Disclosures & Model Limits

Not investment advice

Qwidgets is an analysis tool. Nothing in Qwidgets or in these docs is investment, tax, or legal advice, or a recommendation to buy or sell anything. Qwidgets doesn’t manage money or make decisions for you, and it never holds your account or your funds. Your brokerage and Kalshi hold those; Qwidgets reads from them and sends the orders you place.

Every figure Qwidgets computes is a model output. Models simplify, and the sections below say how. Check anything that matters against your brokerage before you act on it.

Data

  • Delayed data. Without a connected brokerage, stock and option data comes from a delayed feed that is at least 20 minutes behind the market and updated intermittently. Don’t treat it as a live price.
  • Brokerage data. With a connected brokerage, quotes refresh about every 15 seconds, account data about every 30 seconds, and orders about every 10 seconds. Balances, positions, and orders are what your brokerage reports.
  • Market hours. Stock quotes and account data refresh during regular trading hours on weekdays, and pause outside them. Early closes aren’t accounted for, so data may keep refreshing after an early close.
  • Prediction markets. Market data refreshes every few seconds to every few minutes depending on the view, and streams live where the exchange supports it. See the Coverage Matrix.
  • Screeners refresh about every five minutes and always use the delayed feed.

Option pricing

  • Model. Options are priced as American options on a binomial tree, and their Greeks come from the same tree. Qwidgets also estimates the chance an option is exercised early.
  • Dividends. Dividends are projected from the company’s declared ex-dividend date and its usual cadence, assuming evenly spaced payments. When the cadence is unknown or irregular, future dividends aren’t projected.
  • Volatility. Implied volatility comes from the option chain. Where the market doesn’t imply a volatility for a contract, Qwidgets models its Greeks and says so. When no volatility can be read at all, Qwidgets says so rather than guess.
  • Interest rates come from a published risk-free rate curve.

Probabilities

  • Distribution. Probabilities for stocks and options assume prices follow a lognormal distribution with a single volatility. Real markets fall sharply more often than that, so these probabilities understate the chance of large downward moves.
  • Drift. Prices are assumed to drift at the risk-free rate, not at any forecast of returns.
  • Win in range is the chance a trade is profitable given the price lands inside the forecast range you set. It isn’t the chance of profit overall. Probability is the chance of landing in the range at all.
  • Implied distribution. The option chain can compare the distribution implied by option prices with the lognormal one. The two often differ, and that difference is information, not an error.

Book projection

Book management values an underlying’s positions now and at a target date.

  • Path dependence. The book walks through every expiration before the target date, so what happens to an expiring leg affects the value at the target. You choose how expiring contracts are handled: cash-settled, always exercised, or exercised only when covered.
  • Price path. Price sweep moves the price smoothly across a range of outcomes around the forward price. Monte Carlo samples random paths. Both are models of how the price might move, not forecasts of how it will.
  • Exit volatility. Current market keeps each contract’s own implied volatility, including skew. A custom value scales every contract’s volatility together.
  • Dividends paid before the target date lower the modeled share price and are counted as cash.

Price paths for multi-expiration books explains the walk in detail.

Stress test

Stress testing reprices every position in an account at prices and volatility you set for each underlying. Its presets (up or down two standard deviations, up or down 10%, double, halve) are starting points to edit.

  • It’s a scenario, not a forecast. It shows what the account would be worth if those prices happened, not how likely they are.
  • Interim expirations. Contracts that expire before the target date settle at the price modeled for their own expiration date, on the way to your target price, not at the target price.
  • Cash is held flat: it earns no interest and pays none.
  • Dividends aren’t modeled at the account level. The book page for a single underlying does model them.

Margin estimates

The stress tester estimates the account’s requirement two ways:

  • Margin (Reg T) uses exchange-minimum maintenance requirements, times a house multiplier you set.
  • Cash-secured assumes every obligation is fully covered by cash.

These are estimates from Qwidgets’ own prices. Your brokerage applies its own house rules above the minimum, and portfolio margin accounts use a different method entirely. Your brokerage’s figure is the one that counts; the stress tester shows it beside the estimate. The maximum loss on a trade ticket isn’t a margin requirement either; a brokerage will typically hold more.

Prediction markets

  • Implied probabilities are read from each market’s midpoint price. Within an event whose outcomes are mutually exclusive, they’re scaled to add up to 100%.
  • The Prediction Distribution Builder sizes positions from probabilities you assign, using a method you choose, such as Kelly. Its output follows from your probabilities, so it’s only as good as they are.

AI compute

Kalshi lists binary markets on GPU rental prices. Those binaries are real, tradable markets. Everything AI compute builds from them is derived: futures, options, the perpetual, the forward curve, and volatility. They’re computed from the binaries’ prices, aren’t listed anywhere, and can’t be traded. Derived figures are called an implied cost or a model value, never a price or a quote.

  • Distribution. Each ladder of binaries is read as a probability distribution from its midpoint prices, with gaps and inconsistencies repaired inside the bid and ask.
  • No volatility input. No implied volatility is used to price anything. Volatility is shown only as an output of the derived prices.
  • Tails. Above the top strike, value comes from fitted tail models and is shown as a range, with the tail’s share of the value disclosed.
  • The perpetual is a blend of every quoted month, weighted toward the nearest, and ends at the longest quoted month. Perpetual figures published elsewhere use different methods and aren’t expected to match.
  • Liquidity. Ladders are graded on spread, two-sided quoting, and staleness, and thin ladders are marked.
  • Dates. Any compute figure quoted outside Qwidgets should name the date it was captured. The ladders change daily.