They Were Given Tomorrow's Headlines. Half of Them Still Lost
Idea clave
- Given tomorrow's front page for 15 trading days, participants got the direction right 51% of the time — a coin flip, with the answer in their hands.
- One in six lost everything. Not from bad news, from bad sizing: they bet big when unsure and small when certain.
- Five professional macro traders took the same test and all five finished ahead — mostly by sizing, barely by predicting.
Ruta de aprendizajeWhy did it move? Earnings, rates and the newsPaso 6 de 9
Antes que esta:What a Bond Actually Is, and Why Its Price Moves the Wrong Way
A partir de un vídeo de The Economist (@TheEconomist) — YouTube
The experiment
Tomorrow's news, in advance, with leverage. It should have been free money.
Researchers at Elm Wealth took 15 front pages of the Wall Street Journal from between 2008 and 2022. Each participant saw tomorrow's headlines, then traded stocks and bonds on that day with up to 50 times leverage, starting from $50. Positions were closed against what the market actually did.
Why knowing the news does not tell you the direction
Markets do not move on what happened. They move on the gap between what happened and what was expected.
The clip's worked example is exactly the one a beginner needs: “Say tomorrow there's a great jobs announcement, but everybody was already expecting that… the economy added a 100,000 jobs. People expected it was going to add 150,000 jobs. People are actually quite disappointed.” Good news, falling prices — because the headline never carries the benchmark it is being judged against.
That is the same machinery as a stock falling on good news, and it is why the participants managed 51% on direction. Half the information they needed — the expectation — was not on the page.
The models did not rescue it either
AI models took the same test, read direction better, and still handled the money badly.
They got direction right about 60% of the time — a real improvement on 51% — and were, in the clip's words, “pretty much as bad at sizing their bets correctly… took way too much risk.” Some lost money. The bottleneck was never the forecast.
What the professionals did differently
They were slightly better at predicting and dramatically better at deciding how much to put behind a prediction.
Five expert macro traders ran the same test. All five finished profitable and on average more than doubled their money. The clip is explicit that prediction was not the difference: “It was they changed their investment sizing massively according to their confidence. On some days, they chose not to bet at all.”
What the amateurs did
- Bet a similar amount regardless of confidence
- Piled on leverage when least sure
- Never sat a day out
What the professionals did
- Sized in proportion to confidence
- Took nothing when the read was unclear
- Went large only on the clearest days
What this means for someone starting out
You have been optimising the wrong half. Everyone asks what to buy; the experiment says how much matters more.
The closing argument is worth taking literally: investors spend enormous effort on which stock or which bond, while “almost if not more important is the question of… how confident am I that the particular investment I'm making is going to pay off… and how should I size my bet.” That is the entire subject of how much to bet, and this experiment is the cleanest evidence for why it deserves the attention.
- Stop treating news as an edge. If a crystal ball is worth about a dollar on fifty, a push notification is worth less.
- Notice when you have no read. The professionals' most distinctive move was not trading at all on unclear days.
- Make size follow confidence. Not the other way around, which is what wiped out one participant in six.
Prueba esta semana
- Think of the last market headline that made you want to act, and write down what the consensus expectation was. If you cannot, that is the point.
- Look back at your last ten decisions and mark each one confident or unsure at the time.
- Check whether your position sizes matched those labels.
- For the next two weeks, take no position on any day you cannot state your confidence as a number.
Preguntas frecuentes
If I knew tomorrow's news, could I make money in the market?
Far less than you would expect. In the Elm Wealth experiment participants given tomorrow's front page for 15 trading days finished with an average of about $51 on a $50 stake, only around half finished ahead, and one in six lost everything.
Why does knowing the news not predict the price?
Because the price already contains what people expected, and a headline does not tell you what that expectation was. A strong jobs number that falls short of forecasts reads as good news and trades as bad news.
Did AI do better in the experiment?
Slightly, on direction — about 60% correct against 51% for people — but no better on sizing. The models took too much risk when they were not confident, and some of them lost money too.
What did the professional traders do differently?
They varied their position size sharply with their confidence, including declining to trade at all on days when the read was unclear, and going large only on the clearest ones. Their edge in predicting direction was modest; their edge in sizing was decisive.
Does this mean markets are efficient and I cannot beat them?
The experiment does not say that, and the clip explicitly avoids it — its framing is that markets can be rational and still wrong. What it does show is that information alone is not an edge, and that how much you commit to a view matters at least as much as the view.