Innova Castle
Validation

A market thermometer, not an oracle.

Calibrated on the first half of the century. Checked on the second.

The Market Stress Index was fitted on the 12,386 sessions between 1927 and 1977, and then read the 12,386 that followed without being re-tuned once. It measures the state of the market, not what comes next.

Market Stress Index

S&P 500 · historical replay, not a live reading

74/ 100

Elevated

1929

1007550250
Normal
Elevated
High
Severe

Historical median 58 / 100 · states at their real share of the scale

The Stress Thermometer

MSI works as a precision thermometer: it reads the structural tension in the market from its price history, and from the S&P 500 only.

Most risk measures describe how much prices have been moving. This one describes the structural state of the series itself.

  • One number, four states

    The result is a single reading from 0 to 100: normal, elevated, high or severe.

  • White box, and reproducible

    The method is stated rather than hidden, and the same price history always produces the same reading. Nothing in the output is a black box.

  • Held fixed

    The same reading across the whole history. No retraining, no per-period tuning, no regime-switching.

  • A reading, not a signal

    It reports a state to interpret, not an alert to act on. It does not tell you what to buy, sell or hold.

Read the method note

A century, measured

Almost a hundred years of the S&P 500, read by the index. The decline phases marked here were located from price alone, with one mechanical rule: a retreat of at least 20% from the previous high. The index does not know they exist; what this chart shows is what it was reading while they ran.

A century measured

The monthly reading of the index, 1929-2026. Shaded: the century's twelve declines of 20% or more, located from price alone.

S&P 500 · 24,772 daily sessions
19291949196919892009

The six declines with a name of their own, out of the twelve marked above. The fall is the price. The reading is what the index showed while that fall was running, not a warning issued before it.

  • 1929 · Crash
    Price fall
    -86.2%
    Highest reading
    86 / 100
  • 1973 · Oil shock
    Price fall
    -48.2%
    Highest reading
    94 / 100
  • 1987 · Black Monday
    Price fall
    -33.5%
    Highest reading
    72 / 100
  • 2000 · Dot-com
    Price fall
    -49.1%
    Highest reading
    85 / 100
  • 2008 · Financial crisis
    Price fall
    -56.8%
    Highest reading
    94 / 100
  • 2020 · Covid
    Price fall
    -33.9%
    Highest reading
    56 / 100
NormalElevatedHighSevere
Dashed line: historical median, 58 / 100

Read the full finding for the price on its own, the two series together, and the state of every month of the century, with the method and the limits around them.

Measured across almost a century of the S&P 500

The reading was fitted on the first half of the century, and then checked on the half that followed without being re-tuned once. Each figure below says which stretch it is measured on. The decline phases were located from price alone, independently of the index.

Full century
24,772
S&P 500 sessions, 1927-2026
Out of sample
+22.5
points higher during decline phases than outside them
Full century
9 / 12
major declines with average tension above the historical median of 58
Full century
8 / 12
of those, with tension above the median for months beforehand

The figure that carries the argument is the first one: on the half of the history that took no part in any decision, the index reads 22.5 points higher during decline phases than outside them. It separates the two better there than on the stretch it was fitted on, which is the opposite of what an overfitted model does.

The counts cover the whole century, calibration included, and they are counts rather than a rate. Of the twelve major declines, nine ran with average tension above the historical median. In eight of the twelve, that tension had sat above the median for months beforehand, the same yardstick as the count before it. That describes the state that preceded them; it is not a claim that the index announced anything.

High tension is not a coming decline: episodes that corrected without one lasted 95 sessions at the median; those that ended in a decline, 273.

And once tension appears, how long does it last?

How long tension lasts

11 to 26 months

Half of the 17 episodes that reached 78/100 in a century lasted that long, counted from the moment the index enters Elevated until it returns to Normal. The longest ran 59 months.

01y2y3y4y5y
  • Half of the episodes
  • Median
  • Shortest and longest

The same count, split by the calibration cut

Calibration · n=7
12 to 28 months
Reserved, never seen · n=10
8 to 24 months

This is a range, not an estimate, and it is not a forecast. With 17 episodes in a century, no exact figure would be honest, and more daily data does not create more episodes. It says what past episodes lasted, never how long the current one will.

One instrument today. The method is general.

MSI reads the S&P 500 and nothing else, on purpose: a reading is worth what its calibration is worth, and every market has to earn its own over its own history. That is why the scope is one series, and why widening it is data work rather than new research. Nothing below is built, and none of it carries a date.

Today

One instrument. The S&P 500, calibrated on the first half of the century and checked on the half that followed.

Markets that could be read

Data work

Each one calibrated on its own history, never on a profile borrowed from equities. What every candidate needs is a long daily series, not a different method.

Other equity indices

The nearest step, and the one that matters most: the same asset class, read the same way. It is what turns "it works on this series" into "it works on this kind of market".

Rates and sovereign debt

Decades of public daily closes, and a regime that behaves nothing like equities. Its own calibration, or no reading at all.

Commodities

Oil and metals move on supply, weather and politics. The honest test of whether a reading built on an index travels at all.

Readings that could be built

Not built yet

The number already exists. Everything you would put around it is product, and none of it is standing today.

Markets side by side

The same 0 to 100 reading across instruments, so tension can be compared instead of read one instrument at a time.

Continuous watch

Today a reading is taken. Following it and saying when the state changes is the recurring use, and it is the part that does not exist.

One number, integrated

The output is a number and a state. Putting it next to your existing risk measures is integration, not modelling.

Source independence

Every check repeated on a second provider, so no result rests on a single feed. The tooling for it is written; the run is not done.

No dates, deliberately. A roadmap whose dates slip costs more than having no roadmap at all.

What it does not do

Stated here, by us, rather than in small print at the bottom. A reader who finds an undeclared limit discounts everything else on the page, and is right to.

01

It does not predict prices or dates

There is no forecast in the output and no probability attached to what comes next. It reports the state of the market now, and that is the whole claim.

02

It does not announce the start of a decline

We measured that too, and what it catches of the beginnings is not distinguishable from chance. The reading is about the state of the market, not about the turn.

03

It is late to sudden crashes

The reading accumulates instead of resetting each session, which is what lets it show tension that has been building for months. The cost is the fast fall: in 2020 it read 56 during the decline, and 77 in the months after it had ended.

04

High tension is not a decline on its way

While they are running, an episode that will end in a decline and one the market will correct on its own look alike. What separates them is how long they last, and that is only known afterwards.

05

It reads the S&P 500, and nothing else

The calibration comes from that one series and its own history. Pointed at another market it would treat as unusual what is ordinary there, so it is not pointed at another market.

An episode of tension that the market corrected is not a failed reading. There was tension, and it passed.

See what the research becomes.

Read what we have tested and what it does not show, or start a conversation about where it could be useful.