Independent research and education. PropBetEdge does not execute event-contract trades and nothing here is investment, betting or legal advice. Examples marked illustrative use invented numbers to teach mechanics.
01 · Prediction Markets 101
A question, a contract, a price
A prediction market lists event contracts: written agreements that pay a fixed amount if a defined event happens and nothing if it does not. On the major U.S.-facing venues a binary contract typically settles at $1 for YES and $0 for NO, so its price in cents reads naturally as a percentage. A YES contract trading at 63¢ implies roughly a 63% chance, in the market’s collective judgement, that the event happens.
Three words carry the whole idea, and each has a trap:
| Word | What it means | The trap |
|---|---|---|
| Price | The level at which a participant is willing to trade right now | The last trade or the midpoint may not be available to you for any size |
| Implied probability | The price read as a probability: 63¢ ≈ 63% | A market view shaped by liquidity, fees and who is trading — not a fact and not a model |
| Resolution | How the venue decides the outcome, using a named source and rules | The rule can define the event differently from the headline |
Because YES and NO are complements, a NO contract trades near 100¢ minus the YES price (the spread keeps the two from summing to exactly 100). Moving the slider in the reader above shows the same arithmetic. If you already know odds, this is the same idea as Implied probability from the implied-probability lesson — with the advantage that a $1 contract makes the mapping literal.
A 90¢ contract still resolves NO about one time in ten if it is well priced. One outcome cannot tell you whether a price was right; only many outcomes, compared honestly, can. That is what calibration measures.
02 · Reading Market Rules
The title is shorthand. The rules are the contract.
Most avoidable mistakes in event markets are not forecasting mistakes — they are reading mistakes. Before forming any view, decompose the rule into five parts: what is measured, where, over what window, against what threshold and according to which authority. Then look for the edge cases.
KXHIGHNY-26OCT12-T76Will the maximum temperature be >76° on Oct 12, 2026?
If the maximum temperature recorded at New York City (CLINYC) for Oct 12, 2026, is greater than 76° fahrenheit according to The Weather Company, then the market resolves to Yes.
- 1 · Measure. Daily maximum, not the temperature at any one hour.
- 2 · Location. One station code. Your neighbourhood can be warmer.
- 3 · Window. One calendar day, as the source defines it.
- 4 · Threshold. Strictly greater than 76°. Exactly 76° is NO.
- 5 · Authority. The named source settles it — not a forecast app.
Decode five real contracts
Each contract below was read from the venue’s published rules. Switch between them and notice where the headline and the rule part ways.
KXHIGHNY-26OCT12-T76Will the maximum temperature be >76° on Oct 12, 2026?
If the maximum temperature recorded at New York City (CLINYC) for Oct 12, 2026, is greater than 76° fahrenheit according to The Weather Company, then the market resolves to Yes.
| Element | What the rules say |
|---|---|
| Measure | Daily maximum temperature, °F |
| Location | One named station: New York City (CLINYC) — not your neighbourhood or your phone’s location |
| Threshold | Greater than 76° (the venue labels the outcome “77° or above”) |
| Window | The calendar day Oct 12, 2026 |
| Authority | The Weather Company, per the series’ listed settlement source |
| Edge cases | The rules warn that preliminary readings can differ from the final value through rounding and conversion, and allow expiration to be held if the first publication has a material error |
The forecast app on your phone and the settlement source can disagree. The contract settles on one station and one named source — a 76.6° reading that a different app rounds to 77° is still a question for the official value.
KXNFLGAME-26OCT19WASSF-WASWashington wins (Commanders vs 49ers, originally scheduled Oct 19, 2026)
If the game ends in a tie, the market will resolve to $0.50 for each team. If the game is postponed but begins within 48 hours from its originally scheduled start time, the market will remain open…
| Element | What the rules say |
|---|---|
| Measure | The official final result of one game |
| Window | The game “originally scheduled for Oct 19, 2026” |
| Tie | Resolves to $0.50 for each team — neither a win nor a total loss |
| Postponement | Starts within 48 hours: market stays open; not within 48 hours: resolves to a fair price |
| Authority | The governing league’s official result |
A tie is its own outcome with its own payout, and a postponement changes the timeline. A sportsbook habit (“ties push”) does not transfer automatically.
KXCPIYOY-26DEC-T5.0Will the rate of CPI inflation be above 5.0% for the year ending in December 2026?
If the Consumer Price Index (CPI) increases by more than 5.0% in the twelve months ending December 2026 (as represented by the one-decimal place value reported by the Bureau of Labor Statistics), then the market resolves to Yes.
| Element | What the rules say |
|---|---|
| Measure | Year-over-year CPI change for the twelve months ending December 2026 |
| Precision | The one-decimal value the Bureau of Labor Statistics reports — a 5.0 print is not “more than 5.0%” |
| Threshold | More than 5.0% |
| Timing | Trading closes on the release date (Jan 13, 2027 at 13:29 UTC per the market listing) |
| Edge cases | If a federal government shutdown delays the data, the latest expiration extends to the sooner of the release or six months after the shutdown ends |
Rounding is part of the contract. Your own calculation to two decimals is not the number that settles it.
KXFEDDECISION-28JAN-H26Will the Federal Reserve hike rates by >25bps at their January 2028 meeting?
This market is mutually exclusive. Therefore, if the Federal Reserve hikes by 50bps, the 50bps market will resolve to Yes and the 25bps market will resolve to No.
| Element | What the rules say |
|---|---|
| Structure | A ladder of mutually exclusive outcomes — at most one bucket resolves Yes |
| Measure | The decision announced at the meeting on Jan 26, 2028 |
| Edge cases | If the scheduled meeting is canceled and does not occur on its date, “Fed maintains rate” resolves Yes and every other bucket No |
| Authority | The Federal Reserve |
In a mutually exclusive ladder the bucket prices should be read together. “Hike” is not one question here — it is several buckets that cannot all be right.
KXTVSEASONRELEASETHELASTOFUS-27-APRWhen will “The Last of Us” Season 3 be released? — Before Apr 2027
If it has been announced that "The Last of Us" Season 3 will release before Apr 1, 2027, then the market resolves to Yes.
| Element | What the rules say |
|---|---|
| Headline question | “When will it be released?” |
| What actually decides it | Whether a release before Apr 1, 2027 has been announced — an announcement, not the premiere itself |
| Window | Before Apr 1, 2027; trading closes Apr 1, 2027 |
| Authority | A named list of news organisations in the series’ settlement sources |
The title asks about a release. The rule resolves on an announcement. Forecasting the premiere date perfectly can still mean misreading this contract.
Read from Kalshi’s public market data on . Shown to teach rule reading — not a recommendation, and not a current price.
Six kinds of mismatch
| Mismatch | Example from the contracts above |
|---|---|
| Event vs announcement | A “release date” market that resolves on whether a date has been announced |
| Location | One weather station, not the city or the reader’s neighbourhood |
| Precision | A one-decimal CPI print: 5.0 is not “more than 5.0%” |
| Strict vs inclusive thresholds | “Greater than 76°” excludes exactly 76° |
| Special outcomes | An NFL tie pays $0.50 per side; a long postponement resolves to a fair price |
| Structure | Mutually exclusive Fed buckets: at most one can be YES; a canceled meeting has its own rule |
03 · Price vs. Probability
The number on the screen is not the price you get
A market quote has two sides. The bid is the highest price someone will pay; the ask is the lowest price someone will accept. The gap between them is the spread, and the quantity available at each level is the depth. A headline probability is usually the last trade or the midpoint — neither is guaranteed to be available, and neither includes fees.
| Bid size | Price | Ask size | Depth |
|---|---|---|---|
| 66¢ | 400 | ||
| 64¢ | 150 | ||
| 63¢ | 80 | ||
| midpoint 62¢ · spread 2¢ | |||
| 120 | 61¢ | ||
| 300 | 60¢ | ||
| 500 | 58¢ | ||
Illustrative. Venues publish their own fee schedules, and some fees vary with the contract price.
Work the tool: at small size the best ask is the price, but larger quantities walk up the book, and the average price rises. Add a fee and the all-in cost per contract — the number a forecast has to beat — rises again. That all-in cost, divided by 100, is the break-even probability, the same idea taught in Break-Even Rate.
| Signal | What it can tell you | What it cannot |
|---|---|---|
| Wide spread | Participants disagree or few are quoting | Which side is right |
| Thin depth | Modest size can move the price | Whether the move reflects new information |
| Fast move | Something changed — news, data or a large participant | That the move will continue |
| Stale last trade | Where a trade once happened | What is available at this moment |
In a deep, active market a price aggregates many views and moves on information. In a thin one, a single participant can set it. The same 63¢ can be strong evidence in one market and almost none in another.
04 · Research Discipline
Build the forecast first. Then let the market argue with it.
A forecast you form after looking at the price tends to drift toward the price. The discipline is simple to state and hard to keep: write your probability down, with its time and its reasons, before you look.
- Start from a base rate. How often does this kind of event happen? Historical frequencies anchor you before the story does.
- Adjust for specific evidence. What is different this time, and how much should each fact move you? Small adjustments are usually right.
- State your uncertainty. A range (“55–70%”) is more honest than a point, and tells you how much a gap has to be before it matters.
- Record it with a timestamp. A forecast you cannot prove you held is a forecast you will quietly revise.
- Compare, then question both. When you disagree with the market, list what the market might know that you do not.
- Review outcomes in batches. Judge your process across many forecasts with Brier scores and calibration, not by the last result.
Hindsight is the enemy of learning
Once an outcome is known it feels inevitable, and every forecast that missed it looks careless. That feeling is hindsight bias. The defence is the timestamped record: it lets you ask whether the forecast was reasonable given what was knowable then, which is the only question a single result cannot answer. The same reasoning is why pregame and live probabilities are graded separately.
Contract: exact rule and source · Recorded: date and time · My probability: point and range · Base rate: source · Key evidence: three items, with the time each was available · What would change my mind: one or two facts · Market at the time: bid, ask, depth.
05 · PBE Model vs. Market
An independent forecast, compared honestly
PropBetEdge Predictions ↗ is built around one principle: forecast first, compare second. Four rules make the comparison meaningful.
The market never enters the model
Venue prices are kept out of the model’s inputs. A model that learns from the price it is judged against measures nothing but its own echo — see Leakage.
Model the contract, not the headline
Forecasts are built for the exact rule: station, threshold, window and source. A comparison is shown only when the venue contract is the same contract.
Recorded, time-aligned forecasts
Each forecast is stored with its time and the data cutoff behind it. Market observations are compared only when they existed at that time, before lock.
Model states: what a number is allowed to claim
Every model family carries a state, and it moves up only by passing a test that was set before the results were known. A RESEARCH probability is published evidence; it is not yet a validated forecaster. How states work ↗ · Current state of every model ↗
- MONITORINGMarket shown, no model
- BACKTESTINGTested on history
- SHADOWForecasts stored, never published
- RESEARCHPublished, not yet validated
- VALIDATEDPassed a gate set in advance
- OFFICIALEligible for official, locked calls
Calibration is the first test
When a well-calibrated forecaster says 70%, events like that happen about 70% of the time. Calibration is measured over many forecasts, and the size of the miss — how far a model sits from the diagonal on contracts like this one — is the honest width of its uncertainty. Calibration · Brier Score and Log Loss
That width matters for the comparison below. A model that disagrees with the market by four points, while its own error on similar contracts is five, has not found anything yet.
From a model’s own calibration record — the honest width of its uncertainty.
A disagreement smaller than the model’s own error. Indistinguishable from noise — a reason to keep researching, not a conclusion.
A model-vs-market gap is a disagreement between two estimates of an uncertain event. It does not account for information the market may have, can vanish once fees and the spread are paid, may not be available for meaningful size, and says nothing about one outcome. PropBetEdge publishes disagreements as research, alongside its track record ↗, so you can judge the model on evidence rather than on any single call.
Where to see this on PropBetEdge
- Free: the methodology ↗, the model research board ↗, the public track record ↗ and the prediction calendar ↗.
- With All Access: PBE probabilities and model-vs-market divergence on the Predictions desk ↗, official locked picks, and Compare ↗, which sets PropBetEdge calls beside exact venue contracts. About All Access ↗
06 · Risk and Responsible Participation
Know the loss before you think about the gain
A YES contract bought at 63¢ can lose all 63¢. That is the defining risk of a binary contract: the outcome is all or nothing, and a strong probability still leaves room for a total loss on any single event.
Set before looking at any price. Stays in this page; nothing is stored or sent.
Above your limit by $7.00. The limit wins.
| Risk | What it looks like | A discipline that helps |
|---|---|---|
| Total loss per contract | The full price paid is lost if the event does not happen | Size every position as if it will resolve against you |
| Concentration | Five cities’ rain markets on one weather system; several games decided by one injury report | Treat contracts with a common cause as one position |
| Capital exposure | Money committed until resolution, which can be days or months away | Count capital as tied up until the source publishes |
| Volatility | Prices can swing sharply before resolution on news or thin trading | Decide in advance what new information would change your view |
| Eligibility | Access depends on venue, product, age and jurisdiction, and changes over time | Check the venue’s current eligibility; never bypass restrictions |
| Rules risk | Being right about the world and wrong about the contract | Chapter 02, every time |
Related reading: Bankroll and Bet Sizing and Model Confidence vs Prediction Certainty. Probabilities describe uncertainty; they do not remove it.
Set a time and money limit before you start, and keep it. If you feel the need to chase losses or hide activity, stop and talk to someone. In the U.S., call or text 1-800-GAMBLER for free, confidential help. You must meet the legal age and eligibility requirements where you are.
07 · Field guide
Sports, weather, economics and culture
The same discipline applies everywhere, but each domain has its own drivers, sources and timing traps. Each card links to a real contract decoded in chapter 02.
Sports
Decode KXNFLGAME- What moves the probability
- Team strength, availability, rest, travel, weather for outdoor games and game state once play starts.
- What settles it
- The governing league’s official result settles it. Injury and lineup news move prices fastest.
- Timing trap
- Pregame and live prices answer different questions. A pregame forecast compared with an in-game price is not a fair comparison.
- Research prompts
- Team-strength and availability inputs before the price, not after
- Whether a live price reflects the game in progress
- How the specific contract treats ties, overtime and delays
- On PropBetEdge
- PropBetEdge’s sport sites publish model probabilities and track records by sport; Compare ↗ (All Access) sets PBE calls beside exact venue contracts when the IDs match.
Weather
Decode KXHIGHNY- What moves the probability
- Station-level forecasts and their spread, observed conditions as the day unfolds, and how far the threshold sits from the expected value.
- What settles it
- One named station and one named data source. Preliminary readings can differ from the final value.
- Timing trap
- Weather resolves on a calendar day as the source defines it. Intraday observations can make a contract nearly decided hours before close.
- Research prompts
- Point forecasts and their spread for that station, not for the city in general
- How close the threshold sits to the forecast high
- When the official value is published, and what “preliminary” means here
- On PropBetEdge
- PropBetEdge Predictions models daily highs and measurable rain from official station data, with venue prices kept out of the model. See the Intelligence Desk ↗ and the calendar ↗.
Economics
Decode KXCPIYOY- What moves the probability
- Prior releases, the published components, forecaster consensus and its dispersion, and the reporting agency’s exact method.
- What settles it
- The reporting agency’s published value, at its published precision. Release schedules and shutdowns change the timeline.
- Timing trap
- Prices often move hardest in the minutes after a release. Compare forecasts only with quotes from before the release.
- Research prompts
- The published inputs already known (earlier monthly prints)
- Consensus and the spread of forecasts — the disagreement is information
- What happens to the timeline if the release is delayed
- On PropBetEdge
- PropBetEdge Predictions runs economics models in research and private shadow states first; the research board ↗ shows what is public and what is not.
- What moves the probability
- Announcements, schedules, charts and award processes — often driven by organisations rather than measurements.
- What settles it
- Frequently a list of named publishers or an official body. The rule may key off an announcement rather than the event itself.
- Timing trap
- News arrives unevenly. A culture market can reprice in seconds on one report, and quiet for weeks otherwise.
- Research prompts
- Production and scheduling news from the named reporting sources
- What counts as an announcement of a date
- How the bucket ladder (before Jan, before Apr…) fits together
- On PropBetEdge
- Culture and entertainment markets are not modeled by PropBetEdge Predictions today. Read them with the same rule-first discipline.
08 · Market Intelligence Checkup
Ten questions before you rely on a price
Use the checklist on any market you are researching. Then test what stuck with the optional quiz.
Tick each statement only when it is true for the market in front of you.
Saved in this browser only. A checklist is a discipline, not a prediction.
One question per chapter idea, each explained and linked back to the section that teaches it. Practice only — nothing is graded or stored beyond this browser.
Questions readers ask
Does PropBetEdge let me trade event contracts?
No. PropBetEdge is an independent research and education network. It does not execute, route or hold event-contract trades. Venues such as Kalshi and Polymarket set their own rules, fees and eligibility.
If PropBetEdge’s model disagrees with the market, is that a profitable opportunity?
Not by itself. A disagreement is a hypothesis about accuracy. It has to survive fees, the spread, available size, the model’s own error and the possibility that the market knows something the model does not — and even then the outcome is uncertain.
Is the market price the true probability?
It is the level where participants are willing to trade. It often carries useful information, but it is shaped by liquidity, fees, the spread and who is trading. Treat it as evidence, not truth.
Why do the contract examples not show prices?
Live prices go stale within minutes. The playbook shows real contract rules (read on October 12, 2026) and uses clearly labelled illustrative numbers to teach price mechanics.
Is this material free?
Yes. Learn is free to read. Some PropBetEdge tools linked from the playbook, such as model probabilities on the Predictions desk and Compare, are included with All Access.
Keep learning
Prediction Markets Academy
How sports, culture and weather markets differ.
Platform guideKalshi explained
Event contracts, order books and settlement sources.
Platform guidePolymarket explained
Outcome shares, liquidity and geographic access.
LessonExpected Value
Why an average over many outcomes is not a forecast for one.
LessonSample Size and Uncertainty
How much evidence a track record really holds.
GlossaryPrediction-market terms
Event contract, spread, depth, resolution source and more.