How far back can you test CAOS? To May 2017, with TAIL
CAOS (Long Volatility / Tail Risk) has its own monthly data from Apr 2023. Spliced with TAIL it reaches back to May 2017, which adds 5.9 years (63% of the extended history comes from proxies). Data through Sep 2026.
What you'd have missed
- CAOS's own data starts in Apr 2023; TAIL extends it to May 2017, 5.9 more years.
- Its own history misses the 2020 COVID crash and the 2022 inflation bear.
- The extended history's worst drawdown is -37.6% (May 2017–Feb 2023), versus -1.2% (Mar 2024–Apr 2024) in the ETF's own years.
- Its worst calendar year, 2019 (-14.3%), predates the ETF.
- Over the 42 overlapping months TAIL returned 12.8% a year less than CAOS, so the proxy years likely understate what CAOS would have earned by a similar margin.
- TAIL is a same-category stand-in, not a tracker (correlation 0.06 over 42 months). Read the years before Apr 2023 as a stress test of the category, not as CAOS's history.
How the history is built
CAOS (ETF) Apr 2023 → now
TAIL (Cambria Tail Risk ETF) used May 2017 → Mar 2023; its own data runs May 2017 → now
May 2017Sep 2026
Solid bars are the months each fund contributes to the spliced series; the outline shows how far a proxy's own data runs. Each month uses the youngest fund that has data for it; returns are spliced, not prices (dividends reinvested via adjusted close).
How well each proxy tracked CAOS
| Proxy | Kind | Months contributed | Overlap with CAOS | Correlation | Tracking error | Tracking difference | Beta | Vol ratio | Grade |
|---|---|---|---|---|---|---|---|---|---|
| TAILCambria Tail Risk ETF | ETF | May 2017–Mar 202371 months | Apr 2023–Sep 202642 months | 0.063 | 8.71% | -12.79% | 0.27 | 4.26 | Category stand-in |
Statistics are over the months a proxy and CAOS both have data. Tracking error is annualized; tracking difference is the annualized geometric gap (positive: the proxy earned more). Page grade: category stand-in, the worst grade among the proxies that contribute months.
CAOS's own history versus the extended history
| The ETF's own history | Proxy years only | Full extended history | |
|---|---|---|---|
| Window | Apr 2023 – Sep 20263.5 years | May 2017 – Mar 20235.9 years | May 2017 – Sep 20269.4 years |
| Growth of $10,000 | $11,702 | $6,346 | $7,426 |
| CAGR | 4.6% | -7.4% | -3.1% |
| Annualized volatility | 2.0% | 12.4% | 10.0% |
| Sharpe ratio (risk-free 0%) | 2.24 | -0.56 | -0.27 |
| Max drawdown | -1.2%Mar 2024 → Apr 2024; recovered May 2024 | -37.6%May 2017 → Feb 2023; not recovered within this window | -37.6%May 2017 → Feb 2023; not yet recovered |
| Worst complete calendar year | 2025 (2.6%) | 2019 (-14.3%) | 2019 (-14.3%) |
| Best complete calendar year | 2024 (5.3%) | 2020 (6.9%) | 2020 (6.9%) |
| Longest underwater (months) | 6 | 70 | 112 |
| Major declines covered | 0 | 2 | 2 |
The columns cover different years, so a higher or lower CAGR is not a verdict on the fund; the point is which market regimes each window includes. Growth of $10,000 is the same one-asset, buy-and-hold calculation the live portfolio app makes.
Growth of $10,000
Drawdowns
How CAOS did in each major US stock-market decline since 1985
| Decline | S&P 500 | CAOS over the same months | Source of those months |
|---|---|---|---|
| 1987 crashSep 1987–Nov 1987 | -29.8% | not covered | before available data |
| 1998 LTCM sell-offJul 1998–Aug 1998 | -15.3% | not covered | before available data |
| 2000–02 dot-com bearSep 2000–Sep 2002 | -44.7% | not covered | before available data |
| 2007–09 financial crisisNov 2007–Feb 2009 | -50.8% | not covered | before available data |
| 2020 COVID crashJan 2020–Mar 2020 | -19.4% | 23.8% | TAIL |
| 2022 inflation bearJan 2022–Sep 2022 | -23.9% | -4.8% | TAIL |
The decline windows are computed, not chosen: every peak-to-trough fall of at least 15% (month-end) in the S&P 500 series since 1985. The fund's figure compounds its monthly returns from the month after the S&P 500's peak through the trough.
CAOS's own five deepest drawdowns (extended history)
| Peak | Trough | Depth | Recovery | Source of the decline months |
|---|---|---|---|---|
| May 2017 | Feb 2023 | -37.6% | not yet recovered | TAIL |
This covers declines the equity windows miss, such as the bond bear markets of 1994 and 2022.
Caveats
- TAIL is a different ETF with a similar mandate. The measured gap to CAOS while both existed is -12.79% a year.
- Monthly month-end data: intra-month drawdowns were deeper than shown.
- Dividends are reinvested; no taxes or trading costs are modeled.
- Our data source's series for TAIL begins May 2017; the fund itself may be older.
Test it yourself
Open CAOS as a one-asset portfolio in the live app, where every number recomputes from fresh data.
Open CAOS (100%) in the portfolio builder →Used in portfolios: Dragon Portfolio
Used in guides: Dragon
Frequently asked questions
- How far back does CAOS data go?
- CAOS (Long Volatility / Tail Risk) has its own monthly data from Apr 2023. Spliced with TAIL it reaches back to May 2017, which adds 5.9 years (63% of the extended history comes from proxies). Data through Sep 2026.
- What did CAOS do before 2023?
- Before Apr 2023 the series is TAIL's. Those 5.9 years (May 2017–Mar 2023) show -7.4% a year with 12.4% annualized volatility and a worst drawdown of -37.6% (May 2017–Feb 2023). In major S&P 500 declines: 2020 COVID crash, 23.8% versus -19.4% for the S&P 500 (TAIL); 2022 inflation bear, -4.8% versus -23.9% for the S&P 500 (TAIL). The full extended history's worst drawdown is -37.6% (May 2017–Feb 2023).
- How closely does TAIL track CAOS?
- Over 42 overlapping months (Apr 2023–Sep 2026), TAIL had a correlation of 0.063 with CAOS, a tracking error of 8.7% a year and a tracking difference of -12.79% a year (grade: category stand-in).
Methodology
- Source
- Yahoo Finance via MarketHeist, monthly adjusted close (dividends reinvested). Months run to the last complete calendar month.
- Splice rule
- Each month uses the youngest fund that has data for it; returns are spliced, not prices (dividends reinvested via adjusted close).
- Numbers
- Computed by the MarketHeist portfolio engine as a one-asset, 100% buy-and-hold portfolio, the same calculation the live app makes. Sharpe uses a risk-free rate of 0%. Drawdowns are month-end.
- Correlation
- Pearson correlation of monthly returns over the months both funds have.
- Tracking error
- Standard deviation of the monthly return differences (proxy minus CAOS) times the square root of 12.
- Tracking difference
- Annualized geometric return of the proxy minus that of CAOS over the overlap. Positive means the proxy earned more.
- Beta and vol ratio
- Beta is the covariance of the proxy with CAOS divided by the variance of CAOS; vol ratio is the proxy's standard deviation divided by CAOS's.
- Grades
- Close: correlation of at least 0.95 and tracking error of at most 3%. Approximate: tracking error of at most 8% and either a cash-like fund (volatility of at most 3%) or correlation of at least 0.80 with a vol ratio of at most 1.25. Category stand-in: anything weaker.
- Generated
- 2026-10-10; data through Sep 2026. Reproduce hash of the full extended one-asset portfolio:
8263b364b6b4e599
Historical simulation, not investment advice. Past returns, including spliced proxy returns, do not predict future results.