The fear and greed tile on the Macro Risk Scorecard currently reads 52, labelled NEUTRAL. That number is the most useful thing on the page for making this argument, because it is a reading that tells you nothing at all, and understanding why it tells you nothing is the whole skill.
Sentiment is not a continuous signal that gets more informative as it moves. It is a signal with two live zones at the ends and a long dead stretch in the middle. Most of the time it sits in the dead stretch. People get into trouble by treating the middle as if it were a weak version of an extreme, and then acting on a weak version of nothing.
The mechanism that makes extremes different from the middle
A contrarian reading of sentiment works when it works for one reason, and it is worth stating the reason so you can tell when it does not apply.
Extreme sentiment is a statement about positioning, not about mood. When almost everybody is fearful, the people who intended to sell have largely sold, which means the pool of remaining sellers is small and the pool of potential buyers is everybody who is out. The asymmetry sits in who is left to transact, not in how anyone feels. That is why the ends of the range behave differently from the middle in a way that is structural rather than mystical.
In the middle of the range, positioning is mixed. There are plenty of people left to sell and plenty left to buy. Nothing about the distribution of who holds what is unusual, so there is no asymmetry to lean on. A reading of 52 is a description of an ordinary market, and ordinary markets do not owe you a direction.
Notice the second consequence. Because the mechanism runs through positioning, the same headline number can mean different things at different times. Sentiment gauges are built from components with their own histories, and a reading that was extreme when a market was smaller and more retail-dominated may be ordinary now. That is the argument for measuring your own bands rather than inheriting a rule of thumb.
Why I am not handing you the thresholds
The natural way to write this article would be to state that below some number you buy and above some other number you sell. I am not going to do that, and the reason is not caution for its own sake.
The capture in front of me shows the Sentiment tab in the row of the Macro Risk Scorecard, alongside Dashboard, Regime, Credit and Liquidity, Policy, Outlook, Strategy and Trade Analysis. What is confirmed on the page is the header and the tile row, including that fear and greed reading of 52 tagged NEUTRAL. The tab's own detail panels sit below the visible area, and I am not going to describe panels I have not seen or attach percentile cut-offs to them that nothing establishes.
Borrowed thresholds also carry a specific hazard. A number quoted widely enough becomes a level people trade in anticipation of, which changes how the market behaves around it. Your own bands, computed from the history you actually have, are both more honest and less crowded.
Finding your own bands in an afternoon
This is arithmetic, not statistics, and a spreadsheet handles it.
- Pull as long a history of the sentiment series as you can get. Length matters more than precision here, because extremes are rare by definition and a short history will contain almost none of them.
- Convert every reading to a percentile of the history that preceded it, not of the whole series. Using the full series lets information from the future leak into the past, and that single mistake makes almost every homemade sentiment study look better than it is.
- Bucket the percentiles into fives, so the bottom five per cent, the next five, and so on up to the top.
- For each bucket, compute the average forward return of whatever you actually hold, over a horizon you would actually wait, such as one month and three months. Record how many observations sit in each bucket. This number is the one people skip and it is the one that decides whether you should believe the result.
- Look at the shape. If the pattern is a clean gradient across all twenty buckets, be suspicious, because that is what curve fitting looks like. If it is flat across the middle and only turns in the outer one or two buckets, that is consistent with the positioning mechanism and it is what you would expect.
When you get to the end of this, look hard at the observation counts in your outer buckets. If the top five per cent contains twelve observations, you have twelve data points supporting your rule, and twelve data points support very little. This is not a reason to abandon the exercise. It is a reason to size the resulting trade like something you are twelve observations confident about.
What to actually do with a reading of 52
Nothing. That is the answer, and it deserves more than one word because doing nothing is harder than it sounds.
A middling sentiment reading is an invitation to overtrade, because the gauge is right there, it updates, and updating things feel like they are asking for a response. The specific cost for a retail account is that each response carries a spread, possibly a commission, and in a taxable account a realised gain or loss you chose to create. Ten unnecessary decisions a year at a few tenths of a per cent each is a real drag, and it is entirely self-inflicted.
The concrete practice that helps is to write down, in advance, the two band edges you found in your own analysis, and then to only look at the gauge when it is near them. Set an alert at your levels and otherwise leave the tab closed. This converts a continuous temptation into two discrete events a year, which is roughly how often the underlying mechanism is actually present.
The failure modes at the extremes themselves
Even when you get to a genuine extreme, two things go wrong often enough to plan for.
The first is that an extreme is a condition, not a trigger. Sentiment can sit at an extreme reading while the price keeps going the same way, because the positioning asymmetry says something about the eventual distribution of outcomes and nothing about the timing. Anyone who tells you how long an extreme lasts before it resolves is quoting a range as though it were a constant, and the honest version of that number is one you would have to measure yourself.
The second is that extremes cluster in exactly the periods when your other holdings are also under stress. The reading that says the pool of sellers is exhausted tends to arrive when your account is down and your appetite for adding risk is at its lowest. That is not a coincidence, it is the same event viewed from inside your portfolio, and it is the reason the decision has to be written down in advance. A rule you wrote in a calm week is the only version of you that will still be functional in the week the gauge finally reaches your band.