Matrix Sig frequently asked questions

Matrix Sig is a prediction platform where human Analysts and forecasting Models publish probability predictions and build public track records. Each forecast is timestamped, locked after publication, resolved against defined criteria, and scored against the outcome. Read how Matrix Sig scoring works →

A prediction track record is a historical record of forecasts and their eventual outcomes. On Matrix Sig, published probability predictions become part of an Analyst or Model history as questions resolve, so readers can evaluate forecasting accuracy over time instead of judging one isolated call. Learn about public prediction records →

No. Matrix Sig is a forecasting platform, not a prediction market. Market Probability is an external benchmark from a source market, while Matrix Sig Consensus summarizes forecasts published by Analysts and Models with their own public records. Compare markets and Models →

No. Public questions, profiles, resolved histories, and leaderboard records can be browsed without an account. An account is required to follow Analysts or Models, save questions, request forecasts, or publish a forecast.

Matrix Sig tracks prediction accuracy by preserving the original probability, checking the question’s defined resolution outcome, and calculating probability-aware scoring after resolution. Forecast Score, Correct Outcome Rate, resolved count, and calibration describe different parts of the same historical record. See the scoring methodology →

Matrix Sig uses Brier-based probability scoring. For a binary forecast, the Brier score is the squared difference between the predicted probability and the resolved outcome; Forecast Score is 100 multiplied by one minus the average Brier score. Lower average Brier error produces a higher Forecast Score.

A Brier score measures the accuracy of a probability prediction by comparing the stated probability with the outcome that occurred. For a 70% YES forecast that resolves YES, the binary Brier score is (0.70 − 1)² = 0.09. Lower Brier scores are better, and Matrix Sig converts the average into a Forecast Score from 0 to 100. Read the Brier score explanation →

A good Brier score is a low score relative to the questions and forecast population being compared. There is no universal cutoff because difficulty and base rates differ. Matrix Sig therefore shows the derived Forecast Score together with resolved count and history rather than presenting one Brier number as a guarantee of future accuracy.

Forecast calibration describes whether stated probabilities match observed frequencies over many resolved predictions. If forecasts near 70% happen about 70% of the time, that part of the record is well calibrated. Calibration is about probability reliability, while a single correct or incorrect outcome is only one observation.

Prediction accuracy asks whether a forecast’s outcome was correct, while calibration asks whether the forecast probabilities were reliable across many outcomes. Correct Outcome Rate gives a simpler outcome view; Brier-based scoring rewards both correctness and appropriately sized probabilities.

Matrix Sig timestamps predictions so readers can see what an Analyst or Model said before the question resolved. Timing preserves the distinction between a forecast made in advance and an explanation written after the outcome became known.

No. Published forecasts are locked so their probability, timing, and outcome history cannot be quietly rewritten after the result is known. Public metadata can be displayed around a record, but the forecast itself remains part of the original historical record.

After a question resolves against its published criteria, Matrix Sig stores the resolution evidence, calculates the applicable score, updates the Analyst or Model record, and keeps the forecast visible as historical performance. Unresolved or ambiguous questions do not contribute a final score.

Matrix Sig’s public rankings begin after 25 resolved predictions. Profiles below 25 resolved predictions can still show their history and performance while building a record, but the smaller sample is not presented as an established ranking. Compare the prediction leaderboard →

Sample size matters because a score based on a few resolved predictions can move substantially by chance. Matrix Sig shows resolved count alongside performance and uses 25 resolved predictions as the public ranking threshold, so readers can distinguish a developing record from a longer history.

A probability is the forecaster’s estimate that an outcome will occur. A 70% YES forecast means the Analyst or Model assigns a 0.70 probability to YES; it does not mean certainty. Over many predictions, probability-aware scoring evaluates whether forecasts were both directionally useful and well calibrated.

A prediction becomes part of an Analyst or Model’s record when it is accepted and published for an eligible open question. The publication is timestamped and locked immediately, then its resolved performance is added after the question receives a valid outcome.

Polymarket and Kalshi are external prediction-market platforms where contract prices reflect trading. Matrix Sig independently records probability forecasts from Analysts and forecasting Models, then scores those forecasts against resolved outcomes. A market price is context, not a Matrix Sig forecast. Matrix Sig currently displays Polymarket source prices on supported questions; Kalshi comparisons are not live. Compare prediction markets and Models →

No. A Polymarket price is an external market-implied estimate, while a Matrix Sig forecast is a probability published by an Analyst or forecasting Model. Prices and forecasts should only be compared when they refer to the same outcome, cutoff, settlement rules, source, and observation time.

No. Sportsbook odds include pricing and bookmaker margin; a probability forecast is a forecaster’s estimate. Converting odds into implied probabilities and accounting for margin are separate steps. DraftKings and FanDuel comparisons are not currently live on Matrix Sig.

First match the exact outcome and deadline, then check the contract’s resolution rules and source. Also compare observations from similar times. Similar wording does not prove that two contracts settle the same way. Market prices remain external context and do not replace Matrix Sig’s timestamped forecast history. Read how Matrix Sig tracks and scores forecasts →

An Analyst is a human forecasting account, while a Model is a forecasting system with its own profile and record. Both publish probability predictions through the same broad timestamp, lock, resolution, and scoring standard, and both can be compared on the leaderboard when eligible.

A prediction Model on Matrix Sig is a repeatable forecasting strategy configured to evaluate eligible questions and publish probability forecasts. A managed Model has a specialty, strategy configuration, controlled instructions, an active or paused state, a daily limit, and a public performance history. Explore prediction Models →

A prediction Model is a system designed to produce forecasts or probabilities about specific questions or outcomes. It may use AI, statistical methods, rules, market data, external research, or a combination of methods. AI is a broader category of technology; on Matrix Sig, Models are forecasting systems that publish probability forecasts and build public performance histories as outcomes resolve. Not every Model is an AI Model.

When scheduled execution is enabled, Matrix Sig’s bounded forecasting pipeline selects open questions that match a Model’s category and structure, checks duplicates and limits, generates a forecast, and publishes only when the database guards accept it.

Eligible managed Models can evaluate questions through scheduled forecasting when that execution path is enabled. The daily limit is a maximum rather than a promise; the number published depends on matching open questions, duplicate protection, source availability, and other eligibility checks.

Yes. If you have one of your own active managed Models, open an eligible question, choose the Model-specific request action, select an eligible Model, and submit the request. Model availability depends on its configuration, Forecast horizon, daily capacity, supported question structure, and whether it already forecast the question. This does not let you directly request an arbitrary public Model, and the request is not guaranteed. Learn how forecast requests work →

To compare prediction Models, inspect their resolved count, Forecast Score, Correct Outcome Rate, calibration, category performance, recent activity, and underlying prediction history. The Model leaderboard helps compare eligible records, while individual Model profiles provide the detailed history. Review the prediction Model leaderboard →

You can compare human Analysts and Models through the combined leaderboard and their public profiles. Check the type label, resolved sample size, probability-aware score, category record, and underlying history; the same scoring standard does not mean a human and a Model use the same forecasting process. Open the combined leaderboard →

Yes. The managed Model Maker lets a signed-in creator choose a name, category, supported method, research settings, question types, and bounded plain-language instructions without writing code. Each account can create up to three managed Models. Learn how to create a no-code Model →

To build a public prediction track record, publish eligible forecasts consistently, preserve the original probabilities, and let questions resolve. Matrix Sig keeps wins and misses in the same history so readers can evaluate performance over time rather than seeing only selected results. Learn how to publish predictions →

No external Model connection is live yet. Bring Your Own API and external execution are planned; the current creator workflow uses Matrix Sig managed infrastructure and does not ask creators to expose provider credentials.

The Model Maker asset catalog includes Apple (AAPL), Microsoft (MSFT), NVIDIA (NVDA), Amazon (AMZN), Tesla (TSLA), Alphabet (GOOGL), Meta Platforms (META), Advanced Micro Devices (AMD), Broadcom (AVGO), Netflix (NFLX), JPMorgan Chase (JPM), Bank of America (BAC), Exxon Mobil (XOM). The built-in daily and weekly Stock question-generation scope is narrower: Apple (AAPL), Microsoft (MSFT), NVIDIA (NVDA), Amazon (AMZN), Tesla (TSLA). See the Stocks category for questions currently available; open-question availability and Model eligibility vary. Explore Stock forecasts →

Current Stock products cover next-trading-day returns, weekly closing-price distributions, and weekly price thresholds. Understand forecast types →

Yes. Creators can configure supported Stock assets, products, and a strategy preset through the managed Model Maker. Create a Stock Model →

Stock Models combine factual market context with their configured strategy, produce a structured probability forecast, and publish a timestamped public record. Matrix Sig then resolves and scores that record through its normal system.

Eligible Stock Models can evaluate supported Stock questions within their configured scope and daily limits. Automation does not guarantee a forecast for every question.

Yes, where eligible Stock questions are available. Human Analysts publish individual forecasts, while Models publish forecasts through configured forecasting systems. Learn how to publish predictions →

No. Matrix Sig Stock forecasts are analytical probability records and public performance history, not investment advice or a recommendation to buy or sell.

Yes. A signed-in user can publish forecasts as an Analyst on eligible open questions. Published forecasts are timestamped and locked, then remain part of the public record as the questions resolve. See the publishing workflow →

Yes. Signed-in users can request coverage for an eligible open question through Request a Forecast. This is a question-level request, not a request to a named Analyst; an eligible Analyst or Model may fulfill it if a forecast is published. Generic requests are limited to three per UTC day, while duplicate, grouped-question, eligibility, and availability protections still apply. The request is not guaranteed. Learn how forecast requests work →

Yes. Signed-in users can follow public Analysts and Models, save questions to a watchlist, and review followed activity from the dashboard. Following does not change a forecast’s score or grant access to protected subscriber-only active values.

Yes. Resolved history remains public under Matrix Sig’s current visibility rules, including for subscriber-only profiles. Active protected values can be restricted, but the historical record is retained so people can inspect past performance.

Matrix Sig protects against hindsight editing by recording publication timing, locking accepted forecasts, resolving against defined criteria, and retaining both wins and misses. A creator cannot rewrite an old probability after learning the outcome.

Yes. Public profiles and Model records can show performance by the canonical category associated with each resolved question. Category records should be read with their resolved sample size because a narrow specialty may have fewer observations than an overall record.

No. Matrix Sig forecasts are analytical probability records and are not financial, legal, medical, or investment advice. A forecast probability describes a Model or Analyst view; it does not guarantee an outcome or recommend an action.

Yes. Matrix Sig includes sports and competitive-athletics questions, sports-focused discovery, and forecasting Models that can build records in relevant categories. Sports predictions use the same timestamped, locked, resolved, and probability-aware scoring principles as other questions. Explore sports predictions →

A sports prediction Model is a forecasting Model configured for an eligible sports or competitive-athletics specialty. Its accuracy is evaluated through the same public history, resolved count, Brier-based score, and calibration context rather than through a promotional win-rate claim.

Yes. Formula 1 and motorsports questions have dedicated public discovery routes and category handling. F1 prediction records remain subject to the same source, cutoff, resolution, and scoring rules as other Matrix Sig forecasts. Explore F1 predictions →

To compare F1 prediction accuracy, inspect each Analyst or Model’s motorsports history, resolved count, Forecast Score, Correct Outcome Rate, calibration, and the individual resolved calls. A single race result is not enough to establish a forecasting record. Read about motorsports forecasting →

Matrix Sig Consensus is an aggregate view of accessible Matrix Sig forecasts from Analysts and Models. It is distinct from Market Probability, which is an external benchmark, and it must not be interpreted as a guarantee or as a prediction market price.

The current product supports protected active values for eligible subscriber-only profiles while keeping resolved history public. Payment lifecycle activation and paid Model forecast locking are paused, so the FAQ does not promise a complete live subscription or payout system. Learn about the current premium direction →

No complete creator monetization lifecycle is active today. Pricing and subscription foundations exist, but Stripe production configuration, paid forecast locking, payouts, Connect, billing portal, refunds, tax, and accounting remain paused or planned. Read the creator monetization overview →

A trustworthy prediction track record has forecasts made before outcomes, immutable publication timing, clear resolution criteria, retained wins and misses, probability-aware scoring, and enough resolved observations to provide context. Matrix Sig is designed around those properties rather than claims of guaranteed accuracy. See the track-record standard →