In 2025, I ran a prediction game for finance professionals. This year, to keep pace with current trends, I decided instead to pose ten questions about markets in 2026 to several AI large language models (LLMs). Will they do any better than humans, and will they agree with each other?

I used three AI LLMs for this task: ChatGPT (5.2), Gemini (3), and Claude (Sonnet 4.5). Here are the questions and responses:

1) Where will the S&P 500 finish 2026?
ChatGPT: 7,700
Gemini: 7,555
Claude: 6,800

2) Where will the FTSE 100 finish 2026?
ChatGPT: 10,600
Gemini: 10,750
Claude: 9,200

3) What will the ten-year Treasury yield be at the end of 2026?
ChatGPT: 4.5%
Gemini: 3.75%
Claude: 4.50%

4) Where will the GBP/USD spot rate be at the end of 2026?
ChatGPT: 1.28
Gemini: 1.38
Claude: 1.28

5) What will the dollar price of Brent crude oil be at the end of 2026?
ChatGPT: $57
Gemini: $55
Claude: $66

6) Will there be a 20% decline in the S&P 500 in 2026?
ChatGPT: No
Gemini: No
Claude: No

7) What will the dollar price of Bitcoin be at the end of 2026?
ChatGPT: $130,000
Gemini: $135,000
Claude: $78,000

8) Will the Russell 2000 outperform the S&P 500 in 2026?
ChatGPT: No
Gemini: No
Claude: No

9) What will the Federal Funds Target Rate (upper bound) be at the end of 2026?
ChatGPT: 3.25%
Gemini: 3.25%
Claude: 3.50%

10) What will the dollar price of gold be at the end of 2026?
ChatGPT: $4,500
Gemini: $4,900
Claude: $3,100

---

It was reassuring that when I initially asked the AI models to make these predictions, they were reluctant and would only offer heavily caveated ranges – suggesting they may be better calibrated than humans. I had to persuade them to produce point forecasts!

Of course, the models were right to be reticent. No LLM will be capable of making accurate and specific predictions about a system as complex as financial markets. Still, there were some other aspects I wanted to explore.

Were the predictions consistent across models?

Absolutely not. While there was some consistency between ChatGPT and Gemini, Claude made several very bold calls (gold at $3,100, for example). Interestingly, Claude’s rationale was often the most “convincing.” It behaved like a particular type of market strategist – frequently wrong, but articulate and provocative – and therefore attracting the most attention.

Were the predictions internally consistent?

This is an important and difficult question to answer. The models could certainly sound internally consistent when asked to justify their views, but some combinations looked odd. For example, Claude was the most bullish on oil while also being the most bearish on UK equities – a scenario that could happen, but feels like an unlikely pairing (given the resource heavy nature of the UK market).

Are they better than human predictions?

I was asking the models to predict outcomes that, in my view, are impossible for either humans or LLMs to forecast with any real accuracy. Can they do this as well as a human? Almost certainly. Does that make it useful? Outside of marketing copy, probably not.

There are definitely more important things to spend tokens on!

—

* Thank you to my colleague Duncan for suggesting this post.

—

My first book has been published. The Intelligent Fund Investor explores the beliefs and behaviours that lead investors astray, and shows how we can make better decisions. You can get a copy here (UK) or here (US).

All opinions are my own, not that of my employer or anybody else. I am often wrong, and my future self will disagree with my present self at some point. Not investment advice.

Having already written a piece on the similarities between how contestants in the TV game show The Traitors behave and our futile attempts to time financial markets, I thought I was finished with the subject. After indulging in the fourth UK series of the show, however, I realised there are so many excellent examples of human folly and perplexing decision-making that it felt worthy of a sequel.

For those uninitiated, there is a recap of how the game works in my previous post on this subject.

Here are some traitorous observations on investor behaviour:

We like being part of the herd. In the first banishment round of the latest UK series, 16 of the 21 contestants voted to remove the same person from the game. This broad consensus was reached despite there being no meaningful evidence to speak of. Of course, despite the agreement, they were entirely wrong.

We don’t like outsiders. After the first failed banishment, one of the 16 people who voted to remove a Faithful suggested turning their attention to a contestant who had voted differently, on the basis that this made them a potential Traitor. This was despite the majority having been incorrect in their prior accusation. It does not matter whether a group is right or wrong; it is safer to be in it.

We have a narrow circle of competence. A consistent feature of the game is that people in certain professions, such as lawyers and police detectives, believe they will be good at it and are reassuringly terrible. Part of me worries that people whose jobs involve making decisions under uncertainty do not seem to spend much time thinking about how to make good decisions. Another part simply concludes that our circle of competence is far narrower than we like to admit.

We are wildly overconfident. Along similar lines, a highlight of the most recent series is a retired police detective who plays the game with great confidence while aggressively and incorrectly accusing the wrong people of being Traitors. They also decided to keep their previous career a secret from the group, until revealing it to one other contestant who was, with wonderful inevitability, a Traitor. We are not as good as we think we are.

We are in thrall to expertise. Even though participants with so-called decision-making backgrounds are wrong at least as often as everyone else, other contestants remain in awe of their supposed expertise, no matter how often it proves unreliable. This is reminiscent of investors patiently waiting for the next insight from a market strategist who is right about as often as a coin flip.

We are attracted to new things. In the game, a banishment occurs every episode or day, and there is a strange behaviour whereby one day an individual may attract a great deal of suspicion and votes, but if they survive, they are barely mentioned the following day. We are drawn to shiny new things even when the evidence has not changed at all, much like investors latching on to the topic of the month and forgetting what they were previously obsessed with.

We are terribly calibrated. One of the most enjoyable and bewildering facets of the game is how often contestants say things like, “I am 100 percent sure X is a Faithful.” This is typically based on no evidence whatsoever and would be absurd even in situations where we have good reason to be confident. Here, participants have almost nothing to go on. We are very poor judges of how confident we should be about uncertain outcomes.

We need to survive: Although the game is focused on identifying Traitors in the group this isn’t really very important until we get near to its end because banished Traitors are replaced anyway. Participants shouldn’t care that much whether they banish a Faithful or a Traitor – they just need to make sure it is not them. Survival is the most vital and underappreciated thing in both the game and investing.

We cannot read people. Whether it is identifying a Traitor or judging a fund manager, we are awful at reading people yet remain convinced we are good at assessing both honesty and intentions. In a previous season, there was a contestant who was a professional magician whose job involved reading people. He was predictably hopeless at the game.

We ignore the most important things. Given how little useful evidence there is for identifying Traitors, particularly early in the game, the most sensible approach outside of choosing someone at random would be to ask: if I were a producer of this show, who would I select as a Traitor? Either contestants do not think this way, or, if they do, it is not shown on television because it would spoil the magic. My guess is the former.

We use salience as evidence. Judgements about who is a Traitor, especially early on, seem to be driven by what or who is salient, noticeable, or different. We mistake what captures our attention for what is meaningful. The current UK series has seen players frequently banished after making themselves the centre of attention – which is baffling behaviour.

While The Traitors is a light-hearted and convoluted game, it does a great job of revealing some of our most ingrained and peculiar behaviours. When we watch it and feel frustrated by the decisions being made, it is worth remembering that the contestants are simply being human. We do many of the same things in life and certainly when investing.

—-

My first book has been published. The Intelligent Fund Investor explores the beliefs and behaviours that lead investors astray, and shows how we can make better decisions. You can get a copy here (UK) or here (US).

All opinions are my own, not that of my employer or anybody else. I am often wrong, and my future self will disagree with my present self at some point. Not investment advice.

From a behavioural perspective, our ability to meet our long-term investment objectives is really about one thing: managing the conflict between our desire to save for the future and our craving to make ourselves feel better in the present. This has always been an incredibly exacting challenge, but it is being exacerbated by an industry that seems set on innovations aimed at exploiting our weakness for immediate satisfaction.

Back in 1997, economist David Laibson published a paper titled “Golden Eggs and Hyperbolic Discounting”. In it he contends that our long-run savings and investment goals are frequently compromised by the behaviour of our present-biased selves.

There are two key elements to this:

• We fail to save sufficiently for the future because we choose to consume more today.
  
  
• We fail to keep our savings invested for the long run because of our response to short-term volatility and temporary market losses.

Laibson argues that investors are therefore blighted by having “too much” liquidity, which gives us ample opportunity to prioritise how we feel right now over our future needs. If we know that humans have this strong tendency, then illiquidity becomes incredibly valuable – because it protects us from ourselves.

While Laibson refers to the value of illiquidity, he is not advocating a wholesale switch into private markets. What he means is the use of commitment devices: steps we can take to prevent panic selling or otherwise stop us reacting in-the-moment at the expense of our future needs.

Commitments come in many forms – the most obvious being investment vehicles holding genuinely illiquid assets with lock-up periods. Not everything has to be so punitive, however. Soft frictions, such as auto-enrolment or the Save More Tomorrow scheme in the US, can also be extremely effective. Anything that reduces impulsive behaviour can deliver value many years from now.

While tools that encourage commitment can have a dramatic impact on behaviour, there is a problem: we don’t like being asked to commit. There are several reasons for this:

• It feels like a restriction of choice and agency. Adding friction to slow our decision making or reducing our ability to act can feel like a loss of freedom. Imagine if an investment platform introduced a 24-hour “cooling-off” period for every trade. This would likely be effective at limiting hot-state behaviour, but few people would choose to use it.
  
  
• We don’t think we need them. Commitment devices are designed to counter problems caused by our present-biased selves. To value them, we must first accept that we suffer from this behavioural failing. Of course, we all do – but we are far more likely to see it in others than in ourselves.
  
  
• Commitment devices involve trade-offs. Anything that reduces liquidity or limits action for the benefit of our future self comes with a present-day cost. This may not just be frustration at our inability to act; it could also mean being unable to access funds when they are genuinely needed.

Laibson made the argument about the dangers of “too much liquidity” nearly 30 years ago, but his case feels more relevant – and more urgent – than ever. Humans remain inescapably present-biased, yet the investment landscape has evolved in ways that exacerbate (and perhaps deliberately exploit) this bias, to the detriment of our long-term savings goals. There are a host of contributing factors, but these are the most culpable:

• ‘24 hour’ trading. Financial markets are increasingly ‘always open’, providing instant liquidity at all times. There is no restriction on our ability to react in the moment. This can be framed as technology-driven liberation or, more realistically, as a behavioural disaster.
  
  
• Constant portfolio access. Not only can we trade at any time, but we can live the minute-by-minute fluctuations in our portfolio values. Nobody would argue that investors should not have access to this information, but few have seriously considered the behavioural cost.
  
  
• More news and emotional stimulus. We are bombarded by financial news and opinion, a phenomenon greatly amplified by social media. There is more news, more noise, and more negativity. Humans make poor decisions when emotional stimulus is high and friction is low – precisely the environment we now inhabit.
  
  
• The financialisation of everything. Over time, more aspects of life are becoming financialised products to buy and sell – stocks, currencies, cryptocurrencies, sports, even political events. Everything has become something to trade. The line between gambling and investing is increasingly blurred, and may soon disappear entirely.
  

All of these factors increase the likelihood that we sacrifice long-term goals for short-term fulfilment. If Laibson was worried in 1997, he should be petrified today.

We can think of the golden egg as the future consumption funded by our savings, and the goose that lays it as a sensibly diversified investment portfolio. Financial and technological innovation is making it far easier to kill the goose.

—

Laibson, D. (1997). Golden eggs and hyperbolic discounting. The Quarterly Journal of Economics, 112(2), 443-478.

(Laibson frames the ‘golden egg’ in a slightly different way in this paper, but the overarching point is the same).

—

My first book has been published. The Intelligent Fund Investor explores the beliefs and behaviours that lead investors astray, and shows how we can make better decisions. You can get a copy here (UK) or here (US).

All opinions are my own, not that of my employer or anybody else. I am often wrong, and my future self will disagree with my present self at some point. Not investment advice.

Back in December of 2024, I ran a survey which was completed by 276 finance professionals. It asked them 10 questions on what would happen in markets in 2025.

Let’s see how they got on:

Question 1 – What will be the % total return of the S&P 500 in 2025?

Prediction: 6.5% (mean)

Outcome: 17.9%

The 21 participants that predicted a loss of 10% or greater for the US market must have been feeling pretty good in April, but by the close of the year the US market had recovered to produce a total return just short of 18%. Of the 268 responses, only 8% predicted performance of 18% or more.

The average return expectation was for a 6.5% total return from US equities in 2025. While this seems a reasonable number such ‘normal’ performance is quite rare in any given calendar year – equity returns are high because they are very lumpy!  

Question 2 – Will the Russell 2000 outperform the S&P 500 in 2025?

Prediction: Yes (60%)

Outcome: No

The respondents were moderately optimistic about a recovery in the fortunes of smaller US companies, perhaps through a combination of depressed (relative) valuations and the initial hopes for the economic impact of Trump’s second term. As it turned out, the performance of small caps in the US was decent in absolute terms, but they still trailed the mega caps by c.6%. Maybe 2026 will be the year.

Question 3 – Which equity market will produce the highest total return in 2025 (USD terms)?

Prediction: US equities (39%) 

Outcome: Chinese equities (20% of respondents)

There was a strong expectation for a continuation of US equity exceptionalism in 2025, which turned out to be entirely wrong. While near 40% of participants backed the US to be the leading market in 2025, its returns trailed all other options except for India. This was a win for the contrarians, as the much-maligned Chinese market came top of the pile – making 20% of the respondents right.  

Question 4: What is the probability of the US economy entering a recession in 2025?

Prediction: 25% chance of a recession (mean)

Outcome: No recession.

Without the ability to observe parallel universes it is quite difficult to judge the quality of responses here. All that can be said is that most participants were generally sanguine about the prospects for a US recession in 2025 and the outcomes were consistent with this. It was looking a little dicey in April and May, however, with prediction markets pricing in a 60% chance in 2025.

Oh, and the 23 forecasts of a 0 or 100% recession probability were wrong from the start.  

Question 5: Will Nvidia outperform the S&P 500 in 2025?

Prediction: No (62%)

Outcome: Yes. Nvidia returned 39%, against 18% for the S&P 500.

Much like the small-cap view, participants were expecting a broadening out of US equity markets in 2025. It seemed reasonable to believe a company of that size, trading on that valuation with that stratospheric performance history couldn’t outperform for another year, but markets have a history of being wholly unreasonable.

Question 6: What will be the GBPUSD spot exchange rate at the close of 2025?

Prediction: 1.26 (mean)

Outcome: 1.34

Although on average there was an expectation for a slightly stronger sterling / weaker dollar across 2025, the most common prediction was for sterling weakness (1.20). This probably reflects lukewarm sentiment for the UK and (thus far) unfounded optimism about Trump’s impact on the dollar. Only 16% of respondents went for a rate of 1.34 or above.

Here is a resolution for 2026 and for every year for all investors – I will not make currency forecasts.  

  
Question 7: What will the USD price of Bitcoin be at the close of 2025?

Prediction: $108,353 (mean), $95,000 (median)

Outcome: $87,647

Participants were a little too optimistic about the fortunes of this ‘belief asset’ in 2025, however, the forecasts did range from $1m to 0 so it was a little noisy. It is hard to say whether any given prediction is either serious or a joke given the nature of the asset.

Question 8: Will the S&P 500 suffer a peak to trough decline of more than 20% in 2025?

Prediction: No (60%)

Outcome: No (just)

Although US equity returns were strong across 2025, there is also a Wikipedia page for the ‘2025 Stock Market Crash’, which although seeming a little dramatic is a reminder of the negativity that surrounded ‘Liberation Day’ and the policy chaos that surrounded it. If we classify a bear market as a 20% decline, the S&P 500 just missed out in 2025 falling 18.9% between 19 February and 8 April.

Question 9: What will be the annual rate of US inflation (CPI) in 2025?

Prediction: 2.9%

Outcome: 2.7% (November print)

This is a tricky one for a couple of reasons: 1) We don’t yet have the December figure, 2) There was much scepticism from economists about the November print because of the impact of the US Government shutdown. It is probably fair to say that the average inflation expectation for 2025 was in the right ballpark.

Question 10: What will the US Ten Year Treasury yield be at the close of 2025?

Prediction: 4.20% (mean)

Outcome: 4.17%

Who said that forecasting bond yields was difficult? The crowd was wise on this topic with the average prediction being pretty much bang on the yield of ten-year Treasuries at the close of 2025. The dispersion of responses to this question was, however, wide. 20% of participants forecast the yield finishing the year above 5% and another 20% below 3.5%. Lots of wrongs can make a right.

—

Making good predictions about financial market performance over short periods such as one year is incredibly difficult. In essence, we are trying to forecast how other market participants will react to events that we know will happen but don’t know the outcome and events that we don’t know will happen and (by definition) can’t know the outcome. We also need to work out how these events or developments might interact with each other. It would be hard to design a more complex prediction problem. *

(There is a reason why annual investment outlooks don’t spend much time reviewing what they said a year ago.)

Given this it is somewhat baffling that the entire asset management industry can sometimes feel like a giant forecasting game – one where people are constantly making predictions while being incurably afflicted with amnesia about the quality and content of their previous judgments. I was totally wrong last time, but you have to listen to me this time around.

The good news is that for most investors this is all irrelevant noise. Our long-term investment fortunes will not be defined by our ability to foresee what will happen in markets over the next quarter or year. They may, however, be defined by our ability to accept this and act accordingly.

—

* It is not even about getting a single prediction right, you have to do it again and again.

–

My first book has been published. The Intelligent Fund Investor explores the beliefs and behaviours that lead investors astray, and shows how we can make better decisions. You can get a copy here (UK) or here (US).

All opinions are my own, not that of my employer or anybody else. I am often wrong, and my future self will disagree with my present self at some point. Not investment advice.