Open-source political risk intelligence cross-referencing prediction markets, polling data and verified media. Global by design. Brazil 2026 · US midterms 2026.
Brazil: Lula reelected, 51.79% to 48.21%. On the eve, Polymarket gave 83.40% to Flávio Bolsonaro and 9/10 polls had him ahead. Both backed the loser. A Supreme Court justice extended voting to 7pm. As Flávio led the count, Lula's price rose 67.20 to 89.10. https://t.co/pRlWg3iWns
Brazil runoff in two days. Polymarket: Flávio Bolsonaro 77.70%, Lula 22.50% (confirmed reading, Oct 22). Datafolha: a tie at 50% of valid votes. Not a contradiction: price measures the chance of winning, polls measure voting intention.
https://t.co/Jl5j3dEqWt
Two nights, both candidates on Brazil's top newscast. What did the money do?
Polymarket: Flávio Bolsonaro 76.45% before his Oct 16 interview, 76.65% next morning. Lula 23.50% both times.
Polls: Flávio leads 15 of 17 since round one, by 0.9 to 9 pts.
https://t.co/pRlWg3iWns
144 days measuring an election live. Since May 14, AFOS has crossed Polymarket prices, TSE polls and the press. On the eve, 5 of 9 national polls had Lula ahead; since Sep 10 the price had Flávio Bolsonaro ahead. Result: Flávio 47.03% vs Lula 45.16%.
https://t.co/dIrjpicR1P
Brazil, Oct 2: on Polymarket the Flávio Bolsonaro vs Lula gap narrowed from 23.55pp to 15.25pp in under 21h. Vox Brasil, the day's only national poll, went the other way. And TV Globo cancelled the debate after TSE and Supreme Court rulings.
https://t.co/yf0o6PXl9r
Three instruments, one question: who controls the US House.
Binary: 7.50%.
Seat distribution: only brackets it, 5.47% to 11.33%, since 218 falls inside the "215-219" band.
Polling: D+7.58, and it can't answer: it measures votes, the House is decided in seats.
No subtraction.
Polymarket, Sep 26: Flávio Bolsonaro 56.15%, Lula 42.50%. The gap reopens to 13.65pp.
In the 6 national polls of Sep 24, Lula leads the 1st round in 4 and Flávio leads the 2nd round in 4.
The contract pays the election winner, not the 1st-round leader.
https://t.co/P1ExPnQgRz
The gap between Lula and Flávio Bolsonaro on Polymarket moved 25.30 points in five sessions: from Lula +18.60 to Flávio +6.70, the largest weekly move since Apr 14.
The second-place contract confirmed it in mirror: Flávio -13.00, Lula +15.00.
https://t.co/XAV5YxQjxs
AFOS Weekly No. 6.
Republicans reached the top of their House series: 14.50%, above the 13.50% of July 29.
The polling average fell to D+5.17 from D+6.07 with no new poll published. Six rounds aged out of the 30-day window.
https://t.co/HlmOyxBJms
One side of the US midterm control contracts did not move a tick.
Republican House 11.50%, 30 of 30 readings.
Republican Senate 49.50%, 29 of 29.
On USD 63,454 of new money.
The other side's +1.00pp is +0.49pp once you normalize the pair.
https://t.co/10rTNzhsaS
Brazil's presidential contract compressed 9.70pp Aug 24-28, the largest since the series began in April.
Both ends set a record at once: leader -5.00pp, challenger +4.70pp.
Two pollsters published the runoff the same day: 47-42, and 46-41.
https://t.co/hIFk6IJT8x
The US midterms divergence record went public 25 Aug, 70 days before the vote. 379 polls from 63 pollsters, 377 traced to the pollster's own source URL. v1, pre-electoral, and labelled as such: no certified result to check it against yet.
https://t.co/abJPExk1BV
The market pays 86.50% on Republicans losing at least one of their 53 Senate seats.
And still prices control of the chamber as a coin flip.
Both are true: they can lose two seats and keep the majority.
https://t.co/BFm6DDJROI
A poll said 34% undecided. The market said 95.8% certain. The candidate polling at 7% won.
The market did not inherit the poll's error. It inherited its average and discarded its uncertainty.
Wisconsin, August 11.
https://t.co/MPdwvwjA4r
New validated case at AFOS.
In the world's largest election (India 2024, ~980M voters), the prediction market and the polls were wrong together: they called Modi's NDA but projected 350-400 seats. Result: ~293, and the BJP lost its solo majority.
Open data + Harvard DOI 👇
🇫🇷 France 2024 is AFOS's 10th validated case.
The deepest market (US$917k) priced the Rassemblement National ~99% to be the largest single party. Right: 143 seats.
The RN-majority hype only lived at low volume. A divergence needs volume.