As of 2026-08-13 · run #14 · baskets formed 2026-08-03 ·
universe 828 names · daily dashboard · archive
Account placement matters more than anything on this page.
The momentum sleeve replaces roughly a third of its names monthly (historically 20–45%) —
in a taxable account nearly all gains are short-term. Run this in an IRA or equivalent if
you possibly can. Dividends accrue as cash until the next rebalance; don't rebalance
intra-month.
Allocator
stays in your browser (localStorage) — never sent or published
Rebalance
Performance since inception
Barbell since 2026-07-28
+7.4%
base 100 at inception
vs SPY
+2.2pp
SPY +5.0% same window
Realized vol (21d)
35%
12 sessions
VT20 dial
57%
min(1, 20% / vol) — informational
Live out-of-sample record from first formation (2026-07-28),
computed nightly from closes: daily equal weight, new basket accrues from the next session,
5 bps/side membership turnover. Base 100. This is the record the backtests get judged against.
The strategy, honestly
Construction: 40% momentum (12-1 return, top 25 of the ≥$10B universe) +
60% low-vol (lowest-vol quintile, 126-day realized). Rebalance on the first trading day of
each month at the close; positions accrue from the next session. No trend filter.
Expectations: Rebuilt 15y point-in-time engine (live convention): ~0.82 Sharpe, ~+14%/yr, −36% once-a-cycle drawdown, worst calendar year ≈ −6%. Plan around these, not the flattered 3-year numbers.
10m-SMA trend gate
REJECTED 2026-08-04 — lowers Sharpe in every variant; net-negative even inside the 2020/2022 bears.
20% vol-target (21d, cap 1.0)
APPROVED 2026-08-05 as optional crash insurance only: maxDD −36%→−20% for ~1.4pp CAGR/yr. The dial below shows its current exposure; the default book stays unfiltered.
Month-end formation
REJECTED 2026-08-05 — the apparent edge was a cap-vintage artifact; formation stays first-run-of-month.
Cap-lag universe (paper A/B)
Pre-registered live log since 2026-08-05; not adopted. Not part of these instructions.
Low-vol sleeve: three ways to hold it
Option
What it is
Cost vs tested strategy
Full quintile, fractional (default)
All 165 names at equal weight — exactly the tested strategy. Needs a broker
with fractional shares (Fidelity, Schwab, IBKR).
None — this IS the strategy.
Top-40 lowest-vol (whole-share mode below)
First 40 names of the ranked quintile — 65 positions total.
One ticker for the whole sleeve. Changes the strategy: external MSCI methodology,
correlation 0.942 to our quintile.
Backtests comparably (barbell give-up +0.00)
but it is no longer the tested construction.
Execution checklist & FAQ
Baskets form at the evening run on the first trading day of the month (this page updates
that night). Trade at the next session's open — the engine assumes fills at the
formation close, so next-open is the closest realizable implementation (the overnight gap
is unmodeled). "Sometime that week" quietly adds tracking error.
Sell everything on the sell list (full positions), then buy the buy list at the weights
shown by the allocator. Market-at-open orders are fine at these liquidity levels.
Joining mid-month? Buy the current baskets now at allocator weights; expect a few
basis points of tracking noise vs having entered at formation. Or park in BIL until the 1st.
Missed a rebalance? Do it when you notice, using this page's current baskets —
the screener itself forms late the same way when the machine was off on the 1st.
Halted/delisted name? Hold the proceeds as cash until the next rebalance. (The
NAV record instead redistributes the weight across surviving names — a small, known
tracking difference vs your book.)
VT20 overlay (optional, default off): scales the whole book by w = min(1, 20% /
trailing 21-day vol), remainder in BIL. It is crash insurance priced at ~1.4pp CAGR/yr —
see the research table. Toggle it above only if a ~−20% drawdown floor matters more to you
than that premium.