Overall accuracy
For the selected Set and displayed period, accuracy is the number of correct qualifying completed outcomes divided by all qualifying completed outcomes. Percentages from different Sets are not averaged together.
How prediction results are measured
The sections below form this notice. A more specific transaction disclosure accepted at checkout controls for that transaction.
CORRECT PREDICTIONSCorrect eligible outcomes
COMPLETED PREDICTIONSAll eligible completed predictions
MODEL ACCURACYOutcome percentage
Model outcome accuracy is not customer profit, account performance, or a guaranteed future result.
Per-set record
The selected Set’s record shows correct and completed outcome counts, the measured dates, coverage and latest update. Its public labels describe the evidence shown; internal strategy identifiers are not customer-facing results.
Window, coverage, and public delay
Set {01}'s public model-outcome record uses a rolling 90-day window and releases newly settled outcomes only after the 15-minute public reporting delay. Read the stated dates, last update, and coverage information alongside the percentage. Missing source coverage is not evidence that no signals occurred. The public reporting delay is separate from member delivery timing.
Eligibility
Accuracy includes only qualifying settled model decisions under the stated selection and settlement rules. Abstentions, synthetic examples, voids, unresolved outcomes, and duplicate observations do not count as wins or losses. Days without records are identified and left out of the average predictions per day. An absence of eligible decisions is not a losing prediction.
Historical and live outcomes
The record combines retrospective evaluation of collected market data with newer decisions recorded before their contracts closed. A disclosed historical revision can add dated results for an expanded Set at the original contract closing dates. The combined figure is not an independent forward test. Coverage varies; missing periods are not estimated. Historical calculations do not prove member signals were delivered before their outcomes. That requires separate timestamped publication evidence. Overlapping outcomes are counted once, and model results do not represent customer trading returns.
Simulated results and examples
Synthetic examples, simulations, and backtests must be identified as such. They do not represent executed customer trades and may not capture available liquidity, execution delays, market impact, fees, or other real trading conditions. Neither a simulation nor evaluation after the event proves a signal was published before its outcome.
Corrections
Corrections are versioned and visible. Historical records are not silently rewritten to remove losing outcomes or to apply later information to earlier signals.
Required legend
Historical accuracy does not predict or guarantee future accuracy. Accuracy is not profit and loss and does not include entry price, fills, fees, slippage, sizing, drawdown, taxes, or customer behavior.