Five calls, five hits. It’s the first clean sweep since we started, and the calibration note at the bottom explains why it’s worth less than it looks.
The board
| # | Market | Platform | Our call | Score | Market price at publish | Actual result | Hit / Miss | What moved it |
|---|---|---|---|---|---|---|---|---|
| 01 | Bitcoin price on Friday at 5pm EDT? ($82,000 or above) | Kalshi | NO | 31 | Yes 31c, No 69c | Settled near $79,100 | HIT | The rally stalled in the $79K to $81K band and never tested the strike |
| 02 | Core PCE YoY – July 2026 (3.4% or higher) | Polymarket | NO | 18 | Yes 15c, No 85c | 3.3% | HIT | Core came in at 0.2% monthly, exactly on consensus |
| 03 | Ethereum price on Friday at 5pm EDT? ($2,200 or above) | Kalshi | YES | 80 | Yes 82c, No 18c | Held near $2,500 | HIT | ETH spent the whole week above $2,490, roughly 13% clear of the strike |
| 04 | Liverpool FC vs. Nottingham Forest FC (Forest to win) | Polymarket | NO | 24 | Yes 24c, No 76c | Liverpool 2-2 Forest | HIT | Forest took a point at Anfield, not three |
| 05 | Crystal Palace FC vs. Manchester City FC (Palace to win) | Polymarket | NO | 24 | Yes 24c, No 76c | Manchester City won 4-1 | HIT | City scored four at Selhurst Park and the game was over by the hour mark |
01. Bitcoin above $82,000 on Friday. Bitcoin opened Friday at $80,261.86, its highest open since May 15, then drifted lower through the morning to $79,560 by 7:00 a.m. ET. It finished the session inside the $78,000 to $80,000 band, which is where Polymarket’s parallel bracket contract settled. Clearing $82,000 would have needed about a 3.5% push in a single session off an already overbought tape. The 82-level RSI we flagged on Monday did what overbought readings usually do, which is nothing dramatic, just a stall.
02. Core PCE at 3.4% or higher. The BEA published July personal income and outlays on Wednesday morning. Headline PCE held at 3.7% year over year, a tenth above the FactSet consensus, while core rose 0.2% on the month and 3.3% on the year. Core has now printed 3.3%, 3.4%, 3.3% and 3.3% across April through July, so the base case was a repeat and it repeated. One honest caveat: Employ America’s post-print tracker put the unrounded core figure at 3.34% year over year. The contract settled NO on the published rounding, and a few basis points of firmness in healthcare or airfares would have flipped it. An 18 was too confident for a call that close to the line.
03. Ethereum above $2,200 on Friday. ETH opened at $2,511.31, traded at $2,505.68 by 7:00 a.m. ET and sat around $2,496 at midday, above its 20, 50 and 200-day moving averages throughout. A drop below $2,200 would have meant a 12% single-day collapse with nothing on the calendar to cause one.
04. Nottingham Forest to win at Anfield. Forest drew 2-2. They got a result on merit and still lost the contract, which is the difference between a three-way soccer market and a two-way one. The NO covered both a Liverpool win and a draw, and that structural edge is most of why the price sat at 24c.
05. Crystal Palace to beat Manchester City. City won 4-1 at Selhurst Park. Palace had beaten City in each of the last two meetings, which is what kept the price up near a quarter, but two results is not a pattern.
Summary for the week
The week: 5 of 5. Running total: 11 of 15 across the last three editions. Edition 01 stays out of that number until its contracts are checked against the settled pages rather than a paste.
Now the calibration, which matters more than the sweep. Four of the five were NO calls on outcomes the market already priced between 15c and 31c. We agreed with the crowd four times and got paid for it four times. The single call where we took a real position against the price, Ethereum at 80 against an 82c market, was also the one with the least uncertainty in it.
A perfect week built out of consensus is worth less than a 3 of 5 built out of disagreement, because only the second kind tells you the model has an edge. The board this week was short on the thing this publication is for. Next week’s selection needs to rank candidates by distance from the market price, not by how comfortable the call feels.
Disclaimer: PredictGeek is an independent publication. Nothing here is financial advice or a recommendation to trade. Prediction markets carry real risk of loss. Trade only with money you can afford to lose.
