Cost model realism
Cost model realism 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
- Go live date
- 2026-08-07
- Schema
- canli.alphac-cost-model-realism.v1
Crypto sleeve
- All taker
- yes
- Book exhausted count
- 0
- Database exists
- yes
- Fills since go live
- 0
- Fills total
- 24
- Modelled taker bps
- 5
- Notional total
- 159576.536982
- Realised commission bps MEASURED
- 5
Slippage bps MEASURED
- Max
- -1.282038
- Mean
- -3.181039
- Median
- -3.328349
- Min
- -4.409572
Note. Recorded against `modeled_price`, so this IS a real execution measurement — the only one in the book. Negative is favourable. All 24 fills are taker and none exhausted the book.
Window
- 2026-07-05
- 2026-07-30
Equity sleeves
AlphaMax
- Commission
- NOT MEASURABLE — no fee column in this fills table
- Database exists
- yes
- Fills since go live
- 1063
- Fills total
- 3075
- Notional since go live
- 748509.470238
Slippage vs decision price. NOT MEASURABLE — the fills record limit_price, a PADDED marketable limit, not the price the decision was taken at. A fill that beats it is beating the padding.
Submit to fill latency seconds
3 further fields. Read them in the artifact.
AlphaTrend
- Commission
- NOT MEASURABLE — no fee column in this fills table
- Database exists
- yes
- Fills since go live
- 98
- Fills total
- 311
- Notional since go live
- 842220.826596
Slippage vs decision price. NOT MEASURABLE — the fills record limit_price, a PADDED marketable limit, not the price the decision was taken at. A fill that beats it is beating the padding.
Submit to fill latency seconds
3 further fields. Read them in the artifact.
AlphaVintage
- Commission
- NOT MEASURABLE — no fee column in this fills table
- Database exists
- yes
- Fills since go live
- 15
- Fills total
- 15
- Notional since go live
- 1028168.623657
Slippage vs decision price. NOT MEASURABLE — the fills record limit_price, a PADDED marketable limit, not the price the decision was taken at. A fill that beats it is beating the padding.
Submit to fill latency seconds
3 further fields. Read them in the artifact.
Modelled parameters
- Crypto taker bps
- 5
- Default half spread bps
- 2.500000
- Equity borrow bps annual
- 50
- Equity commission bps
- 1
- Equity half spread bps
- 3
- Equity taker bps
- 1
- Impact coef
- 1
- Latency addon bps
- 2
Verdict. The one checkable component MATCHES exactly. The rest is not checkable from what is recorded, and the latency component is modelled as the wrong kind of quantity. No cost parameter should move on this evidence; the schema should.
What cannot be asked and the fix. Slippage against the decision price is not computable for the equity sleeves: the fills record `limit_price`, a PADDED marketable limit, so a fill beating it by 54bp is beating the padding, not the market. Equity commission is not computable either — there is no fee column. THE FIX IS ONE FIELD: record the reference mid at `submitted_ts` on every order. That single addition turns implementation shortfall from unanswerable into a daily measurement, and it costs nothing to capture at the moment the order is built.
What matched. Crypto commission. Measured 5.00bp against a modelled taker fee of 5.0bp — exact. The one cost component in the book that can be checked against reality checks out.
⚠️ crypto has not traded since go live
- Fills since go live
- 0
Last fill window
- 2026-07-05
- 2026-07-30
Note. Zero crypto fills since the 2026-08-07 re-baseline, with the last on 2026-07-30, against a weekly rebalance cadence over a fifteen-day record. The sleeve is marking equity but not trading. Surfaced, not diagnosed — it may be a no-trade band holding, and it may not.
⚠️ the latency finding
- Median submit to fill hours
- 5.515539
- Modelled latency addon bps
- 2
Why this matters. Median submit-to-fill latency across the equity sleeves is about 5.5 HOURS, not seconds — orders are submitted after the close and fill at the next open. The cost model represents latency as a flat 2bp add-on, which is a microstructure quantity. An overnight gap is not a spread: it is unhedged exposure to whatever happens between submission and fill, and its cost is a distribution with a fat tail rather than a constant. This is not evidence the model is wrong by a specific amount — it is evidence that this component is modelled as the wrong KIND of thing, and nothing here measures its size.
Check it yourself
Every figure above is read from /glassbox/cost_model_realism.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.