Learning path

This is the Parachute curriculum: a competency-based path from "what is an event contract" to designing and operating your own strategies. It is organized by what you can do, not by chapters — each level lists the competencies to check yourself against and the platform surface where you practice them. Work the levels in order; each builds on the previous one.

Two framing rules for the whole path:

Terms keep their industry meanings. Universal quantitative terms (edge, calibration, drawdown, slippage, …) mean here what they mean everywhere — the knowledge transfers to any other market or research environment. Venue terms (strike, settlement, YES/NO) are specific to event contracts. Parachute component names (Overpriced-Edge Guard, Tail Hedge, …) are our implementations of universal ideas, and their docs say which idea each one implements.
Nothing here is a promise of profit. Every level teaches evaluation and risk before construction. A strategy that passes every check can still lose; the point of the path is knowing exactly what your evidence does and does not show.

Level 0 — Instrument literacy

You can explain, in your own words:

What an event contract is, and what the YES and NO sides represent.
How the binary payout works: a contract settles at either 0 or 100 cents, nothing in between.
How prices are quoted, and why a contract price resembles an implied probability.
Why the displayed price is not automatically a well-calibrated probability.
When trading closes, how settlement is determined, and why "market close" and "final settlement" are different moments.
How the spread and fees change what a trade is actually worth.

Practice on Parachute: Concepts: edge & settlement covers all of this; the Positions tab shows live YES/NO quotes and the History tab shows settled outcomes to make it concrete.

Level 1 — Model & probability literacy

You can distinguish:

A forecast from a fair value, and a fair value from the executable price.
A probability from confidence in that probability.
Edge from profit — a modeled edge is an input, not a result.
Discrimination (can the model rank better opportunities above worse ones?) from calibration (when the model says 70%, do comparable contracts settle that way about 70% of the time?) from tradability (does the edge survive spread, fees, slippage, and fill uncertainty?). A model can pass any one of these and fail the others.
A large apparent edge from a trustworthy one — exceptionally large modeled edges more often indicate stale data or a broken input than a gift. This is the principle behind the Overpriced-Edge Guard.

Practice on Parachute: Tuning: Model & signal explains the fair-value model and edge; the Analysis tab shows how modeled edge related to outcomes historically.

Level 2 — Backtest literacy

You can identify, when reading any historical result:

Look-ahead bias and data leakage.
In-sample tuning and multiple testing — why the tenth variation you tried is weaker evidence than the first.
Selection bias and small-sample results.
Unrealistic fill, fee, latency, or settlement assumptions.
The difference between a replay, a backtest, paper/scan-only observation, shadow execution, and a live record — and why they are ascending levels of evidence, not interchangeable labels.

Practice on Parachute: Backtest literacy is this level's core text; run a backtest and read its qualifiers before its numbers, then compare with a free What-if replay.

Level 3 — Strategy construction

You can represent a strategy as a composition:

Model → Entry rules (signal, filters, sizing) → Exit rules (brackets, hedges), with account-level risk controls deliberately outside the strategy.
You can name the archetype a strategy belongs to (carry, arbitrage, trend, mean reversion, longshot, …) and its expected source of return.
You can state where the strategy is expected to lose — every archetype has a known failure regime, and "I don't know when this loses" means the strategy is not yet understood.
You can explain which parts are structural (changing them changes the strategy) and which are buyer dials (schedule, sizing, take-profit / stop-loss).

Practice on Parachute: the component vocabulary and the Develop tab's canvas — duplicate a template, read its composition stage by stage, and explain each block's job before changing anything.

Level 4 — Risk & sizing

You can reason about:

Position limits, portfolio exposure, and concentration by underlying and by expiration — contracts on the same BTC move are one bet wearing several tickers, not diversification.
Fractional Kelly sizing, and why estimation error in the probability makes full Kelly dangerous.
Drawdown, tail loss, and risk of ruin.
The daily-loss limit and kill switch as out-of-loop trust controls — deliberately simple, outside the strategy, and not to be confused with in-strategy take-profit / stop-loss exits.
Capacity: why an edge has a size beyond which it degrades.

Practice on Parachute: Tuning: Sizing, the Money Safety card in Bot Control, and the Positions tab's exposure view.

Level 5 — Execution

You can explain:

Bid, ask, spread, and depth — and why the displayed spread is not the whole execution cost.
Marketable versus resting orders, partial fills, queue position, and fill probability.
Slippage, latency, and stale quotes — especially when the reference market moves faster than the event-contract book.
Adverse selection: why the fills you get easily are disproportionately the ones you shouldn't want.
Why simulated edge may not be captured live, and why closing early is a second execution event with its own costs.

Practice on Parachute: the Trading Operations tab (order outcomes, non-executed trades) and the Analysis tab's Execution Quality section — compare intended prices with filled prices on your own account.

Level 6 — Deployment & operations

You can:

Read the bot's status and interpret order states, skips, and failures.
Reconcile what the dashboard shows against what the exchange reports, and say which is authoritative for what.
Distinguish a model problem from an execution problem from a venue or infrastructure problem.
Use scan-only observation — and shadow execution, as that capability ships — before committing capital to a change.
Explain the safe-state behavior: when data goes stale or reconciliation fails, the bot's job is to stop adding exposure, not to improvise.
Roll back: identify the previously deployed configuration and restore it.
Say which controls cannot be overridden from inside a strategy, and why that boundary exists.

Practice on Parachute: the dashboard guide, Bot Control (the cockpit), the trading calendar, and the config version timeline in Trading Operations.

Level 7 — Advanced research

Open-ended territory, entered only after Levels 0–6 are solid:

Volatility estimation and jump models (the family behind our fair-value model).
Probability-calibration methods and calibration drift over time.
Regime detection and regime-conditional behavior.
Walk-forward methodology and live experiment design.
Multiple-testing discipline for parameter sweeps.
Portfolio construction across strategies, capacity analysis, and execution optimization.

Practice on Parachute: the strategy archetypes survey is the map of this territory; the backtest engine plus the Develop library is the lab.

Tracks

The same levels serve different goals; go as deep as your role needs:

Running a strategy (most users): Levels 0–2 thoroughly, then 4 and 6. You should be able to interpret what you're running, its evidence, and its risk controls — not necessarily design it.
Building strategies: everything through Level 6, with Level 3 as the center of gravity. Level 7 as your ideas demand it.
Evaluating someone else's strategy (the marketplace habit): Levels 1, 2, and 4 are the defense kit — calibration, evidence quality, and risk shape matter more than headline returns.

Work through the levels at your own pace; the fastest honest path is to pick one real strategy in your Develop library and carry it through every level's questions.