What the research actually says about trader intuition

Some traders describe a feeling. They look at a chart and something arrives before the reasoning does: this is not holding, or that breakout is wrong, or this pullback is not like the last three.

It gets called a sixth sense. There is a plainer explanation, and a body of research behind it.

Intuition, in the research, means recognition#

The cleanest statement of it comes from Herbert Simon, who spent years studying how chess masters see a board. Writing in Psychological Science in 1992, he put it this way:

The situation has provided a cue; this cue has given the expert access to information stored in memory, and the information provides the answer. Intuition is nothing more and nothing less than recognition.

That is deflationary on purpose. Simon was impatient with the mystique. On his account an expert's fast judgement is not a shortcut around thinking — it is a memory lookup that happens to be quick, built out of a very large number of previously encountered situations.

The chess work is where this came from. Strong players do not win by calculating every branch. They see a position and a small number of sensible candidate moves are simply there, because the structure in front of them resembles structures they have met before. Experience does not only change what you know. It changes what you see.

Someone tested whether markets can be read this way#

In 2010, Antoine Bruguier, Steven Quartz and Peter Bossaerts published Exploring the Nature of "Trader Intuition" in The Journal of Finance. They built experimental markets in which some participants held private information, then showed the order flow to subjects who had none, paused it repeatedly, and asked which way the next transaction price would go.

The subjects were good at it. Their forecasts were correct in approximately two-thirds of cases on average. For comparison, the authors report that randomly choosing between up, down and unchanged would have been right 33% of the time, and a stronger naive rule — always predict the previous outcome — would have been right 56%.

The spread between people was wide. The worst subject managed 46%, slightly worse than the naive rule. The best managed 78%.

So something was being read out of the order flow. The interesting part is what.

The authors found that performance correlated with scores on two Theory of Mind tests — the capacity to infer intent from behaviour — and their hypothesis was that it would be uncorrelated with mathematical and logical reasoning. Their reading is that subjects were not only doing arithmetic on prices. They were picking up on other people acting: hesitation, urgency, a change in rhythm.

That is a useful way to think about a chart. What you are looking at is not a sequence of numbers. It is the residue of decisions.

The same paper is explicit about the ceiling#

This is the part that usually gets left out when the study is cited, and it is the most important sentence in it. The authors write:

Superior forecasting performance does not necessarily translate into superior investments.

They are direct about scope: reading a market and predicting a price are two of the three steps toward successful investing, and they explicitly leave the third — turning a prediction into a position, the trading itself — to future research.

So the study supports a narrow claim: in some market environments, people can extract real information from price and order flow, and that ability looks like fast pattern recognition. It does not support the broad one, that practising chart reading produces reliable forecasts of real markets, still less better outcomes for anyone trading them.

When is an intuition worth trusting?#

Daniel Kahneman and Gary Klein spent years disagreeing about intuition in public — Klein studying firefighters and other experts who make good fast decisions under pressure, Kahneman documenting the ways human judgement goes wrong. In 2009 they wrote a joint paper about where the boundary sits.

Their answer needs two things to be true. The environment has to supply cues that are genuinely informative about the situation. And the person has to have had a real opportunity to learn those cues.

Both, not either. Where an environment is close to unpredictable, long exposure does not reliably produce skill. It can produce something that feels identical from the inside: more confident guessing.

Kahneman and Klein are notably cautious about financial markets specifically, and long-horizon stock forecasting is one of their standard examples of an environment where the cues are weak enough that experience does not convert into expertise.

Which leaves a distinction worth holding onto: getting better at reading a market is not the same as the market becoming predictable. You can genuinely improve at noticing that volatility just expanded, or that this leg is shaped differently from the last one, and none of that entitles the next move to agree with you.

Screen time is not practice#

There is a second trap, and it is the reason "I have watched charts for years" is a weaker credential than it sounds.

Ericsson, Krampe and Tesch-Römer's 1993 work on expert performance separates deliberate practice — effortful activity designed to improve a specific thing, with feedback — from simply accumulating hours at an activity. Repetition alone is not the mechanism.

Watching a market move and then telling yourself I knew that would happen is repetition. It also has a specific failure mode: you cannot tell, afterwards, whether you actually knew. The story assembles itself around the outcome. Do it for long enough and the skill you are training is post-hoc explanation.

The loop that avoids this is boring and strict:

See the information → commit to a call → lock it → see the outcome.

The lock is the whole point. Once the call is recorded, the outcome cannot quietly rewrite it.

What this means for how SwipeTA is built#

We cannot claim that using SwipeTA will make anyone better at real markets. No study shows that, including the ones above, and the honest position is that the transfer question is open.

What we can do is take Kahneman and Klein's two conditions seriously as design constraints, because they describe exactly what a practice environment has to provide.

Cues that are genuinely informative. Every question is a real US equity intraday session, replayed from 5-minute bars, frozen at a decision point, showing only what was visible at the time. No hidden future data reaches the screen. The move being asked about is filtered to between one and three times ATR — below that band the outcome is mostly noise and there is nothing to read; above it, the move is a shock rather than something the preceding chart implied. That filter exists to keep the cue-to-noise ratio high enough that reading is possible at all.

A real opportunity to learn them. The horizon is fixed at 15 minutes and set when the question is written, never chosen afterwards to suit the answer. The bank is balanced 50/50 between up and down, so always guessing one direction earns nothing. And the answer arrives seconds after the call, which is the part real markets are worst at — a position you hold for a week teaches you almost nothing, because by the time you find out, you no longer remember precisely what you saw.

The numbers behind all of that are on our methodology page.

The thing actually worth training#

Kahneman and Klein make one more point that is easy to skip: subjective confidence is not reliable evidence that a judgement is correct. I am sure and I am good at this are different sentences, and they come apart most often in exactly the environments where feedback is slow.

Which suggests that the goal is not a louder first reaction. It is a first reaction you have enough data about to know when to listen to it — and, more usefully, when it is only telling you what you want to hear.

That part is measurable. It needs a record of what you called, and what happened next.