ZetaMindsQUANTITATIVE MARKET RESEARCH
Knowledge · Market regimes

Seven kinds of trading day

Every NYSE trading day from 15 April 2019 to 24 September 2026 (1,871 days), sorted into seven market regimes by how markets behaved over roughly the prior month: trend, fear, rates, AI leadership, growth vs value and cross-asset moves.

1,871
trading days classified
7
regimes
20
descriptors in 6 themes
0.84
bootstrap stability (ARI)
Latest day · 24 September 2026

Regime 7: Speculative / AI melt-up, day 2 of its spell

S&P 500 on a log scale, each day shaded by its regime. Hover or tap to read any day.

1 Crash / panic2 Bear market / rate shock3 Rebound, growth-led4 Reflation / rising yields5 Late-cycle calm (flat curve)6 Steady low-vol bull7 Speculative / AI melt-up

Source: Yahoo Finance, ZetaMinds regime model. Daily, 2019-04-15 to 2026-09-24.

The regimes

From crash to melt-up

Numbered from most bearish to most bullish by how far the S&P sits from its 200-day average. Figures are averages over each regime's days.

Regime 1 2.4% of days

Crash / panic

Markets in free fall. The S&P sits 11% below its 200-day average, the VIX near 50 and the 10-year yield drops 41bp in a month.

S&P vs 200-day average
−10.8%
Drawdown from 1-year high
−20.1%
VIX
49.1
10-year yield · 20-day change
0.88% · −41bp
Curve, 10Y − 3M
+0.56pp
AI stocks vs Nasdaq, 20 days
−1.3%
Growth vs value, 60 days
+10.9%
Bitcoin, 20 days
−10%
Average spell
22.5 days

Main periods: Feb–Apr 2020. Essentially one event: COVID.

Regime 2 10.5% of days

Bear market / rate shock

A grinding sell-off with rising yields. The S&P sits 7% below trend, the VIX near 28, and value beats growth by 9% over 60 days.

S&P vs 200-day average
−7.3%
Drawdown from 1-year high
−16.6%
VIX
27.5
10-year yield · 20-day change
3.45% · +14bp
Curve, 10Y − 3M
+0.65pp
AI stocks vs Nasdaq, 20 days
−0.2%
Growth vs value, 60 days
−8.8%
Bitcoin, 20 days
−10%
Average spell
15.1 days

Main periods: 2022, April 2025.

Regime 3 10.4% of days

Rebound, growth-led

Recovery after a sell-off, led by growth. The S&P is back just above trend, fear is still elevated, growth beats value by 9% and Bitcoin rises 12% in a month.

S&P vs 200-day average
+1.7%
Drawdown from 1-year high
−9.1%
VIX
23.2
10-year yield · 20-day change
2.72% · −1bp
Curve, 10Y − 3M
−0.07pp
AI stocks vs Nasdaq, 20 days
+0.6%
Growth vs value, 60 days
+8.7%
Bitcoin, 20 days
+12%
Average spell
9.3 days

Main periods: Mid-2020, 2023, mid-2025.

Regime 4 10.5% of days

Reflation / rising yields

Stocks rising with yields. The 10-year climbs 20bp in a month on a steep curve, value beats growth by 8% and AI stocks lag the Nasdaq.

S&P vs 200-day average
+5.0%
Drawdown from 1-year high
−3.6%
VIX
21.4
10-year yield · 20-day change
2.86% · +20bp
Curve, 10Y − 3M
+1.20pp
AI stocks vs Nasdaq, 20 days
−2.5%
Growth vs value, 60 days
−8.4%
Bitcoin, 20 days
−2%
Average spell
9.4 days

Main periods: 2021, early 2022, Feb–Mar 2026.

Regime 5 24.3% of days

Late-cycle calm (flat curve)

Quiet markets on a flat or inverted curve. The VIX sits near 18 and no style leads for long.

S&P vs 200-day average
+6.4%
Drawdown from 1-year high
−3.5%
VIX
18.4
10-year yield · 20-day change
3.30% · −0bp
Curve, 10Y − 3M
−0.18pp
AI stocks vs Nasdaq, 20 days
−0.7%
Growth vs value, 60 days
+1.8%
Bitcoin, 20 days
0%
Average spell
7.0 days

Main periods: 2019, 2023–24.

Regime 6 31.3% of days

Steady low-vol bull

The most common day. The S&P sits 10% above trend, the VIX near 16, rates are steady and growth edges ahead.

S&P vs 200-day average
+9.8%
Drawdown from 1-year high
−0.6%
VIX
15.6
10-year yield · 20-day change
2.97% · −3bp
Curve, 10Y − 3M
+0.05pp
AI stocks vs Nasdaq, 20 days
−0.6%
Growth vs value, 60 days
+5.1%
Bitcoin, 20 days
+9%
Average spell
11.7 days

Main periods: 2019–21, 2024–25.

Regime 7 10.6% of days

Speculative / AI melt-up

The strongest trend, with froth. The S&P sits 11% above trend while AI stocks beat the Nasdaq by 7% in a month and Bitcoin gains 22%.

S&P vs 200-day average
+11.0%
Drawdown from 1-year high
−1.0%
VIX
19.7
10-year yield · 20-day change
2.85% · +10bp
Curve, 10Y − 3M
+0.82pp
AI stocks vs Nasdaq, 20 days
+6.7%
Growth vs value, 60 days
+0.4%
Bitcoin, 20 days
+22%
Average spell
8.6 days

Main periods: Late 2020, 2021, 2026.

2019 and 2023–24 were calm; 2022 was a bear market

Share of each year's trading days in each regime.

2019
2020
2021
2022
2023
2024
2025
2026

2019 starts 15 April 2019; 2026 runs to 24 September 2026.

Regimes persist: about 9 days in 10, today's regime is yesterday's

Row: yesterday's regime. Column: today's. Percent of days.

From1234567
1 96220000
2 19322200
3 02891432
4 03189413
5 00228681
6 00106912
7 00135389
Regimes × stock segments

One growth-vs-defensive axis that regimes flip

Annualised return of each stock segment above or below the equal-weight universe, by regime. Pre-profit growth and cash burners lead in rebounds and melt-ups; low-beta value and yield lead in reflation and late-cycle calm.

Segment1 Crash2 Bear3 Rebound4 Reflation5 Calm6 Bull7 Melt-up
1 Mature franchises +14% 0% 0% +1% +5% +1% −16%
2 Quality growth +66% −3% +38% −29% +6% +17% +6%
3 Ex-growth cyclicals −14% −9% −6% +10% −6% −6% +3%
4 Low-margin growers −31% +3% +6% −1% 0% +5% +9%
5 Banks & lenders −53% +14% −23% +22% −1% −2% +10%
6 Low-beta value −8% +23% −27% +42% +13% −21% −32%
7 Yield & real assets −11% +1% −38% +28% +20% −20% −42%
8 Pre-profit growth +37% −36% +59% −72% −35% +23% +56%
9 Cash burners & broken growth +65% −18% +33% −64% −37% +18% +52%

Faded cells are not statistically reliable (|t| < 2, spell-clustered, or too few spells). Crash / panic is essentially one event. Segment labels are from the September 2026 snapshot and the regimes are fitted on the full sample, so this table describes the past with hindsight. Full technical report.

Educational analysis, not investment advice. Past behaviour does not predict future returns.

Methodology

How the regimes are built

The same idea as customer segmentation, with the trading day as the customer: cluster on how markets behaved, then describe each cluster with outcomes not used to build it.

Twenty descriptors, six themes

  • Trend: S&P vs its 50- and 200-day averages, 1-year drawdown, RSI.
  • Fear: VIX level and its trend, realised volatility, VIX ÷ realised volatility.
  • Rates: 10-year yield, its 20-day change, 10Y − 3M curve.
  • Tech & AI: Nasdaq vs S&P, AI ETF vs Nasdaq, AI ETF vs its trend.
  • Style: growth minus value over 20 and 60 days.
  • Cross-asset: gold, oil and Bitcoin over 20 days.

Clustering

Each descriptor uses a 20–60 day trailing window, so a regime reflects the last month rather than one day's noise. Descriptors are robustly scaled (median and interquartile range, clipped at ±4), each theme weighted equally, then grouped with k-means.

Seven regimes was chosen from 3 to 10 on stability: when 20% of days are dropped at random, seven regimes come back with an agreement of 0.84, against 0.72 or less for every other choice.

What it cannot tell you

  • Regimes are zones on a continuum, not islands (silhouette about 0.12). About 29% of spells are single-day blips at the borders.
  • The scaling and centroids are fitted on 2019–2026, so past labels use hindsight. A live version must refit on an expanding window.
  • Because the 10-year level is a descriptor, some regimes are partly tied to rate eras.
  • Regimes describe the backdrop. They do not forecast returns.

Full technical report