BEGINNER TUTORIAL · ABOUT 20 MINUTES
Understand Polymarket. Practice forecasting for fun.
Learn to read a question, interpret a price, check evidence, and keep a paper forecast journal. You do not need a trading account.
1. Read the question before the percentage
Polymarket is a prediction-market platform where participants trade outcome shares. In a standard binary market, the winning outcome pays one unit at resolution and the losing outcome pays zero. This tutorial studies those mechanics on paper; it does not ask you to trade.
The headline is a summary. The full rules define what counts, which source decides it, and when it must happen. Copy the exact threshold, deadline, time zone, source, and exceptions. An announcement, an implementation, and an official confirmation are different events.
Practice question: Will the fictional Town A weather station report at least 1 mm of rain between midnight and midnight tomorrow, local time? Identify the station report as the deciding source. A photograph of wet pavement is not enough.
2. Understand what a price can tell you
A 60-cent price in a one-unit payout market is often read as roughly 60% implied probability. It reflects participants’ willingness to buy and sell, shaped by information, incentives, liquidity, and constraints. It is not an objective measurement of the future.
The bid is an offer to buy; the ask is an offer to sell. The spread is the gap between them. A displayed midpoint is not necessarily a price available for a transaction. Polymarket’s help center says the display uses the bid–ask midpoint, or the last traded price when the spread exceeds ten cents.
Paper example: a 48-cent bid and a 52-cent ask give a 50-cent midpoint. Record all three numbers and their timestamp. Do not treat a chart point as a guaranteed execution price.
3. Separate market activity from evidence
An order book lists buying and selling interest. Orders execute when compatible interest is matched. A limit order sets a price boundary but may never fill; an immediately executable order may consume several price levels. Slippage is the difference between an expected price and the actual execution price.
Liquidity concerns how much can be exchanged at available prices. Volume describes completed activity over a period. High volume does not guarantee a narrow spread now, and a tiny transaction can move a thin market. Fees and other costs depend on the platform and market; check current official information instead of assuming zero.
For this tutorial, observe only. A move from 50% to 60% may follow new information, but the move itself does not prove the information is accurate. Ask what changed in the underlying evidence.
4. Distinguish an outcome from its resolution
An event happening and a market being formally resolved are different steps. Polymarket’s process applies the published rules, with proposal and dispute mechanisms. A news headline or the market end time alone is not the final settlement record.
Check clarifications and exceptions on the original market. Some questions require a particular publication, measurement window, or authority. A related data provider may be useful context while still not being the named resolution source.
Paper example: the fictional station reports 0.8 mm, while a nearby station reports 1.2 mm. Under our named-station rule, the answer is NO. If the required report is missing, record unresolved and consult the exception rules rather than guessing.
5. Make a forecast you can explain
Write an initial estimate before copying the crowd. Start with a relevant base rate: how often did comparable events happen? Then record what makes this case different. Express uncertainty honestly; ‘I do not have enough evidence’ is a useful conclusion.
Keep a source ledger with the link, publication time, claim, relevance to the rule, and any contradiction. Prefer the named primary source. Five stories repeating one press release do not constitute five independent observations.
Use AI to organize questions and inspect supplied material, then check its claims. Jev’s confidence in an article classification is not an event probability. A price-history forecast is not a prediction of how the event resolves. A precise percentage can still rest on weak evidence.
6. Run a paper-only forecasting session
Choose one low-stakes question for curiosity. Write its rules, your initial probability, the current time, and one reason your estimate might be wrong. No wallet, deposit, or purchase is part of this exercise.
Open the Market Browser to inspect the question, Research to organize sources, and the relevant how-to guide before trying a calculator. When a form says dollars, bankroll, or bet size, use fictional points solely to explore the arithmetic.
Set a review time. Update only when you can name new evidence or correct a mistake, and preserve the old estimate in your notebook. A calm ten-minute observation session is sufficient; there is no need to watch every price change.
7. Learn from outcomes without rewriting history
After formal resolution, record the outcome and source next to the original forecast. Keep unresolved questions unresolved. Retain misses as well as successes, and avoid quietly changing deadlines or deleting inconvenient entries.
Calibration asks whether events assigned similar probabilities happen at similar frequencies. Across many comparable 70% forecasts, about 70% should resolve YES. A single NO outcome does not disprove a 70% forecast. Small samples and selectively chosen questions can be misleading.
For a binary Brier score, convert the percentage to a decimal, subtract the outcome (YES = 1, NO = 0), and square the difference. A 70% forecast scores 0.09 for YES or 0.49 for NO. Average over the same set and compare with a simple baseline. PolyMath’s current calibration page includes seed data; keep your personal journal separately.
8. Keep the activity enjoyable and bounded
Measure success by what you learned: a clearer question, a corrected assumption, or a better explanation. Use fictional points, stop at your chosen time, and never turn a practice score into a reason to chase losses or commit money.
These guides are for fun and education, not instructions for speculation or gambling. Real-money event contracts can lose value, and platform eligibility and local rules still apply. An entertainment label does not change those facts.
Before ending a session, ask: Did I read the rules? Did I check the source and timestamp? Can I distinguish a price, my estimate, and an AI label? Did I record uncertainty? If yes, you have completed a useful learning session regardless of the eventual outcome.
Copy this paper-journal template
Question and exact YES condition: Resolution source / deadline / time zone: Forecast date and probability: Market quote and timestamp (optional): Supporting evidence / contradicting evidence: What would change my mind: Next review time: Revised estimate and reason (keep the original): Final outcome, resolution source, and score:
Check your understanding
Does a displayed 70% guarantee YES?
No. It is a price-based signal. Even a well-calibrated 70% forecast allows NO about 30% of the time across comparable cases.
Can a bullish company article resolve a market about an official earnings number?
No. It may provide context, but the exact required publication and threshold decide the question.
Is Jev’s 90% classification confidence a 90% YES forecast?
No. Confidence concerns classification of the supplied evidence, not the chance of the event occurring.
Do you need to put money at risk to complete this tutorial?
No. Every exercise can be done with notes, hypothetical examples, and fictional points.
Official sources and further reading
Mechanics checked September 21, 2026. Platform processes can change; consult the current rulebook for the specific market.