Weekly Forecast Results: August 3-9, 2026

Written by:

PredictGeek

, Editorial Manager | Last updated:

August 10, 2026

We said every call would get graded, wins and losses both. For the week of August 3-9, PredictGeek got four hits and one miss, but that miss was an outlier.

#MarketPlatformOur callScorePrice at publishResultHit / Miss
01Cardinals beat Panthers, Hall of Fame GameKalshiYES71~49¢Carolina won 33–30Miss
02Typhoon Dolphin makes landfall in ChinaPolymarketYES8052%Landfall in Zhejiang, Aug 9Hit
03Disney beats $1.86 non-GAAP EPSPolymarketYES8488.5%$2.06Hit
04July payrolls at 150,000 or higherKalshiNO28ladderPayrolls fell 23,000Hit
05Seven or more M5.5+ earthquakesPolymarketYES7462%Exactly 7Hit

01. Cardinals beat Panthers. Carolina won 33–30 on the final play, an undrafted Haynes King touchdown run with no time left. Our published note said the honest estimate was nearer 56% than 71, so this is a coin flip landing the wrong way rather than a broken read. The one thing we got wrong was assuming veteran arms decide preseason openers. Both rookie quarterbacks outplayed the depth chart.

02. Typhoon Dolphin makes landfall in China. Dolphin came ashore near Yuhuan in Zhejiang at 5:30 p.m. local time on Sunday with sustained winds of 42 metres per second, then made a second landfall near Wenzhou. More than a million people were evacuated. The thesis was that 52% was a stale price written on July 30 and never updated after three agencies converged, and that held up. Our timing was off by two days, and only the contract’s Aug 15 backstop covered the slip.

03. Disney beats $1.86. Adjusted EPS came in at $2.06, a beat of about 10.8%, with record fiscal-Q3 revenue at Experiences. We marked ourselves at 84 against a market at 88.5 because rising consensus raises the bar, and that caution proved unnecessary. Under-confident, not wrong.

04. July payrolls at 150,000 or higher. Payrolls went negative, falling 23,000 against a consensus near +80,000, below every forecast on Bloomberg. May and June were revised down by a combined 103,000. This is the call we’re happiest with, because it came from reading the revision pattern rather than the headline. At 28 we were still too generous to the YES side.

05. Seven or more M5.5+ earthquakes. The USGS catalogue recorded exactly seven qualifying events, including two M6.3 shocks on August 5 near the Kermadec Islands and Sarangani in the Philippines. Seven is the minimum for a hit, so this landed by one event. Our base rate implied 8.9 per week, and the true rate this year is closer to 7, which is why the market sat at 62% rather than higher. Right answer, thinner margin than the score suggested.

Summary for the week

Four of five, 80%. More useful than the hit rate is the shape of it.

Our two most confident calls, Disney at 84 and Dolphin at 80, both landed, and the miss was our least confident call at 71. Confidence tracked accuracy, which is the only thing that makes a scoring system worth reading.

One change going into next week. A single preseason football game carries variance no amount of research removes, and the Geek Score rule pushed an honest 56% up to 71 to clear the threshold.

We’ll keep sports on the board, but we’ll favour series and season-level markets over one-off games, where the evidence can actually support the number.

Photo of author

Editorial Manager

The PredictGeek Editorial Team is a group of writers, researchers, and industry experts specializing in prediction markets, forecasting, data analysis, and probabilistic thinking. Our team combines rigorous research with clear, accessible writing to help readers understand the trends, platforms, and events shaping prediction markets.