Output
alpha_simulate, alpha_submit and alpha_get return one JSON object. There is no daily series.
| Field | Meaning |
|---|---|
id |
Library id; null until the alpha is submitted and saved. |
family |
Your name for the idea. Versions count within a family. |
version |
Library version; null until saved. |
expression |
The canonical text form of what was tested. |
grade |
held_up (Awesome Idea) / leans_yes (Good Idea) / no_sign (Not sure) / cant_test_yet (Impossible to verify). |
passed |
True when every submission check passed. |
trial_no |
How many distinct versions of this family you have tested. The bar rises with it. |
as_of |
The day the test ran. |
data_end |
The last day of data in the test. |
report |
Whole-period numbers and the same numbers for each calendar year. |
checks |
The submission checks, each with passed, value and limit. |
breadth |
Effective number of names held on an average day. |
curve |
Up to 250 [date, equity] points starting at 1. Present only when breadth is at least 5. |
Report numbers
The report and each entry of report.years carry these numbers. report also carries delay, rebalance and years; each year carries year.
| Number |
|---|
days |
sharpe |
annual_return |
turnover |
max_drawdown |
fitness |
margin_bps |
sharpe is annualised over 365 days, or 252 weekday sessions for a stock universe. turnover is the mean daily traded share of gross, per session for stocks. fitness is sharpe times the square root of abs(annual_return) over max(turnover, 0.125). margin_bps is net return per unit traded, in basis points.
Simulation
History starts 2000-01-01 or at the first day of data. The 365 days before the current quarter are a sealed holdout. Past cases are 30-day blocks. Perps fill at the 02:00 UTC hourly average after the close, stocks at the next open. Returns are net of fees, spread, impact, borrow, margin and perp funding.