How Journalists Use Betting Markets to Forecast Big Events

The 2:37 a.m. refresh: why odds sneak into newsrooms

The copy desk is quiet. The coffee is not. A result ticker blinks on a side screen. Someone mutters, “Check the odds.” That late, it is not a bet. It is a quick read on new info. Prices move before quotes do. A market can react to a leaked exit poll, a weather hit, or one strong line in a debate. A good editor wants that signal, fast but safe. Used well, odds are not a headline. They are one light on the dash. This guide shows how to use that light with care.

Markets are thermometers, not oracles

Think of a betting or prediction market like a thermometer. It tells you how hot the consensus is right now. It does not see the future. The price mixes three things: what informed traders know, what they fear, and how much cash is in the pool. Thin cash can make wild moves. Thick cash can hide slow bias. Rules of the market also shape price. If the payout rule is odd, the price can drift from true chance.

If you want a deeper base, a classic paper gives a clear map of what these markets are and how they work. See a seminal overview of prediction markets from NBER by Wolfers and Zitzewitz at this research note.

When to use odds in a story: when a fresh event hits and you need a quick view of new data; when polls are stale; when expert views split and you need a crowd check; when you track risk over time. When to hold back: when the market is thin, when there is a clear rumor mill, when the contract rules are odd, or when the odds would crowd out stronger facts. Keep one line clear in your head: markets price beliefs, not fate.

Translating prices into probabilities (and where writers slip)

Readers think in chances, not odds. So translate, and show your math. Here is the short map:

  • Decimal odds (2.50) → implied probability p = 1 / 2.50 = 0.40 (40%).
  • Fractional odds (3/2) → decimal = 1 + 3/2 = 2.50 → p = 1 / 2.50 = 0.40.
  • American moneyline: Positive (+150) → p = 100 / (150 + 100) ≈ 0.40. Negative (−150) → p = 150 / (150 + 100) ≈ 0.60.
  • Positive (+150) → p = 100 / (150 + 100) ≈ 0.40.
  • Negative (−150) → p = 150 / (150 + 100) ≈ 0.60.
  • Positive (+150) → p = 100 / (150 + 100) ≈ 0.40.
  • Negative (−150) → p = 150 / (150 + 100) ≈ 0.60.

Need a quick refresher? A clear, short guide sits here: converting odds to implied probability.

Three common slips in news copy: first, not removing the house margin (also called the overround). If a sportsbook quotes both sides, the sum of p can be over 100%. Normalize by dividing each p by the total p-sum. Second, mixing markets. A “to win the vote” price is not the same as “to win the seat” if rules differ. Third, no time stamp. Prices move. Always log the time (UTC), source, and, if you can, the traded size or liquidity. Snapshots beat vibes.

A quick reality check: when markets beat polls—and when they don’t

On big nights, markets can add signal that polls miss. They ingest late mail talk, legal risks, and turnout weather. A clean example is how desks tracked the 2020 U.S. vote hour by hour. For a smart look at what markets can and cannot tell you on elections, see this graphic deep-dive from The Economist.

But do not dump polls. Polls measure stated intent and can show who is still unsure. Good polls with sound weight can beat noisy prices, esp. when markets are thin or when big traders “anchor” the line. If you want caveats on polls and how they did in a recent cycle, see this note from Pew Research Center. The best desk runs both: polls for base rate, markets for live moves.

Sourcing and hygiene: where odds come from and how to vet them

Not all sources are equal. An academic market like the Iowa Electronic Markets has long, careful rules and a research bent. For background and access, start at the Iowa Electronic Markets site. Exchange markets, where people trade with each other, can be fast and deep on big events. See how a major exchange structures its prices and rules here: what is the Betfair Exchange?

Hygiene list for your desk:

  • Log the exact contract name and resolution rules.
  • Record a UTC time stamp and a snapshot of price and size.
  • Prefer exchange last-trade or mid-price over a single quote.
  • Note liquidity (volume, matched bets) to flag thin markets.
  • If you cite a book, remove the margin before you write the %.
  • Store the raw file (CSV or JSON), plus a screen grab if policy allows.

Method box: from odds to Brier scores

If you want to test if markets help your beat, score them. A simple and fair metric for binary events is the Brier score. For one event, Brier = (p − o)2, where p is the implied probability you wrote down, and o is 1 if the event happened, 0 if not. Lower is better. Over many events, average the scores. Also plot calibration: when markets say 60%, do those things happen about 6 times in 10? A short, clean definition sits in the NOAA verification glossary at this Brier score entry.

Below is a small audit table you can adapt. Each row is a real event. Times and prices are snapshots, not live streams. Always add notes on liquidity and key news near the time.

2016 UK EU referendum (Leave) Exchange (Betfair) 2016-06-23 21:00 0.25 (25% Leave) Poll avg: slight Remain edge Leave (0.25−1)² = 0.5625 Price flipped late as early counts showed strong Leave areas
2020 U.S. presidency (Biden wins) Iowa Electronic Markets 2020-11-03 16:00 0.64 (64%) Poll avg: Biden popular vote ≈ 52% Biden (0.64−1)² = 0.1296 Mail-in count created sharp intranight swings on exchanges
2022 FIFA World Cup winner (Argentina) Exchange (pre–Round of 16) 2022-12-01 12:00 0.18 (18%) Top model odds ≈ 20% Argentina (0.18−1)² = 0.6724 Thin prices early; liquidity rose as bracket firmed
2022 U.S. “red wave” in House (≥250 GOP seats) Exchange contract 2022-11-07 20:00 0.35 (35%) Generic ballot: small GOP edge No (0.35−0)² = 0.1225 Hype ran ahead of district-level data; price eased on E-Day

Method notes: Implied probabilities use mid-price where possible. Polling/model lines are rough at-snapshot values and may differ by source. Brier is shown for the event as phrased in column one.

Ethics, law, and risk: the line between reporting and promotion

Your job is to inform, not to push bets. Be clear: “This is a market snapshot, not advice.” In the U.S., some “event contracts” fall under rules set by the CFTC. Their advisory explains the risks and limits; see this CFTC page on event contracts. In the UK, the Gambling Commission tracks harm and sets strict rules on ads and safer play. For figures and policy, see these statistics from the UKGC.

House rules that help: label any affiliate tie; use rel="sponsored" or rel="nofollow" on such links; never frame a price as a sure thing; avoid “go place a bet” lines; include a responsible gaming note; and respect local law notes in your CMS for geo pages.

Write it in: five patterns that respect readers

Here are five small lines that fit in straight news and do not overclaim:

  • “Markets currently price a 38% chance of X, as of 09:00 UTC.”
  • “Implied probability from exchange odds suggests Y is now the base case.”
  • “Odds diverge from polls by about 10 points; liquidity remains thin.”
  • “Prices moved after the court filing; volume was light.”
  • “This is not a forecast; it is a snapshot of what traders pay today.”

Want to see how desks handle real-time data on big nights? Nieman Lab has a good wrap on live data craft at this article on election-night coverage.

Pitch it to editors: why markets belong in the budget

Odds are not your headline. They are your sidebar meter. They add speed, they add a check on bias, and they give readers a clear number they can track across days. Teams that use them well also learn to score their own calls, which raises trust. For a research angle to include in your pitch, compare exchange prices with groups of trained forecasters. The Good Judgment Project hosts papers and notes at this research archive. Editors like formats that teach the audience; a small “market watch” box can do that week by week.

Where independent reviews fit in your workflow

Now and then you need a neutral list of licensed platforms, their data access rules, and their market types (fixed odds vs exchange, live vs pre-match, etc.). We keep a short list in our desk notes so we can check rules fast and avoid gray sites. For a plain, non-promotional overview you can scan in two minutes, see this page. Treat it as a directory, not a signal to place bets. If your org has affiliate ties, label them in your “About” or “Disclosure” page and on first mention.

FAQs and edge cases

Can markets be manipulated?

Yes, in thin markets. A small stack of cash can nudge price for a short time. Signs to flag: big moves on low volume, one trader crossing the spread in bursts, or price spikes around PR drops. Note this in copy (“liquidity is light; prices can whipsaw”). Community forecasting sites study these effects and how to improve calibration; see the work at Metaculus Research.

Is it legal to cite odds?

In most places, yes. You are reporting market info, not taking bets. Still, check local rules on linking to operators and include safer-gambling notes. If your outlet blocks links to betting sites, cite the source name, time, and price without a URL.

What if markets and polls split wide apart?

Write both and explain the gap. Markets may see late mail, court risk, or turnout shifts. Polls may catch age groups that do not trade. If one source is thin or stale, say so. If both are thin and stale, say that too, and lean on base rates and expert quotes.

What about black swans?

Rare shocks are, by nature, hard to price. If a tail risk is in play (war, sudden ban, new rule), describe the scenario and why prices may lag. Avoid big claims off one bar of data.

Toolkit and update plan

Tools that help:

  • CSV/JSON exporters from exchanges or academic markets.
  • A small script to convert odds to implied p and remove margin.
  • A chart that plots market p vs poll p over time.
  • A shared log of snapshots with UTC stamps and liquidity notes.
  • A style card for hedging lines and disclaimers.

Update plan: set a review date each quarter to refresh links, redo the audit table with 3–5 new cases, and check that your law notes are current. For a sense of how reader trust shifts over time and by country, bookmark the Reuters Institute’s Digital News Report at this benchmark.

Mini checklist before you hit publish

  • Did you translate odds to probability and show the time and source?
  • Did you check rules of the market and remove the margin?
  • Is liquidity thin? If so, did you say so?
  • Do you balance markets with polls and base rates?
  • Are disclosures and safer-gambling notes in place?
  • Do you avoid language that urges readers to bet?

Responsible use and disclosures

This guide is for editorial use. It is not betting advice. Gambling can be addictive. If you choose to take part, do so legally and responsibly, and only if it is allowed in your country or state. If this article links to any betting or review site and there is any commercial tie, that link should be labeled accordingly (rel="sponsored" or rel="nofollow") and a clear disclosure should appear on the page.

Method and sources

  • Core concepts: prediction markets as information aggregators (see the NBER paper linked above).
  • Odds math: implied probability formulas (see Investopedia link above).
  • Scoring: Brier score definition (see NOAA glossary link above).
  • Context: election graphics (The Economist), polling review (Pew Research), newsroom live-data practices (Nieman Lab), forecaster research (Good Judgment), community forecast research (Metaculus), trust benchmarks (Reuters Institute), and compliance (CFTC, UKGC).

Editor’s note: Keep a living copy of this guide in your internal wiki. Add your own cases, with snapshots, each time you cite a market. Over time, you will learn where markets shine on your beat—and where they do not. That is the point.