Forex Sentiment
Live 2026-09-04 04:36:09 Jerusalem

Findings

What actually moves these pairs, measured

← All findings

The contrarian trade failed overall — does it work in specific conditions?

Fading the crowd does not work in any condition we could find

We split the same trades by market fear, by carry, and by what the big funds were doing. Fading the crowd lost money in every single group.

In numbers

The worst case was panics. When volatility was high — exactly when you would expect the crowd to be capitulating and wrong — fading it lost 0.13 % per trade across 125 trades.

How to use this

  • Do not go looking for the condition that makes this work. We looked, with the obvious candidates, and there was not one.
  • The one group that came closest to breaking even was fading the crowd while earning the interest differential. Even that was slightly negative.
  • When retail and the funds are on opposite sides, the funds were not reliably the ones to follow. Trading against the crowd in that situation lost money too.
  • Use positioning to understand where risk is concentrated and where stops sit. That is what it is good for. It is not a direction signal in any regime we tested.

The detail

A failed strategy often hides a working one inside it. The obvious hypothesis after the unconditional test failed was that the crowd is only wrong in particular circumstances — during a panic, say, when positioning becomes capitulation.

So the same 474 entries were split three ways using data this site already holds: by the VIX reading at entry, by whether fading the crowd meant earning or paying the interest differential, and by whether the CFTC funds were positioned with the crowd or against it.

Every bucket lost money. Not one condition produced a positive result, let alone a significant one.

The volatility split is the most informative, because it points the opposite way to the hypothesis. Fading the crowd performed worst when volatility was highest. In a stressed market the crowd's positioning was more right, not less — which makes sense if panics are when trends run hardest, and the crowd is positioned with the trend rather than against it.

The institutional split is worth noting too, because it is the one this site surfaces most prominently. When retail and the funds disagreed, betting with the funds against the crowd still lost. Disagreement between the two groups is interesting information about the market. It is not a trade.

Every condition tested

the numbers behind it RECOMPUTED 03 Sep 2026
ConditionTradesWin rateMean returnt-stat
Volatility — normal 335 49.3% -0.015 -0.700
Volatility — calm 14 28.6% -0.125 -3.000
Volatility — stressed 125 39.2% -0.129 -3.020
Carry — we earn carry 293 46.8% -0.020 -0.810
Carry — we pay carry 119 42.0% -0.091 -2.560
Carry — roughly neutral 62 50.0% -0.099 -1.840
Institutions — funds with crowd 31 45.2% -0.056 -0.880
Institutions — funds against crowd 156 45.5% -0.088 -2.310
What this does not show. Splitting 474 trades three ways produces small groups, and the calm bucket has only 14 trades and should be ignored. Seven months of sentiment history covers one regime. This shows no edge was found, which is not the same as proving none exists.

Other findings