A parallel paper deployment carries a wider universe — 19 allowlisted assets, 20 attested, 22 rules — and exists purely to accumulate evidence without risking funds.
+Alongside it runs a paper deployment — separate database, separate config, synthetic cash. It is not a toy: it is the evidence pipeline that is supposed to feed live. The next section is about why it currently does not.
+ + +Paper → live
+A rule is not supposed to touch real money until it has earned it. keel encodes that as a status ladder with a deterministic gate between the last two rungs — and the gate is doing its job, which is how we know the strategy cannot clear it.
+ +| Item | Detail | Figure |
|---|---|---|
| Mode | Paper — synthetic equity, no venue orders | $500 → $550 |
| Allowlist | The live six plus SOL, LTC, LINK, ZEC, NEAR, AVAX, UNI, FET, ICP, DOT, CRV, ALGO, BCH | 19 |
| Attested assets | Human-classified, each with a source | 20 |
| Rules on the books | 19 × turtle_breakout at paper, 1 at candidate, 2 × dca disabled | 22 |
| Rules that actually load | Only status=paper loads in paper mode — the other three are inert | 19 |
| Signals produced | To date | 0 |
| Orders, positions, outcomes | To date | 0 |
A rule moves candidate → paper → live. Status is not a label: it decides whether the agent loads the rule at all. Promotion to live requires clearing every floor below, and the floors were deliberately not relaxed together — when the win-rate axis was loosened for trend-following, min_trades was held at its canonical value on the grounds that the two axes are independent and only one had a justification.
| Floor | Required | Status against measurement |
|---|---|---|
min_trades | 100 | Unreachable — see below |
min_expectancy | > 0 | Untested at scale |
min_rr | 1.5 | Untested at scale |
min_win_rate (trend class) | 0.30 | Relaxed from 0.55, deliberately and alone |
| PBO / overfitting gate | required | Blocks while unset |
min_trades = 100 is checked per rule, per product — it does not pool across assets. Measured trade rates for the daily turtle:
Waiting for the gate is not a slower path to the same place; it is no path. And adding assets does not help — the floor is per rule, so a new asset yields a new rule with its own ~14 trades, not a pooled 114.
+Five of the seven live rules carry promoted_at = NULL. They were seeded straight to live and have been left there. That is a recorded decision, not an oversight — the reasoning sits in the live config, reviewed and dated. Only the BTC dca rule and the DOGE turtle were promoted through the normal path.
What bounds the risk is therefore not the promotion gate, which never ran. It is the position caps, the nineteen rails, and the allowlist — and those are separate mechanisms that all did run. The bypass is of the evidence gate, not the safety rails.
+This is also why paper matters more than its zero signals suggest. It is the only apparatus that could ever produce the evidence the gate demands, and the honest reading of 0 signals is not that paper is broken — it is that a rule firing twice a year cannot generate a hundred trades inside a human attention span.
Shariah basis
+The compliance screen is not a curated list of "approved coins". It is a set of named juristic tests, each traceable to a source, each enforced in code that fails closed.
+ +| Criterion | What it tests | Source | Waivable? |
|---|---|---|---|
attestation | Sector, backing and yield are human-attested with a source. Nothing is derived from price data, so an unclassified asset is unknown — and unknown is a rejection | §28.4 | Never |
haram_sector | The token's core business line is not an excluded one | §28.4 | Never |
riba_yield | Holding it carries no guaranteed or expected return — that is riba-like, and not a bare spot position | §28.4 | Never |
backing | 'ayn (a tangible, owned thing) versus dayn (a debt claim on an issuer). A pure claim is refused | §65.5 · §67.2 | Never |
| └ gold & silver | If the backing is gold or silver, the stricter bay' al-sarf treatment applies — relevant to PAXG | §65.5 | Never |
instrument_wrapper | Admission must name the contract, not just the asset: the allowlist is not juristically homogeneous, and a CFD can spell itself exactly like spot | §71.4a | Never |
withdrawal_capability | Qabd — possession holds only while nothing prevents withdrawal on demand. Rail 17 halts new entries when it cannot be shown | §65.4 | Never |
settlement | USD-settled spot only; every other contract class is refused mechanically | rails 18/19 | Never |
history | Four years of daily bars before a rule can be validated | evidence, not fiqh | Yes, documented |
Only one criterion in the whole screen can be waived, and it is the one that is not a religious question. The code does not consult the waiver map for any of the others, and the CLI will not accept them — expanding that set is a deliberate decision, not something a passing test can quietly do.
+ +| Ref | Source | Role |
|---|---|---|
| §28 | Andreas Jobst, The Economics of Islamic Finance and Securitization — IMF Working Paper 07/117 (2007) | The original compliance foundation; its §II definition grounded the first screen |
| §65 | Muhammad Ayub, Understanding Islamic Finance — Wiley Finance (2007, 544pp) | The foundation source, displacing §28. Derives the prohibitions from primary texts, names the juristic tests, and cites OIC Fiqh Academy and AAOIFI resolutions |
| §66 | Four trading-specific papers on bay' al-sarf, qabd and forex fiqh — including Syed Faiq Najeeb (INCEIF) | Exchange-specific rather than banking-specific; the sharpest-focused material in the base |
| §67 | Al-Jarhi, Abuzaid & Oweida, Handbook of Islamic Finance — Ankara Social Sciences University (2022, 322pp) | The 'ayn/dayn distinction the backing axis rests on |
| §71 | IIFA Resolution 237, the SRB material, and the IIUM literature review | Sought specifically to close the crypto gap below |
| §72 | Four academic papers on cryptocurrency and Shariah — the māl/thaman question | The first direct engagement with whether crypto is Shariah-recognised property |
The field's highest authority has expressly declined to rule on the central question. IIFA Resolution 237 carries a heading reading "Second: Shariah Ruling" — and there is no ruling under it. The Academy identified the foundational questions as unresolved. It has ruled on electronic possession, which is precisely why the qabd test behind rail 17 is well-grounded; it has not ruled on whether crypto is Shariah-recognised property. The IIUM piece is a literature review and issues no ruling of its own.
+An earlier note in the record puts it more bluntly: at the time, not one compliance source addressed crypto directly — and silence is not permission. §71 and §72 were sought specifically to close that, and they narrowed it rather than closing it.
+So keel does not claim the question is settled. It enforces what is settled — possession, riba, sector, backing, contract class — and records the operator's own attestation, attributed and sourced, for what is not. That is the whole design: keel is an enforcement engine for a ruling you supply, not a source of rulings.
+Asset sweep
Venues
Three venue expansions were studied and three answered no. Two are permanent; one is "not yet".
+Four venues have been assessed beyond Coinbase spot. None is live, and the reasons are not interchangeable — two are permanent refusals on the instrument, one waits on data, and one is built and deliberately unwired.
| Trade Nation | Refused | Nothing in the product range is spot ownership — CFDs and spread bets throughout. |
| US equities | Not yet | Coinbase does have an equities order path — the brief was wrong. But market data is absent: n=0 candles across 21 trading days, preview refused 403, three conflicts with our execution model. |
| Coinbase futures / perps | Refused | Rails 18/19 reject any non-USD-settled-spot contract mechanically, regardless of narrative. |
| Robinhood | Built, unwired | The only venue with a working adapter — keel-broker-robinhood ships translate, transport and adapter layers behind the broker port, with its own test suite. It is deliberately unreachable from live: the executor still calls Coinbase's signature directly, _common.py constructs a CoinbaseClient, and nothing calls load_broker(). |
The distinction matters for planning: a CFD venue is never, because the instrument conflicts with the charter. Equities are not yet, because the blocker is data and execution capability, both of which can change.
+The gap is a property of the venue rather than a bug to fix: Robinhood's market orders accept only asset_quantity — there is no quote-amount market order. keel opens every position with MarketIOCByQuote, sizing in dollars rather than coins, so under the current entry model Robinhood cannot open a position at all. The adapter is honest about it: it declares market_ioc_quote unsupported and raises rather than approximating.
That is why the port work is finished and the wiring is not. Issue #198 exists so the migration trips over this deliberately, on the bench, rather than discovering it with real money.
+ +The distinction matters for planning. A CFD venue is never — the instrument conflicts with the charter. Equities are not yet — the blocker is data and execution capability, both of which can change. Robinhood is not like this — it needs either a different entry model or a venue-side capability that does not exist today.