Market memory
Argentina · since 1999 · 24 crises · data as of Oct 28, 2026
Which days and which crises in Argentina's history today looks like: every trading day since 1999 is a fingerprint of 13 variables in historical percentile, and similarity is measured between fingerprints. A cell map shows where today lands in 27 years of territory.
How to read it: the score is a similarity, not a probability. A random day scores around the floor shown in the hero; above the resonance threshold, today's mix of country risk, FX pressure, inflation and activity looks like that crisis's week. It says nothing about what happens next.
Today looks like
72/100
YPF 2012 · Apr 16, 2012
- +3 over 30 sessions
- Deval. 2014 · 65
- IMF 2018 · 62
- press: no reading
- 13 of 13 variables
A random day scores 34 · resonance starts at 60
A trading day is one market session. Every session since 1999 is a 13-variable fingerprint in historical percentiles. Similarity with the past, not a forecast.
Where today lands, since 1999
Tap a cell, a ring or TODAY to see what is there.
How to read the map
- Cell: the sessions that landed there. Size is how many; color is the p90 similarity with today among its comparable sessions.
- Ring: one of the 24 crises, colored by family.
- TODAY, with the trail of the last 90 sessions and marks at 30 and 60 sessions.
Color by similarity with today
max today 72
Families
- Debt
- Currency
- Capital controls
- Political
- Pandemic
Distance on the map correlates 0.72 with the real distance across the 13 variables; two neighboring cells may not look alike: the light comes from the 13 variables, not from position.
The press axis is not drawn on the map.
The 24 crises
similarity with today · random 34 · press (0-50)
- Dec 3, 2001
Capital controls
39no reading - Dec 23, 2001
Debt
26no reading - Jan 6, 2002
Currency
27no reading - Mar 3, 2005
Debt
30no reading - Mar 11, 2008
Political
47no reading Nationalization of the pension funds
Oct 21, 2008Political
51no readingStart of the currency controls
Oct 31, 2011Capital controls
46no reading- Apr 16, 2012
Political
72no reading - Jan 23, 2014
Currency
65no reading Technical default over the holdouts
Jul 30, 2014Debt
37no readingEnd of the currency controls 2015
Dec 17, 2015Capital controls
49no reading- Apr 22, 2016
Debt
27no reading Start of the 2018 currency run
Apr 25, 2018Currency
57no reading- Jun 7, 2018
Debt
62no reading Monday after the 2019 primaries
Aug 12, 2019Political
58no reading- Aug 28, 2019
Debt
49no reading - Sep 1, 2019
Capital controls
48no reading - Mar 20, 2020
Pandemic
61no reading - Sep 4, 2020
Debt
41no reading - Jul 2, 2022
Political
49no reading - Aug 14, 2023
Currency
42no reading - Dec 13, 2023
Currency
23no reading Capital controls
57no reading- Sep 19, 2025
Currency
53no reading
How it is read
The fingerprint
Every session since 1999 is a vector of 13 variables, each expressed as a percentile of its own history (0 = the lowest value on record, 1 = the highest). Missing data stays empty; nothing is imputed. The 7 market variables weigh 1.0; the 6 regime variables, 0.5.
| Variable | Abbr. | Unit | Weight |
|---|---|---|---|
| Country risk | RP | pb | 1.0 |
| Country risk, 30-day change | RP30 | % | 1.0 |
| FX gap (blue vs official) | BR | % | 1.0 |
| Merval in USD, 30-day change | MV30 | % | 1.0 |
| Real interest rate | TR | pp | 1.0 |
| Reserves, 30-day change | RS30 | % | 1.0 |
| Official dollar, 30-day change | OF30 | % | 1.0 |
| Monthly inflation | INF | % | 0.5 |
| Real effective exchange rate | TCR | idx | 0.5 |
| Activity (EMAE), year-on-year change | EMAE | % | 0.5 |
| Unemployment | DES | % | 0.5 |
| Primary balance over tax revenue | FIS | % | 0.5 |
| Real wages, year-on-year change | SAL | % | 0.5 |
Similarity
Similarity = 100 · (1 − d / d_max), clipped to 0-100, where d is the weighted root-mean-square distance between the two fingerprints over the variables both have (at least 9 in common). Current d_max: 0.630. Method version: 1859088e6dd4.
The press axis
It measures how much today's headlines resemble those of each crisis's week, aggregated by day. The pilot uses a single press source, so the scale is its own (0-50) and 0 is frequent: it means the headlines do not resemble each other, not that they are missing. When they are truly missing it says 'no reading' (today) or 'no data' (the crisis).
The map axes
Horizontal: average of the percentiles of the FX gap, 30-day official dollar change, monthly inflation and the inverted real rate. Vertical: country risk, its 30-day change, and the inverted 30-day changes of Merval in USD and reserves. Each axis needs at least 2 of its 4 components. The 0.05–0.95 domain of each composite is stretched to 0-100 and the map is not recalibrated from one day to the next.
Two components carry zeros imposed by the panel: the FX gap is 0 before 2011 (no parallel-rate series before capital controls) and the official dollar change is 0 until January 2002 (fixed at 1). That is why 1999-2011 sits shifted to the left of the map.
The cells
The data area is split into 18 columns of hexagons. Each cell counts the sessions that landed in it (size) and takes the 90th percentile of the similarity with today among its comparable sessions (color): those from the last 90 days do not count, so today's cell does not light up by itself. Distance on the map correlates 0.72 with the real distance.
Publication jumps
Inflation, activity, unemployment and the fiscal balance arrive with a lag and as steps: the day INDEC publishes, today's fingerprint changes at once and the 30-session delta can move without the market doing anything.
Institutional access
Full fingerprints, the day-by-day similarity matrix and resonance alerts for asset managers, trading desks and family offices. Leave your contact and we will arrange access.