StuPredicts

LET’S GOOOO! The US midterms are done and prediction markets X has been an absolute goldmine this week. Big wins, some painful misses, traders celebrating, and people still fighting over whether prediction markets actually beat the polls. • Joe Weisenthal (@TheStalwart): Absolutely loved election night. Says prediction markets did exactly what they were supposed to do, making confident calls before the election and updating ridiculously quickly as votes came in. • Nate Silver (@NateSilver538): Not having any of the victory-lap nonsense. Argues that the best polling models were ahead of the markets, and that traders were largely following information the forecasters had already produced. • G. Elliott Morris (@gelliottmorris): Has some receipts of his own. His FiftyPlusOne model gave Democrats a 98% chance of taking the House and 79% for the Senate. Says it was ahead of prediction markets on the big outcomes. • Luana Lopes Lara (@luanalopeslara): Points out that the NYT only launched its election needle at 8:37pm, when Kalshi was already pricing a Democratic House at 99% and Senate at 98%. • Luke Kawa (@LJKawa): Probably my favourite description of the whole evening. Prediction markets are sales, TV is quality assurance. • Andy Masley (@AndyMasley): Watching the election at a party and wondering why all the traditional election coverage feels so slow compared with prediction markets. • Justin Slaughter (@JBSDC): Pushes back on claims the markets had a disastrous night. Points to Kalshi’s figures showing favourites winning 479 of 487 resolved races and explains how he was following the markets to understand the developing Democratic wave. • Arin Dube (@arindube): Makes a useful point amid the celebrations. Yes, markets did well at picking winners. But they, along with other forecasters, underestimated how big the Democratic wave would be. • Jeremiah Johnson (@JeremiahDJohns): Gets stuck into the Brier scores and reckons The Argument/Split Ticket could finish with the most accurate election model, depending on the remaining results. • @JamesSurowiecki: Notices something quite strange on Election Day. Despite all the noise, Senate market prices barely moved for long stretches. Iowa was one of the exceptions. • @tracewoodgrains: Has a brilliant response to Kalshi comparing itself with Silver Bulletin and Split Ticket. If traders are reading those models and then performing about as well as them, surely that’s a pretty good advertisement for the models? • @JoePostingg: Reckons prediction markets are especially useful once votes start coming in. Before that, they’re often following the same forecasts as everyone else. Once the counting starts, the speed becomes genuinely useful. • Benjamin Freeman (@benwfreeman1): Challenges reporting suggesting election trading was stagnating. Says the comparisons weren’t giving Kalshi’s growth a fair hearing. • Emily Nicolle (@emilyjnicolle): Looks at the other side of the volume story. Polymarket’s US-election trading before the midterms was much lower than ahead of the 2024 presidential vote, while Kalshi was broadly flat. Interesting when everyone is celebrating record activity. • @MarshallCohen: Shares five takeaways from the election, including what market watchers saw regarding suspicious trading and why the Democratic victory could make life more difficult for the industry in Washington. • @DustinGouker: Puts the election volume into perspective. Kalshi reportedly traded $1.9bn across the midterm cycle. It did $4.2bn in sports in a single recent week, without parlays. • Krown (@KrownCryptoCave): Questions how an apparent 75% favourite could suddenly become an underdog. Then gets a Community Note pointing out the candidate never actually reached those odds. And that’s before getting into the actual money people made and lost! I’m also putting together a much bigger write-up on what the midterms really told us about prediction markets.

🇺🇸 US MIDTERMS: KALSHI VS POLYMARKET. WHO GOT IT RIGHT? With election results coming in, I’ve been checking how the two prediction markets performed. And crucially, how early they got it right. Here’s the scoreboard across 14 selected outcomes that have been called, using market prices from November 2, the day before the election. 📊 THE SCOREBOARD @Kalshi: 12–2 ✅ 12 correct favourites ❌ 2 incorrect favourites @Polymarket: 11–3 ✅ 11 correct favourites ❌ 3 incorrect favourites 🏛️ BREAKDOWN Senate: Kalshi 4–1 | Polymarket 3–2 Governors: Kalshi 6–1 | Polymarket 6–1 House district: Both 1–0 Senate control: Both 1–0 ❌ WHERE DID THEY GO WRONG? Two significant misses for both platforms: South Carolina Senate: Annie Andrews won despite being priced at… • Kalshi: 32.5% • Polymarket: 31.5% Texas Governor: Gina Hinojosa won despite being priced at… • Kalshi: 37.5% • Polymarket: 39.5% ⚖️ WHY DID POLYMARKET HAVE AN EXTRA MISS? Iowa Senate. Kalshi gave eventual winner Josh Turek a 51.5% chance. Polymarket gave him 49.5%. One technically got the favourite right, the other didn’t. But with prices that close, I wouldn’t award Kalshi much extra credit. ⏱️ THE BIG QUESTION: WHEN DID THEY GET IT RIGHT? Here’s how the scoreboard would have looked six days before the election, on October 28, using the same 14 eventual outcomes. Kalshi: 9–5 Polymarket: 9–5 Some forecasts had been pointing in the right direction well before then. Democratic Senate control, October 28: • Kalshi: 64.5% • Polymarket: 66.5% Texas Senate, James Talarico: • Kalshi: 66.5% • Polymarket: 68.5% ⏳ AND THE HOUSE? This could be one of the most interesting early calls. Democratic House control, October 28: • Kalshi: 90.9% • Polymarket: 92.5% By November 2: • Kalshi: 93.8% • Polymarket: 95.5% We’re still awaiting an AP call, so House control isn’t included in the scoreboard. 🎯 MY EARLY TAKEAWAY Kalshi 12–2. Polymarket 11–3. Very similar results, with the only difference being a virtual coin flip in Iowa. But a correct favourite days before an election deserves more credit than one that only emerges hours before the result. And even a 65% favourite should lose sometimes. The really interesting comparison will be how both platforms performed at 30 days, seven days and 24 hours before the election. That’s where we’ll get a clearer picture of how useful these markets actually were as forecasting tools.

Gonna do some live posts about the midterm election results and whether prediction markets got it right. Let’s begin… 🇺🇸 VIRGINIA SENATE: AP CALLS IT @MarkWarner (D) defeats @GenBert2026 (R), securing a fourth term. AP called the race at 7pm ET, as polls closed. At 7:45pm ET, with 12% reporting: 🔵 Warner: 57.88% 🔴 Mizusawa: 41.83% The first Senate race called on election night. Prediction markets had this one priced well in advance. • Oct 21: Warner around 98% on Kalshi and Polymarket. • Nov 1: 99% on both. • Nov 3, 6:56pm ET: Still 99%, four minutes before the AP call. Trading activity: • Kalshi: ~$237,135 in volume • Polymarket: ~$24,900 in volume

🇺🇸 US MIDTERMS × PREDICTION MARKETS SPECIAL The big day is here! • @Kalshi had Democrats at 94% to win the House and around 75% to take the Senate in its latest election-eve posts. A pretty emphatic forecast. • @Reuters reports more than $1.48bn traded on the midterms. @washingtonpost puts the wider election-market total at almost $1.9bn. Different datasets, but enormous figures. • @acdatacollectiv counted $431m in congressional trading through 29 October, across more than 28,845 markets on Kalshi, Polymarket and Polymarket US. • ACDC also found five contests accounted for 78% of volume among the 15 busiest congressional races. Texas Senate alone attracted $66.9m. • @DustinGouker shares @TickerTracker_ estimates showing elections make up just 0.5% of Kalshi's year-to-date contract volume. Sports and combinations account for 84.1%. Quite a contrast with the publicity. • @Kalshi says around five people view its midterm markets for every trader, with total visits 50 times the number of traders. A lot of people are watching without betting. • @Kalshi says Democrats are favoured in 22 of 36 governor races. But Republicans @ByronDonalds and @GregAbbott_TX were still around 61% favourites in Florida and Texas. • @CNBC reports the Texas governor race tightening, with @GinaHinojosaTX challenging Abbott partly over the local costs of AI data centres. • @KalshiPolitics had Democrats favoured in all seven key Senate races: Iowa 52%, Kansas 63%, Ohio 71%, Alaska 72%, Maine 73%, Texas 76% and Michigan 78%. • @DeItaone shares markets pricing a 73% chance of Democrats taking both chambers, against just 5.8% for a Republican sweep. 1/4