Brazil presidential election 2026: Lula vs Flávio Bolsonaro

Kalshi · PolymarketClosedNo clear news effect

99.9%Lula

Leading outcome on Polymarket, closed Oct 26. All times UTC.

moved the price 7 didn’t 30? no news found 3

Loading prices…

Question 1 of 5

Does news move this market?

No clear effect of news on this market yet (12% of stories against 10% of random moments)

Lift over random1.2×Intervals overlap: no clear effect
Stories that moved the price12%85 of 731 · 95%: 10–14%
Random moments10%95%: 7–12%
Stories measured731Sep 15 – Oct 26

By source type

The bar is the share of that type’s stories that moved the price, its 95% interval lighter; the black tick is the random-moment rate (10%).

  • Newsroom10% 42 of 432
  • Unlisted accounts17% 18 of 109
  • Journalist17% 14 of 81
  • Wire12% 8 of 69
  • Unlisted sites4% 1 of 27
  • Candidate15% 2 of 13

Question 2 of 5

How does it react?

−14−70+7+14−1 hPost+15 min+1 h+4 h+24 h+72 h
All 85 stories that moved the price Newsroom (42) Journalist (14) Unlisted accounts (18) Random moments, 95% band

Each story is signed by its own change at +1 hour, so the curve always rises to +1 h. Read what comes after: how much of the move is still there at +24 h and +72 h. Several stories inside one move share its peak.

Median time to peak4.1 hhalf within 1.8 h–10.3 h
Left after a day60%median share of the peak
Mostly reversed21%under a quarter left at +24 h
Price moved first8 of 852+ pts the same way in the hour before

Time to peak, one dot per story

0h6h12h18h24h

Share of the peak left after a day

0%50%100%150%

Question 3 of 5

Which sources move it?

Sources with five or more stories, by how often their stories moved the price. Under 15 stories the interval is too wide to rank on, so those rows are greyed.

SourceTypeStoriesMoved the price · tick: random 10%Median change
g1.globo.comNewsroom2520% 5 of 253 pts
www1.folha.uol.com.brNewsroom2020% 4 of 202.1 pts
@GugaNoblatJournalist7120% 14 of 712.9 pts
apnews.comWire1817% 3 of 184.4 pts
noticias.uol.com.brNewsroom4914% 7 of 493 pts
@CNNBrasilNewsroom1613% 2 of 163.35 pts
bbc.comNewsroom4411% 5 of 442.6 pts
poder360.com.brNewsroom157% 1 of 152 pts
oglobo.globo.comNewsroom605% 3 of 603.9 pts
veja.abril.com.brNewsroom205% 1 of 203.3 pts
metropoles.comNewsroom353% 1 of 352 pts
cnnbrasil.com.brWire230% 0 of 23—

Question 4 of 5

Which moves had no news?

10 of 21 sharp moves had no credible post in the 30 minutes before them. For those on Polymarket, the money that went the move's way in that half hour, and who put it in.

WhenMoveEarly moneyWalletsTop walletLargest early traderAt peak
Oct 25, 22:33———Not computed—
Oct 25, 21:46———Not computed—
Oct 25, 19:58———Not computed—
Oct 15, 15:59———Not computed—
Oct 4, 02:06———Not computed—
Oct 3, 23:21———Not computed—
Oct 3, 21:34———Not computed—
Sep 21, 00:44———Not computed—
Sep 20, 23:45———Not computed—
Sep 18, 17:46———Not computed—

Question 5 of 5

Do the venues agree?

Kalshi repriced first4of 21 detected moves
Polymarket repriced first44 ties, 9 can't tell
Usual gap1.2 ptsmedian, 2 h before a move
Gap after a move3.5 ptsmedian peak, 2 h after

With 8 decided cases, the split is what a coin flip would give (sign test p = 1.00): no venue leads.

Resolution rules

  • Kalshi resolves on the AP's call, or on whoever is inaugurated if the AP names no winner; Polymarket on a consensus of credible reporting, with the TSE if ambiguous.
  • If the result isn't known by June 30, 2027, Polymarket resolves to “Other”; Kalshi stays open until a year after the election.
  • Both settled to Lula, so neither difference came into play.

News that moved it

Every event

Sort

Findings

  • Tentative

    The average story doesn't move this market

    85 of 731 stories (12%, 95% interval 10%–14%) were followed by a price change bigger than 95% of this contract's changes at random moments, against 10% of random moments themselves. Most coverage repeats what's already priced in; the moves come from a few specific stories.

  • Tentative

    Stories with X posts move prices more often

    Stories that include an X post moved the price 14% of the time (50 of 365); newsroom-only stories 10% (35 of 366); random moments 10%.

  • Tentative

    Journalists lead the sources that move it

    Journalists 17% (14/81), unlisted sources 14% (19/136), wires 12% (8/69), newsrooms 10% (42/432); random moments 10%.

  • Strong

    Moves after news mostly stick

    A day later, the median move kept 56% of its peak; 26% had mostly reversed (85 moves).

  • Strong

    The price rarely moves before the story

    In 7 of 85 stories that moved the price, it had already moved 2+ points the same way in the hour before the first post or article.

  • Tentative

    Kalshi usually reprices first

    Where it could be told: Kalshi first 9, Polymarket first 3, within a minute 4. 14 of 85 moves opened a Kalshi−Polymarket gap 3+ points wider than usual.

  • Tentative

    Election days are when news moves this market

    On the two voting days (first round and runoff, counts included), 19% of stories moved the price (22 of 115, 95% interval 13–27%), about twice the rate on other days (10%, 63 of 616). The rest of the campaign was mostly priced in.

  • Tentative

    Polls on their own rarely moved it

    Stories about polls moved the price 11% of the time (9 of 84, 95% interval 6–19%), about the same as random moments. The Quaest poll the night before the first round was the exception: g1 posted it 25 minutes before Flávio rose 10 points.

    g1: Quaest, 2º turno: Flávio 44%, Lula 42% ↗

  • Tentative

    News-first collection found 4 of the 11 known moves

    Of the 11 moves the backtest traced to a story, collecting news first (without looking at the price) found the Quaest poll and the first-round count, but not the Palver poll, the Luchsinger interview, the runoff count opening, or the two runoff-night scoops. X added context before the voting extension — Noblat 19 minutes before, on the TSE weighing an extension; Exame 32 minutes before, on Flávio's campaign petitioning the TSE — but not the decision itself.

    Noblat, 19 min before the extension move ↗Exame, 32 min before ↗

  • Caveat

    Busy nights need denser sampling than we used

    Each X query returned the 20 most engaging posts per 3-day slice, and election night produces hundreds. Guilherme Amado's scoop 14 minutes before Lula's 44-point rise (14.8k likes) didn't make the top 20 for his account group that night; the price-anchored search in the backtest did find it.

    Amado's scoop ↗

  • Caveat

    Newsroom headlines can't be timed to the minute

    Search indexes return an article's current headline with its first publish time. g1's article reads 'Lula amplia vantagem e vence eleição' with a time 16 minutes before the 22:08 move, about 90 minutes before the result was known; VEJA's live page reads 'Lula é eleito presidente' with a time before the polls closed. Live pages are excluded; ordinary updated articles can't be told apart, so X timestamps date the breaks.

    g1 article (headline rewritten after publishing) ↗

  • Caveat

    A price gap between venues here isn't clean arbitrage

    Kalshi resolves on the AP's call (or the inauguration); Polymarket on a consensus of credible reporting with the TSE as tiebreak. Gaps that opened after news can partly reflect those rules.

Method and limits

  • A story “moved the price” when the change within 1 or 4 hours was 2+ points and larger than 95% of the same contract’s changes at random moments. At a 9% random rate, about half of tagged stories could have qualified by chance: one tag is weak evidence, a market’s rate is not.
  • A story first seen from 30 minutes before a detected move to 10 minutes after it is merged into that move. Moving stories with no sharp move within 4 hours are shown as slow drifts.
  • “First post” means the earliest post found. X search returns a sample, and view counts are as of collection.
  • Wallets exist for Polymarket only; Kalshi accounts are private. Wallets that put $200+ into an early window are shown by public address. Trading early is not proof of inside information, and lifetime figures are not profit on the trade.
  • A venue gap is not arbitrage where the resolution rules differ. Here: Kalshi resolves on the AP's call, or on whoever is inaugurated if the AP names no winner; Polymarket on a consensus of credible reporting, with the TSE if ambiguous.; If the result isn't known by June 30, 2027, Polymarket resolves to “Other”; Kalshi stays open until a year after the election.; Both settled to Lula, so neither difference came into play..