The failure your risk stack was built for
Picture the trader a prop firm's risk system is built to catch. A loss lands, the trader wants it back, and the next position is too big. The revenge trade breaches the daily loss limit or the trailing drawdown, the account is flagged, and the risk desk has a record of exactly what happened. It is loud, it is fast, and it is legible. The whole detection stack, max-loss caps, position-size limits, breach alerts, is tuned to catch it.
That failure is real, and it is the easy one to see. The study below points to a second, quieter one.
The failure it cannot see
Kandasamy and colleagues at Cambridge found the opposite of the intuition when cortisol stayed high. Working from stress-hormone levels measured in active traders, they ran a controlled study in which volunteers received hydrocortisone over eight days. The volunteers became more risk-averse: the certainty equivalent of a gamble, the sure amount they valued as highly as the gamble itself, fell 44%, from £25 to £14. A single acute dose produced no such effect, so this is the multi-day mechanism rather than the one that fires in the minutes after a loss.
Play that forward on a funded account. The study did not test traders, so this part is our extrapolation. Picture a trader under sustained stress who responds the way the volunteers did: they shrink rather than blow up. Winners get cut early to lock in relief, size drifts below the plan, and the setups that fit the strategy go untaken. The account breaches nothing. It bleeds down through a hundred small, cautious, defensible decisions until there is nothing left to trade.
No max-loss rule fires on that, because no limit was broken. A position-size cap only cares about positions that are too large. Breach detection only speaks the language of breaches. The quiet fade is a failure mode assembled entirely out of behaviour that stays inside every rule, which is exactly why the detection stack is silent while it happens.
| The loud failure | The quiet fade | |
|---|---|---|
| The behaviour | Oversizing after a loss | Undersizing, cutting winners, skipping setups |
| The clock it sits on | Seconds after a loss | Days of sustained stress (extrapolated from a study of volunteers) |
| Your limits | Breached | Never touched |
| On the risk desk | Flagged | Invisible |
How does cortisol affect a trader's risk-taking?
Held high for days, it made volunteers in a controlled study more cautious, not more reckless. Two stress responses fire after a loss, on two different clocks. Adrenaline arrives within seconds. Cortisol rises behind it through the HPA axis and peaks 21 to 40 minutes later. Kandasamy linked this slower hormone to risk aversion only when it stayed high for days, a separate effect from the minutes after a loss, and in volunteers rather than traders. We cover the full mechanism and its timeline in the cortisol glossary entry and in how long tilt lasts; here, only the direction matters. The loud failure is a trader doing too much. The quiet one, if the volunteer result carries over, is a trader doing too little.
"We already have an academy"
The reflex response to any behavioural problem is to teach harder. It does not reach this one. Even mild acute stress that a person cannot control impairs prefrontal functions such as working memory (Arnsten 2009). The review does not study trading, so the next step is ours: a trader who has read every risk lesson can still struggle to apply one straight after a loss. We have made the full version of this argument in the education gap. The narrower point for the risk-averse fade is sharper still: education cannot correct a failure nobody names. Risk courses warn traders against oversizing. A firm can check whether its own course says anything about the opposite: sizing down until the account fades. You cannot teach your way out of a pattern your curriculum does not recognise.
What your detection layer cannot see
For a Head of Risk, the consequence is uncomfortable and specific. The instruments on the desk, the dashboards, the drawdown monitors, the alerting thresholds, are all calibrated for the failure that announces itself. They are doing their job. They are watching the wrong door. A trader fading this way would show up as a run of unremarkable sessions: smaller, quieter, more hesitant, every one of them compliant. By the time the number on the account was low enough to notice, the decisions that got it there would be weeks gone and never flagged, because not one of them broke a rule. Each of those accounts cost something to acquire: our working range is $200 to $2,000 per trader, and a firm's own figure is in its acquisition data.
This is not an argument for tighter limits. Tighter limits catch the loud failure sooner; they do nothing for a failure built from compliance. It is an argument that a detection stack made only of thresholds is blind on one side, and that a firm should find out how many of its funded accounts end there.
Where the call fits
If the failure is behavioural and invisible to thresholds, the response has to read the behaviour itself, and it has to arrive while the trader can still be reached. The window we design for is the few minutes between a trigger and the next trade. It belongs to the fast clock, not to the multi-day cortisol effect above, and a review the next morning still arrives after the trades are placed. A dashboard the risk desk checks tomorrow is a record, not an intervention. At active tilt we put a voice call in those minutes rather than a notification: a message can be swiped away, and a ringing phone asks for an answer. Coaching, not financial advice.
Which conditions push the stress axis hardest, and how much of that a funded evaluation supplies by its own design, is a separate post. This one stays on the slow clock.
What the quiet fade looks like
The fade is invisible to limits, but it need not be invisible to a layer watching the behaviour. It would show up as a signature that never touches a threshold: session frequency dropping off, position sizes compressing below the trader's own baseline, winners cut faster than the plan allows, longer and longer gaps before re-entry. Read individually, each is defensible. Read together, they may be a trader fading out under load. The five churn metrics and the case for a pre-breach behavioural layer cover how a firm turns that signature into a signal it can act on. The detection problem this post describes is the reason that layer has to exist alongside the limits, not instead of them. The wider retention picture, acquisition cost included, sits in our churn statistics library and the trader retention playbook.
~75% of retail traders quit within 90 days. The loud failures are counted, flagged, and understood. The quiet ones, if a firm has them, are in its active cohort now: traders sizing down week on week, breaching nothing.
Sources and notes
- Cortisol and risk aversion (certainty equivalent fell 44%, £25 to £14, under sustained cortisol): Kandasamy et al., PNAS 2014, Cambridge (DOI 10.1073/pnas.1317908111). Peer-reviewed. Related: Coates and Herbert, PNAS 2008, on cortisol rising with market volatility among London traders (it did not measure risk preferences).
- The two stress clocks (adrenaline within seconds via the sympathetic-adrenal-medullary response; cortisol peaking around 21 to 40 minutes from onset via the HPA axis): standard stress physiology; the cortisol peak window is from Dickerson and Kemeny, Psychological Bulletin 2004, a peer-reviewed meta-analysis of 208 laboratory studies of volunteers, not traders. Summarised in our cortisol glossary entry and how long does tilt last.
- Prefrontal impairment under uncontrollable acute stress: Arnsten, Nature Reviews Neuroscience 2009. Peer-reviewed. Note: even mild acute stress that a person cannot control impairs prefrontal function; the region is not "physically blocked." The review does not cover risk-taking or financial decisions, and its mechanism work is largely in rats and monkeys. No reliable speed ratio between the amygdala and the prefrontal cortex exists in the literature, so the multipliers that circulate online should be treated with caution. LeDoux's animal work recorded amygdala responses within about 15 milliseconds, and the human magnitude is debated.
- The intervention window (the few minutes between the trigger and the next trade): Discentra operational construct, not a peer-reviewed term.
- Acquisition cost ($200 to $2,000 per trader): a Discentra working range from our own market research, not a published industry figure; it varies by channel and segment.
- "~75% of retail traders quit within 90 days": industry-cited working figure, not a regulator-published statistic. Regulators publish loss rates, a different measure.
