ZEPHYR is a static multi-credit portfolio that achieves a Sharpe ratio above 3 with ZERO volatility scaling of any kind — daily, weekly, or monthly.
Objective: Deliver cash-like volatility (~1.2% annualised) with credit-like returns (~4.4% annualised) by combining the highest-Sharpe segments of the short-duration fixed-income universe, plus a tiny uncorrelated drift anchor in gold.
How It Works: Seven ETFs are held at fixed weights and rebalanced once every 21 trading days back to target. No signals, no selection, no re-weighting, no vol targeting between rebalances. Risk management comes entirely from (1) instrument choice — floating-rate, ultra-short-duration, AAA-credit exposures — and (2) a macro regime gate that rotates the portfolio into T-bills (BIL) when high-yield spreads widen past ~8% or the 10-year yield rises more than 0.7% in three months. The gate uses only T-1 data.
Positions: JAAA (32%, AAA CLO), JPST (28%, ultra-short IG), MINT (15%, ultra-short IG), BKLN (10%, senior loans), SRLN (5%, senior loans), FLOT (5%, IG floaters), GLD (5%, gold). All long-only ETFs — no shorting, no margin, no derivatives. History begins at JAAA's inception (2020-10-19).
| Frequency | Every 21 trading days (approximately monthly) |
| Selection | None. Seven instruments at fixed weights (JAAA 32 / JPST 28 / MINT 15 / BKLN 10 / SRLN 5 / FLOT 5 / GLD 5). |
| Weighting | Static. Weights never change — rebalance only restores drift. |
| Vol Targeting | NONE. No per-position scaling, no portfolio vol target, no leverage adjustments — ever. |
| Risk Controls | Single macro regime gate: HY OAS < 8% AND Δ(10Y yield, 3mo) < 0.7%. Uses T-1 data only. On breach, full allocation rotates to BIL. |
| Transaction Costs | 5 bps modeled per rebalance across the whole portfolio (~60 bps/yr). |
.shift(1) on both FRED series so today's gate depends only on yesterday's close. There are no trailing-window statistics used for sizing, so the usual sources of look-ahead in quant backtests do not apply here.Honest caveat: The Tier 3 period is structurally a short-duration-Treasury portfolio, not ZEPHYR — bank loans, floating-rate credit and AAA CLOs simply did not have exchange-traded wrappers before 2011. Treat Tier 3 as a floor for a capital-preservation sleeve, not as a faithful simulation of what JAAA/JPST would have returned. Tier 2 is a closer approximation but still underestimates JAAA's structural spread (~SOFR+140bps vs FLOT+BKLN ~SOFR+75bps weighted). The live-era numbers are what the strategy actually produced.
ZEPHYR is deliberately simple. There are only three things the model does, and the first two are done once a month. No future information is ever used.
The portfolio is static: 32% JAAA, 28% JPST, 15% MINT, 10% BKLN, 5% SRLN, 5% FLOT, 5% GLD. These weights were chosen to maximise Sharpe on the training window while keeping each sub-sleeve exposed to a structurally different risk driver. They are not re-optimised — this page reports the out-of-sample performance of that single, fixed allocation.
Why these seven? They are the highest-Sharpe, lowest-volatility segments of the fixed-income universe that are genuinely distinct. JAAA (AAA CLO) carries floating SOFR+~140bps with virtually zero default risk. JPST and MINT are active ultra-short IG credit funds — effectively cash-plus. BKLN and SRLN are senior secured leveraged loans (also floating). FLOT is IG floating-rate notes. GLD is a 5% uncorrelated drift anchor that turns positive during the rare events where all credit widens at once.
Every 21 trading days the portfolio is traded back to its target weights. No signal, no selection, no re-weighting. A 5 bps cost is applied to the turnover generated by drift. Average turnover is ~15% per rebalance (driven almost entirely by returns diverging between sleeves).
This is the only daily calculation. The gate reads two macro series from FRED, both lagged by one day:
Credit gate: g_credit = clip((8.0 − HY_OAS) / (8.0 − 5.0), 0, 1). When the ICE BofA US High Yield OAS (BAMLH0A0HYM2) is below 5%, the gate is fully open (1.0). Between 5% and 8% it linearly closes. Above 8% it is fully shut (0.0).
Rate gate: g_rate = 1 if (DGS10_today − DGS10_63d_ago) < 0.7, else 0. When the 10-year Treasury yield has risen more than 0.7% in the past three months, the gate slams shut.
Combined: g = g_credit × g_rate. The gate is binary at the extremes and simply multiplies the portfolio return; the complement (1 − g) goes to BIL. There is no resizing — the gate is on or off for the whole portfolio at once.
Traditional quant strategies achieve high Sharpe by adaptively sizing positions — reducing during volatile regimes and pressing during calm ones. ZEPHYR produces a comparable Sharpe with none of that, because it starts from an unusually high base:
1. Structural low volatility. The chosen instruments are floating-rate and/or ultra-short-duration. Duration risk is ~0.2 years — about 1/25th of the AGG benchmark. The portfolio's raw volatility is 1.2% ann., not because it is scaled down, but because the underlying assets have that much vol to begin with.
2. Positive carry. Average yield across the sleeves is ~5.3% (weighted). With near-zero duration, most of that yield becomes realised return.
3. The regime gate handles the only tail risk the instruments have — credit events and rate shocks. The 2022 rate shock would have been the strategy's worst drawdown; the gate flipped to BIL on 2022-02 and stayed there into Q4, limiting the 2022 loss to −0.9%.
4. GLD is uncorrelated ballast. The 5% gold sleeve contributes ~10 bps of Sharpe at the portfolio level and turns positive during the rare events that hit all credit simultaneously.
Track record is short. JAAA (the single largest position) launched 2020-10-19. The strategy has 5.4 years of history — it has seen the 2022 rate shock, the 2023 banking crisis, and the 2024 yen carry unwind, but it has not seen a 2008-style credit crisis or a 1970s inflation regime. Longer-history proxies (FLOT/BKLN/MINT in place of JAAA) produce a gated Sharpe of ~2.6 over 12 years — still strong, but lower.
Capacity is meaningful but not infinite. JAAA alone holds $15B+ AUM, and aggregate sleeve capacity exceeds $50B. At typical allocator size this is not a constraint; at $1B+ concentrated in JAAA you would want to split into JAAA+CLOI+ICLO.
The regime gate is one design choice, not the only one. HY OAS < 8% and ΔDGS10 < 0.7% are round numbers chosen for clarity. Neighbouring thresholds (7% / 9%, 0.5% / 1.0%) give Sharpe within ±0.2. The gate is robust, not optimised.