Where does the economic cycle stand?
A selection of indicators computed every night from public sources: US corporate guidance, central banks, rates, liquidity and commodities.
Cycle phase diagnosis, an explanatory note every Monday, week-on-week comparison, full archive and PDF reports. Registration is free.
How to read Antiktj
No indicator is trusted on its own. Antiktj cross-checks the guidance of large and small companies, commodities, equity prices, central banks, liquidity and the real economy, and keeps as signals only those that preceded past turning points across nine stock markets.
The more signal families light up together, the rarer the configuration: vigilance, confirmation, then full alignment. In the past, a turning point followed about one week of confirmation in two, against one random week in four.
It does not time markets: its signals precede turning points by six to eighteen months, and may light up with no follow-through. It gives no advice. It describes; judgment rests with the reader.
Demonstration: what Antiktj would have shown in 2006-2008
Week after week, from 2005 to mid-2009: the state of the four signal families and the level of the scale, computed with today’s rules, next to the S&P 500 and the CAC 40 (base 100 in January 2005). Vigilance lights up on May 12, 2006, confirmation on August 4, 2006: ten months before the CAC 40 peak and fourteen months before the S&P 500 peak.
Note: 2007 is part of the period on which the rules were tuned for the S&P 500. The company rule, recalibrated without the 2007 episode, would have detected it fourteen months ahead, and the CAC 40 was never used for tuning. A retrospective demonstration: it shows the method at work, with no forecasting value.
Companies
Central banks
Rates
Liquidity
Real economy
Commodities
Latest central bank decisions
Method
Indicators are recomputed every night. For each one, the deviation from normal compares the latest value with its average over the previous two years, in standard deviations; a card marked in orange deviates by more than 1.5. The corporate indicator is computed by Antiktj from earnings releases filed with the SEC.
This page shows figures only, with no commentary or interpretation, and reproduces no raw price series.
Sources: SEC (EDGAR), Federal Reserve Bank of St. Louis (FRED), OECD, European Central Bank, Federal Reserve, IMF.
Method and validation: protocol, results and limits
Appendix: method and validation (results as of October 12, 2026)
Definitions
- Turning point: a decline of at least 13% in a stock index from its high of the previous twelve months; intermediate technical rebounds are merged into a single episode.
- Qualification: each episode is classified as internal (economic and financial cycle) or external (pandemic, disaster, political decision, brief market shock) by the analyst, market by market, before any calculation. Only internal episodes are used to judge the rules.
- Detection: an alert within the 18 months before the episode's peak. False alarm: an alert not followed by any 13% decline within 18 months.
- Deviation from normal: distance of a value from its average over the previous two years, in standard deviations. It measures distance, not probability.
Protocol
- Rules calibrated on the S&P 500, 2004 to 2012.
- Cross-validation: each internal S&P 500 episode (2007, 2015, 2018, 2022) is left out in turn, the rule is recalibrated without it, then judged on it.
- Rules frozen, then applied without any adjustment to eight other markets never used for tuning: CAC 40, DAX, FTSE 100, Ibex 35, Nikkei 225, Hang Seng, Shanghai, Sensex (28 internal episodes from 2004 to 2025, corresponding to six or seven major global turning points: they are not independent).
Frozen rules and results outside the United States
| Rule | Definition | Seen ahead | Missed | False alarms |
|---|---|---|---|---|
| Operating net balance | Guidance raises minus cuts (excluding accounting, currency and scope effects) above +1.25 standard deviations | 24 / 28 | 2 | 10 |
| Copper | Year-on-year change beyond ±2.25 standard deviations | 22 / 28 | 0 | 9 |
| Wheat | Year-on-year change above +1.25 standard deviations | 22 / 28 | 3 | 7 |
| Geopolitical threats | Caldara-Iacoviello index (threats) above +2 standard deviations | 19 / 28 | 3 | 4 |
| Small vs large caps gap | Net balance of a sample of 300 non-S&P 500 companies per quarter, minus the S&P 500's, below -1.25 standard deviations | 26 / 28 | 0 | 9 |
| "Conflicts" theme | Share of earnings releases mentioning conflicts, above +1.75 standard deviations | 18 / 28 | 2 | 3 |
| "Mortgage credit" theme | Above +1 standard deviation | 20 / 28 | 5 | 5 |
| "Inflation" theme | Above +1.25 standard deviations | 20 / 28 | 1 | 10 |
| Industrial production | Year-on-year growth above 3.9% | 19 / 28 | 2 | 9 |
Reading the table: for each rule, seen ahead + seen late (alert after the peak) + missed = 28. Seen late: operating net balance 2, copper 6, wheat 3, small vs large caps gap 2, geopolitical threats 6, conflicts 8, mortgage credit 3, inflation 7, industrial production 7.
Equity prices (rules set on the S&P 500, each seeing 3 of the 4 internal declines ahead): Nasdaq gap to its four-year average above +2.25 standard deviations; at least 46 S&P 500 records in twelve months; S&P 500 gap to its four-year average above +2 standard deviations.
Concordance scale
Vigilance: 2 of the 3 validated rules (operating net balance, copper, wheat) triggered within six months. Confirmation: 3 of 4 families, the fourth being equity prices. Full alignment: 4 of 4 families.
| Situation (nine markets, weeks from 2004 to mid-2024) | Followed by an internal turning point within 18 months |
|---|---|
| No rule active | 10% |
| Reference: random week | 24% |
| Vigilance | 44% |
| Confirmation | 50% |
| Full alignment (only 22 weeks, almost all in 2021) | 73% |
Contribution of each family (confirmation level, outside the S&P 500): complete, 20 of 28 turning points seen and 3 false alarms; without companies, 6; without wheat, 7; without equity prices, 15; without copper, 20, copper partly overlapping with the others at this level. Independence: two families are active together only 15 to 30% of the time either one is.
Maximum-sensitivity scale (complementary reading, added on October 13, 2026): it counts the small vs large caps gap among the validated rules. Vigilance = 2 of 4 rules; confirmation = 3 of 5 families; full alignment = 5 of 5. Outside the United States, its confirmation preceded 26 of 28 turning points with none missed, and 8 false alarms (against 20, 7 missed and 3 false alarms for the main scale). A turning point followed 39% of the weeks it was reached, against 50% for the main scale. It was tested against a criterion set in advance (no more false alarms than the main scale), which it does not meet: the main scale remains the diagnosis, and this one a complementary reading.
Limits
- Few independent episodes: these frequencies describe the past and are not probabilities for the future.
- Six- to eighteen-month horizon: a signal may stay on long before a turning point, or switch off without one.
- Company indicators rely on US firms; monthly or quarterly series are published with a lag (see the "Latest data" column).
- Context layers (M3, debt, crypto-assets, regulators, themes not retained) are tracked without established signal value.
Sources: SEC (EDGAR), Federal Reserve Bank of St. Louis (FRED), European Central Bank, OECD, IMF, Yahoo Finance, DefiLlama, Caldara and Iacoviello (Federal Reserve), releases from the SEC, FCA, ESMA, AMF, FSB and BIS.