Sleeve discovery
Sleeve discovery is one of 21 measurements this engine publishes in full. It is regenerated from a real run rather than transcribed, and it is shown here with its own claim boundary so the number and its limits arrive together.
The measurement
- Schema
- canli.sleeve-discovery.v2
- Program
- ALPHAC breadth program
- Admission contract schema
- canli.alphac-sleeve-admission-contract.v6
- Admission contract public path
- /glassbox/sleeve_admission_contract.json
- Source path
- config/sleeve_discovery.json
Objective
- Portfolio max drawdown target
- 0.110000
- Target kind
- honest out-of-sample research objective, never a promised result
- Target sleeve count
- 14
- Minimum new sleeves
- 10
- Candidate atlas minimum
- 200
Portfolio sharpe target
- 2
- 2.500000
Drawdown target kind. approximate research objective evaluated out of sample and under stress, never a loss limit or promised result
Portfolio requirement. Improve the combined book through independent economic return sources, not leverage or selection of the prettiest backtest.
Admission rule. A target never creates a sleeve. Fourteen is reached only if ten or more new identities independently clear every admission gate; otherwise the book remains smaller.
Admission gates
- Point in time data
- yes
- Net of costs
- yes
- Walk forward only
- yes
- Requires capacity curve
- yes
- Requires kill criteria
- yes
- Book deflated sharpe min
- 0.950000
- Pbo max
- 0.200000
- Average pairwise correlation max
- 0
- Average pairwise correlation upper 95 max
- 0.100000
- Pairwise correlation max
- 0.350000
- Stressed pairwise correlation max
- 0.500000
- Minimum oos observations
- 756
- Net sharpe min
- 0.150000
- Capacity usd min
- 500000
Credential policy
- Storage
- environment variables or managed secret store only
- Request timing
- Only after a candidate passes its key-free feasibility screen.
Never
- source code
- research artifacts
- chat transcripts
- client-side bundles
Feasibility reviews
Nested beyond what this page renders. Read it in the artifact.
Audit
- Performed at
- 2026-08-15
- Public source
- glassbox/kill_log.json
Finding. The v1 queue accidentally relisted eight previously tested families. They are preserved below as retired, not presented as fresh research.
Candidates
Nested beyond what this page renders. Read it in the artifact.
| Id | Public result | Verdict | Net sharpe | Dsr | Newey west t | Average correlation | Max pair correlation | Capacity p05 usd at 1pct adv | Reason |
|---|---|---|---|---|---|---|---|---|---|
| futures_curve_carry | futcarry_xs and xasset_carry | KILLED | — | — | — | — | — | — | — |
| options_variance_premium | equity_vrp and crypto_vrp | KILLED | — | — | — | — | — | — | — |
| post_earnings_drift | equity_pead | KILLED | — | — | — | — | — | — | — |
| crypto_basis_curve | crypto_dated_basis and funding_termstructure | KILLED | — | — | — | — | — | — | — |
| rates_curve_relative_value | rates_curve_carry | KILLED | — | — | — | — | — | — | — |
| short_horizon_reversal | PRE_REGISTRATION Sleeve E and reversal IC campaign | KILLED | — | — | — | — | — | — | — |
| fx_value_carry | fx_carry_ratediff, fx_carry_trend_overlay and xasset_value | KILLED | — | — | — | — | — | — | — |
| cross_asset_defensive | quality, BAB and tail_overlay campaigns | KILLED | — | — | — | — | — | — | — |
| insider_purchase_clusters | artifacts/probe/insider_purchase_clusters/result.json | KILLED | -0.243230 | 1.644e-6 | -0.785510 | -0.064277 | 0.057034 | 5167351.429733 | The sleeve was genuinely diversifying and capacity-feasible, but had negative OOS returns, DSR effectively zero, and reduced the fixed-weight ALPHAC book in three of four leave-one-year-out recomputations. The sign is not inverted and the parameters are not retuned. |
| commodity_inventory_seasonal | artifacts/probe/eia_petroleum_inventory/result.json | KILLED | -0.589203 | 1.395e-9 | -1.844977 | 1.577e-4 | 0.032074 | 14919.439129 | The one-shot petroleum proxy was exceptionally orthogonal but had strongly negative OOS returns, failed both standalone product checks, failed mean-zero and leave-one-year-out book stability, and was severely capacity-constrained by UGA. The sign is not inverted and the parameters are not retuned. |
Retired admission gates
Deflated sharpe min. Retired as a per-sleeve GATE in contract v6 and replaced by book_deflated_sharpe_min at the same 0.95 against the complete union of hypothesis identities. A 0.95 per-sleeve floor required an annualized Sharpe of 1.184 even at the least deflation the formula permits, about eight times the declared net Sharpe floor, and no sleeve in this book clears it. The per-sleeve figure is still measured and published for every candidate; evidence that omits it fails closed.
Frontier arithmetic
- Identity
- S_book = s_bar * sqrt(N / (1 + (N - 1) * rho_bar))
- Target sleeve count
- 14
- Psd floor at target n
- -0.076923
- Correlation gate in force
- 0
- Gate permits objective floor
- no
- Gate permits objective ceiling
- no
Psd floor note. Average pairwise correlation cannot fall below -1/(N-1) for any real correlation matrix. At fourteen sleeves that floor is far from binding, which is why a negative average is arithmetically available here and is not at 250 sleeves.
Book sharpe ceiling at the gate
- S bar traded basis
- 1.979337
- S bar four curve basis
- 1.736129
Quality precondition at the gate
- S bar required for 2 / 25
- 0.601338
- S bar required for 3 / 0
- 0.801784
- S bar measured traded basis
- 0.529000
- S bar measured four curve basis
- 0.464000
Note. What the tightened correlation gate alone does NOT buy. At the gate exactly, reaching each end of the objective band requires at least this average standalone Sharpe across fourteen sleeves.
Correlation required at measured quality
Note. The other route to the same place: hold quality where it is measured today and read off the average pairwise correlation each end of the band demands.
Traded basis
2 further fields. Read them in the artifact.
Four curve basis
2 further fields. Read them in the artifact.
Honest reading. Tightening the correlation ceiling from 0.15 to 0.00 removes a ceiling that sat BELOW the objective at every plausible quality, which is why the v4 gate could not have produced the objective at any sleeve count. It does not on its own deliver the objective. At the gate exactly, fourteen sleeves of today's measured quality reach the figure recorded above, and the remaining distance has to be bought with per-sleeve quality, genuinely negative correlation, or both. Publishing this is the point: a target sitting beside a gate that forbids it is not a stretch goal.
Check it yourself
Every figure above is read from /glassbox/sleeve_discovery.json, the
artifact the engine wrote. Nothing on this page is typed by hand: the page is generated from that
file, so a figure that moves in the artifact moves here and a figure that is not in the artifact
cannot appear here at all.