The falling knives that don't bounce — down 25%+ from the high, with no path back.
Backsliders is a member of the Blaque Baux family. The core repo is the engine and blueprint — a governed, systematic platform (Julia) with a venue-agnostic execution controller and a Layer-3 live-money safety gate. Backsliders points that engine in its own direction and inherits the governance wholesale.
Not investment advice. Educational/research software. Nothing here is validated. See LICENSE.
git clone --recursive https://github.com/blaquebaux/backsliders.git
julia --project=engine -e 'using Pkg; Pkg.instantiate()' # one-time engine setupA SHORT sleeve, and an honest counter-test. The base found high-drawdown names tend to BOUNCE short-term (Blunt #5, Bore), so Backsliders' hard problem is separating that transient bounce from terminal decline: names down 25%+ with broken fundamentals and no recovery path, which keep sinking.
- Fallen-from-high screen — drawdown >=25%, then a quality filter for 'no path' (dilution, going-concern, broken growth).
- Beat the bounce — confirm these keep underperforming AFTER the short-term bounce window (else you are shorting into the Blunt #5 rally).
- Borrow / liquidity discipline — shorts cost borrow, and many are small and illiquid (Bottom's caveat).
Nothing above is implemented or validated. This is the map, not the territory.
Full detail in research/README.md. The scorecard:
| # | Question | Verdict |
|---|---|---|
| 1 | Does shorting the fallen pay? | ❌ no — short −0.28 (long +0.25); negative at every horizon |
| 2 | Can price isolate "no path"? | ❌ no — shorting "still-falling" is worse (−0.17); they bounce harder |
The synthesis: Backsliders is the negative image of the drawdown-bounce keeper. Fallen names bounce more often than they keep sinking (in tradable space), so shorting them is shorting into the rally — it loses at every horizon (−0.28), and a price "still-falling" filter is worse (−0.17). Survivorship cuts against the short here (the opposite of Bottom): the ideal targets — names that go to zero — delist and leave the sample, so this data both understates and can't access the real terminal-decline short. That short exists in principle but is a fundamental, specialist play (dilution / cash-burn / going-concern + borrow tolerance), not a price sleeve.
Research: first pass complete — null (research/). The short side is the wrong side; the
edge is the long drawdown-bounce (Bore / Blunt #5). No live driver. Nothing validated to the
spine's bar.
Blaque Baux is a quantitative research initiative and a subsidiary of Carter Warrens. BlaqueBaux.com is the home for the work; the code lives here on GitHub — open to study, test, and build bespoke strategies on top of.
Anyone can point an AI at a market. The edge is understanding what the data actually says — and turning it into something you can act on. We test relentlessly and put most of it on the record as rejected, with the reason; what survives is built, governed, and validated before it is ever called real. That combination — honest research, reproducible evidence, and execution you can trust — is why Carter Warrens leads on strategy and implementation, not merely uses the tools everyone now has.
This repo is one sleeve of the Blaque Baux family — a single governed engine steered in many directions. The core repo is the base/blueprint and holds the full family roster.
engine/ the Blaque Baux platform (git submodule -> blaquebaux/base)
research/ two Path-A sketches (short the fallen, the no-path test) + scorecard
live/ governed live drivers (once a sleeve graduates to paper A/B)
backsliders_2: strip the short leg entirely and go long-only the maximum 1-month losers, filtered to the bottom
quartile of short interest — isolating the oversold-bounce / "lottery ticket" premium while explicitly neutralizing the
short-squeeze risk that blows up long/short distressed books. backsliders already finds the long bounce is the edge; this
leg asks whether that bounce survives once the squeeze-prone (heavily-shorted) names are removed. Needs short-interest
data (not in Alpaca). (Folded from the batch-3 inversion set.)
MIT. (c) 2026 Carter Warrens.