (take-profit / stop-loss). These are yours to tune to your own risk appetite.

This split is why some changes go through the evidence gate (they alter builder components and therefore the strategy's structure) while adjusting a buyer dial does not.