Market data methodology
Understand AutoSignal chart calculations, data coverage, model assumptions, source timestamps and research limitations.
Reading a data point
Start with the instrument, exchange or source scope, timeframe, units and observation timestamp. A spot price, a perpetual last-trade price and a mark price describe different markets or price bases. A displayed period is the requested observation window, not a guarantee that every source has that much history.
AutoSignal combines market observations and calculated research. Proprietary signal rules, model weights, tuning, source-selection logic and internal infrastructure are not published. The descriptions below explain what a reading means and its limits.
Prices, candles and timeframes
Quotes and candles come from the instrument and price basis associated with the displayed source. Candles summarize observed open, high, low, close and available volume over their interval. Changing a chart interval changes aggregation; changing a data window changes the history under review.
Sources update independently and may return bounded history, delayed observations or corrections. When an input is missing, its value must not be treated as zero. A recently retrieved response may still contain older source observations.
Open interest and funding
Open interest describes outstanding derivatives exposure for the reporting contracts. We normalize supported native readings into the displayed units and aggregate only the included contracts. Options, dated futures, perpetuals and different exchanges are not interchangeable scopes.
Funding is the rate reported for the relevant contract and settlement interval. A next-funding estimate is not necessarily the rate ultimately paid. Compare the exchange, price basis, contract coverage, interval and timestamp before comparing totals or rates with another platform.
Recorded liquidations
Recorded liquidations represent forced-order events observed in the available source coverage. A long liquidation closes a long position; a short liquidation closes a short position. An order list can be shorter than the event set used for a period summary.
Coverage and reporting can differ by venue and time. An empty returned window is not proof that no liquidation happened elsewhere. Exchange-reported events, modeled future liquidation levels and changes in open interest are separate measurements.
Liquidation heatmaps and levels
The heatmap presents modeled concentrations of potential liquidation exposure over price and time. These are estimates because full position entries, leverage, margin arrangements and later position changes are not observable. A bright band indicates stronger modeled concentration on the displayed scale; it does not guarantee a price target or liquidation event.
Cursor readings refer to the selected price/time cell. A USD amount is shown only when that source supplies an amount for the applicable cell or explicitly identified current price band. Empty cells do not receive invented dollar exposure. Current pool summaries and dated historical readings must be distinguished. Modeled exposure is not the same as a recorded forced liquidation.
Signals, tracking and returns
AutoSignal publishes automated research setups with direction, instrument, timeframe and any supplied entry, stop and targets. Publication, entry eligibility, active tracking, target status and a closed paper result are different states. A notification communicates a setup; it does not execute it.
Paper results use the source’s recorded setup prices and tracking status. They do not verify execution on a user’s exchange. A leverage-adjusted percentage changes a displayed return calculation, not the original entry, stop or target. Fees, funding, slippage, liquidation and taxes can make actual results materially different. Selected examples and past outcomes are not forecasts of future performance.
Indicators and quantitative research
Indicators summarize price, volume or other available inputs using their stated settings. Oscillators, trend overlays, breadth measures, market-structure tools and scenario models answer different questions and can disagree without describing the same quantity.
Warm-up history, missing volume and changes to the input interval can affect a calculation. Model probabilities, forecast bands and confidence-like scores are estimates, not verified success rates. Proprietary thresholds, combinations and optimization rules remain confidential.
Order flow, books and profiles
Executed order flow describes recorded buys and sells. Book liquidity describes resting orders, which can move or cancel without executing. Footprints group observed executions by price and interval. Volume profiles describe the distribution over the stated data window; candle-derived profiles may estimate distribution rather than use every trade.
Trade size, turnover, delta, book size and open interest have different units and meanings. These views do not establish a trader’s identity, intent or future order. Limited samples and capped histories affect interpretation.
Exchange wallet flows and ETFs
Exchange wallet flows track supported on-chain token transfers involving attributed exchange wallets. The attribution boundary is limited: an unlabelled address may belong to an exchange, and internal sweeps can cross the boundary. Transfers are not evidence of an executed buy or sell. Native token units are not silently treated as exact USD value.
ETF views present the available reported fund readings and trading dates. Net flow, net assets, trading volume and cumulative flow are different fields; missing reported values are not filled with fabricated readings. A reporting date or historical window is not a live intraday promise.
Options, positioning and public wallets
Options exposure depends on the included strikes, expiries, pricing inputs and available positions. Exposure and settlement scenarios are calculated sensitivities or estimates; they are not a direct measurement of every dealer’s inventory.
Long/short account shares count accounts in their defined population and do not measure position size. Public-wallet positions and P&L reflect the source’s scope and timestamp. They do not imply ownership of an identified person or reproduce another platform’s entire market universe.
Referral accounting
Referral amounts come from confirmed attributed payments, not market models or signal performance. The current rate is 15%, including renewals and repeat payments. Crypto and Stripe holds are measured from the relevant payment confirmation, for 24 hours and 21 days respectively.
Refunds, disputes and reversals adjust commission. Available commission excludes unexpired holds and amounts already reserved or paid. Manual payout records identify the receiving wallet and transfer reference; a pending request is not paid. Leaderboard figures describe participating partners’ recorded net commission, not a projection of future earnings.
Corrections and interpretation
Unavailable, delayed and corrected data can affect a chart or derived reading. We preserve source context and distinguish observed events, estimates and paper tracking rather than presenting them as the same thing.
Use the visible units, settings, time and coverage when interpreting a result. If a feature cannot provide a valid reading, a blank or unavailable state is more accurate than an invented number. AutoSignal is a research workspace; it does not know your execution, financial situation or ability to absorb a loss.