# Odd Lots: Coming First in Round One Was the Worst Thing That Happened to Him

Source: Realpha Blog (blog.getrealpha.com)
Original article and charts: https://blog.getrealpha.com/en/blog/oddlots-2026-11-05-james-bosworth-on-what-comes-after-lula-s-stunning/

> Notes on the 5 November 2026 episode of Odd Lots, where hosts Tracy Alloway and Joe Weisenthal talk with James Bosworth — non-resident fellow at the Stimson Center and author of the Latin America Risk Report — about Brazil's upset: the night before the vote Bosworth had Flavio Bolsonaro at a 64% chance of winning, the prediction markets had him at 80% and briefly 87%, Lula won by four million votes, and turnout moved less than 2% between the two rounds. Educational listening notes only — no investment advice, no price targets, and the judgements here may be wrong.

Published: 2026-11-05
Locale: en
Tags: Brazil election, podcast-notes, prediction markets, Latin America, election forecasting


![A wide avenue at dusk, an empty bus standing dark at the kerb in the foreground over wet asphalt, and behind it a queue of voters stretching away down the avenue toward a lit school gate far in the distance, the figures thinning as they recede into the evening](/covers/oddlots-2026-11-05-james-bosworth-on-what-comes-after-lula-s-stunning-cover.png)

> Order and disorder are a matter of fortune; want and success, of fate; high and low station, of timing. ... The tree that stands out from the wood will be broken by the wind; the mound that rises above the bank will be scoured by the current.
>
> —— Li Kang, "On Destiny" (Wei, c. 3rd century; my translation)

In the 5 November 2026 episode of Odd Lots, hosts Tracy Alloway and Joe Weisenthal talk with James Bosworth — non-resident fellow at the Stimson Center and author of the Latin America Risk Report — about why Brazil's election flipped (the episode was recorded on the afternoon of 3 November). Bosworth says that the night before the vote he had Flavio Bolsonaro at a 64% chance of winning, the prediction markets had him at 80% and at one point 87%, and Lula won by four million votes. His mechanism: the first round was a referendum on Lula, and once Flavio came first, all the attention moved onto him, turning the run-off into a referendum on Flavio instead. That explanation needs Flavio's own baggage to be heavy enough — and in the same episode Bosworth says that with Sao Paulo governor Tarcisio de Freitas as the candidate, Lula would have lost.

## The strongest version of his case

For the past six to eight years Latin America — and, Bosworth says, much of the world — has been in an anti-incumbent environment where every election is a referendum on the incumbent. Lula sat at 45% in the polls, so the first round was a referendum on Lula and had little to do with who Flavio Bolsonaro was. Coming first in that round, Bosworth says, was the worst thing that could have happened to Flavio: the spotlight swung onto him, and Lula got room to catch up from behind.

![Two panels side by side. In the left panel a cone of light shines down from above onto the figure on the left while the figure on the right stands in shadow; in the right panel the cone has moved onto the figure on the right and the left figure has gone dark, so the light has switched sides entirely.](/figures/oddlots-brazil-spotlight-shift-en.svg)

There was plenty to put under that spotlight. Flavio's father Jair Bolsonaro was convicted over the attempted coup of 2022–2023 — the 8 January 2023 storming of government buildings in Brasilia, shaped much like 6 January 2021 in Washington. Flavio himself was accused of money laundering five years ago as a congressman; this year he was tied to Daniel Vorcaro of Banco Master, a large Brazilian financial scandal, and was caught on WhatsApp asking the banker for millions of dollars. Just before the vote a Brazilian magazine connected him to pharmaceutical lobbyists and prostitutes on a boat. When he fled Brazil, the plane that carried him from Argentina to the United States was tied to a money-laundering investigation in Argentina and a drug-trafficking case in Texas.

Asked at a talk in New York what Flavio could do better, Bosworth answered: change his last name.

Lula's side changed how it played. He got his first smartphone, and his team started a WhatsApp group called Lula's Army with more than ten million people in it — over 5% of Brazilian voters. On election day certain mayors shut down bus services, and the group went into gear and started driving voters to the polls itself. In the debates Lula landed a run of thirty-second clips, the right length to share on social media; Bosworth says he does not know whether Lula meant to, and that the effect was there either way. Brazil is among the most online countries on earth, and Lula had never understood social media.

The arithmetic underneath is small. Turnout moved less than 2% between the rounds, and another 1% to 2% came from people who had cast blank or invalid ballots the first time. In Brazil that is a couple of million votes, and this election was decided by four million.

![A long bar standing for the entire vote, with only a hair-thin sliver marked at its left end; blown up, that sliver splits into a turnout shift of under two percent and a one-to-two percent swing in spoiled ballots, followed on the right by a very large block standing for the winning margin.](/figures/oddlots-brazil-thin-margin-en.svg)

The same pendulum explains the "orange wave." Bosworth leans toward reading the run of pro-Trump winners as a swing in an anti-incumbent environment, and puts less weight on a broad rightward shift. Trump has been going around saying he is nine for nine in recent elections, a number pulled out of the air — though before Brazil it was not far off. He expects an orange tilt to persist for two more years, because every Latin American leader, right, left or centre, has to respond to Trump: Lula won, and still has to deal with tariffs and with Trump, as does Claudia Sheinbaum in Mexico, until Trump leaves office in January 2029.

![On the left, a pendulum swung out to its right-hand extreme, with an arc and a return arrow showing it will swing back; on the right, a one-way ramp rising to the upper right, with an arrow that never turns around.](/figures/oddlots-brazil-pendulum-vs-ramp-en.svg)

The Iran war travels through the oil price. When fuel and prices rise, incumbents suffer — Chile, Colombia and Peru all show it. Brazil exports oil and soy, commodity prices have been higher over the past six months, and while the Caribbean and Central America were hit hard, Brazil's net effect is close to a wash. That may have helped Lula, and Bosworth says it will be a long time before the data lets anyone see it clearly.

## The other side

Joe asked the question that most needed asking: all of the negatives were already true before the first round, and Flavio still came first. So what the mechanism has to explain is why the same material changed what it did between the two rounds. "Attention moved" sounds reasonable, and there is no way to measure it in advance.

The vote arithmetic makes that worse. A turnout shift under 2% plus 1% to 2% of spoiled ballots switching is a gap any story fits inside. Tracy Alloway said it in the wrap: everything makes sense in hindsight. That same 2% can be written up as a referendum changing sides, or as Michelle Bolsonaro's public spat with Flavio costing him another slice of an already weak share of women voters, or as the weather.

There is a further layer. Bosworth's own model had Flavio winning. He gave him 64%, said he was essentially at a coin flip before the first round, and that all the data and all the polls pointed to Flavio. His words: he cannot base a prediction on vibes. The side that won had nothing but vibes going for it until the ballots were counted. The mechanism he now offers is one he assembled after knowing the answer.

![A horizontal axis split left and right, with the centre line standing for a coin flip. All three forecast points sit on the right half, while the marker for the actual result stands alone on the left half.](/figures/oddlots-brazil-forecasts-one-side-en.svg)

Two explanations crowd each other inside the same episode. Bosworth says that with Tarcisio de Freitas, who won the governorship of Sao Paulo, or any other right-wing candidate, Lula would have lost — he is convinced of it, and says every Brazilian analyst is convinced of it. That claim moves the weight onto the candidate's personal quality: if swapping the person flips the result, the "first round becomes a referendum on the leader" mechanism did very little work, and Flavio's scandals and his surname did most of it. Both can be true at once, but the explanatory shares have to add to one, and neither can be full.

Joe pushed once more. In much of the world, he said, being seen as too close to Trump should carry an electoral penalty — Trump's Iran war pushed costs up everywhere, and Nigel Farage's polling in the UK has been sliding on exactly that reading. Bosworth's answer: it has not shown up in any poll he has seen; people are angry at inflation without making the logical leap, all elections are local, and higher prices get blamed on the incumbent. That answer stops at the polls he has seen. Not showing up and not being measurable are hard to separate here.

![A bar of fixed length cut into two blocks by a divider in the middle, with an arrow on each side pushing against the divider while the bar's total length never changes.](/figures/oddlots-brazil-explanation-budget-en.svg)

One last thing I noticed. In the same conversation Bosworth hands out several high-confidence numbers: 96% that the Democrats take the House; the Senate closer to a coin flip than Nate Silver had it the week before, so no call; and better than 90% that Maria Corina Machado wins a free and fair Venezuelan election held tomorrow, with the odds rising if expatriates can vote — about a quarter of Venezuela's population has left the country. Those numbers come from the same person on the same day as the miss he just owned. His verdict on the prediction markets is that they missed and were overconfident on this election; he missed too, by less. As he puts it, who got it less wrong becomes a real question in these situations.

## Which side holds up

Start where readers get stuck. When the person everyone said would win loses, the first reaction is usually "so forecasts are useless," or the reverse, "80% always leaves two chances in ten." Both are too quick. Bosworth names the fork himself: you always wonder whether this was the 20% landing or whether they got it wrong. His answer is the second — they were overconfident.

So how do you tell? He gives the test in passing, and I think it is the most valuable line in the episode. All the data pointed to Flavio, he says, and you have to factor in that the polls were herding in one direction and correlated with each other. A cluster of correlated samples looks like a lot of evidence while being one piece of evidence copied many times. At 87%, the prediction markets were looking at a dozen polls saying the same thing; if those dozen share one set of sampling assumptions, they are a dozen copies of one number. This has nothing to do with the size of the lead and everything to do with independence.

![In the left panel twelve circles sit almost on top of one another and look like a single clump; in the right panel the same twelve circles are spread across the whole band. The count is identical and the spread is not.](/figures/oddlots-brazil-correlated-polls-en.svg)

So my read: the half of his case that holds up is the institutional half. A two-round system exists so this can happen — Bosworth says the first-round winner usually wins, and that the point of the run-off is to give voters that second look and let a president build a majority coalition. Attention shifting from incumbent to challenger has a structural place in that design. What does not hold up is the share: it cannot explain why the same scandals did no work in round one, and it competes for the same explanatory space as "Tarcisio would have won."

He also hands over the condition that separates the two readings: Argentina. There is an Argentine election next year, and the start of this right-wing wave was Javier Milei. If Milei holds on to re-election, the rightward-shift reading gains; if he goes the way of other incumbents, the anti-incumbent pendulum gains. The other thing to watch is fuel inflation — whether voters keep charging it to their own incumbent or start charging it to Trump. Both resolve inside a year, and the two outcomes are distinguishable, which is what a condition should look like.

![Two balance scales side by side, the left one tipping to the right and the right one tipping to the left, as two outcomes of the same event push the weight onto opposite stories.](/figures/oddlots-brazil-argentina-scale-en.svg)

Venezuela offers a different kind of condition. Bosworth says there will be no election this year, because any election needs a six-to-nine-month run-up. The current Rubio plan is for Delcy Rodriguez to declare a vacancy in the presidency, seat a new electoral board, and hold a vote six months later, ideally next July. Rodriguez would prefer something closer to 2028 or 2029. One timetable, pulled from both ends. The oil signal is cleaner: Chevron is in, was in before, and will be in for the next fifty years; Exxon and Conoco are interested, Venezuela is a sea of oil, and no oil major will go in before it is certain that contracts will be respected. What goes in now are small firms doing quick deals. A ten- or thirty-year investment is a bet on whether the rule of law survives a change of government.

![A timeline running from next year out to 2029, with an arrow above pulling the voting date toward the near end and an arrow below pulling it toward the far end, the two forces working in opposite directions.](/figures/oddlots-venezuela-timeline-pull-en.svg)

## Worth a look

- The 5 November 2026 episode of Odd Lots, the source for this piece (recorded on the afternoon of 3 November, as the hosts note)
- James Bosworth's newsletter, the Latin America Risk Report, where the Brazil and Venezuela reads discussed in the episode appear
- The Stimson Center, where Bosworth is a non-resident fellow
- The Polymarket contract on whether Venezuela holds an election before year-end, quoted at about 8% when the episode was recorded
- The New York Times report, published shortly before the recording, that the US is in talks over a military presence in Venezuela

## The one thing to take with you

**Being in front changes how people look at you, which makes the lead itself a new risk.**

Coming first in round one, he says, was the worst thing that happened to Flavio. That line stopped me, because it runs against instinct — winning the first round is good news on paper, and its real effect was to swing every spotlight onto one person. The material on him was the same. What changed was how many people were looking.

Here is something I have tried that works outside investing too. Next time you are about to call something a sure thing, or eight-in-ten, write down two things: the number, and what you would have to see before admitting you were wrong rather than saying the two-in-ten landed. The second is harder to write, because it asks you to picture conceding before anything has happened. Put it away, and open that page once the thing has played out.