AutoBrain Sim

A lightweight Python client library for interacting with the WorldQuant Brain platform API. Supports authentication, alpha simulation submission, and result retrieval.

Features

• Multiple authentication methods (direct, credentials file, interactive prompt)
• Submit alpha expressions for simulation
• Poll for simulation completion automatically
• Retrieve alpha details and PnL record sets

Install

pip install autobrain-sim

Requirements

• Python 3.7+
• requests

Install the dependency:

pip install requests

Files

File	Description
brain_client.py	Core client library (BrainClient, SimulationResult)
example.py	Standalone usage example
main.ipynb	Jupyter notebook walkthrough

Quick Start

Authentication

Method 1 — Direct credentials

from brain_client import BrainClient

client = BrainClient(email="your@email.com", password="yourpassword")
client.authenticate()

Method 2 — One-liner login

client = BrainClient.login("your@email.com", "yourpassword")

Method 3 — Interactive prompt (asks at runtime if no credentials are found)

client = BrainClient()
client.authenticate()

Method 4 — Credentials file

Create ~/.brain_credentials (or any path) as a JSON array:

["your@email.com", "yourpassword"]

Then load it:

client = BrainClient(credentials_file="~/.brain_credentials")
client.authenticate()

Credentials priority: direct args → credentials_file → ~/.brain_credentials → interactive prompt.

---

Simulate an Alpha

sim = client.simulate(
    expression="close / ts_mean(close, 20) - 1",
    settings={
        "region": "USA",
        "universe": "TOP3000",
        "neutralization": "SUBINDUSTRY",
    }
)

result = sim.wait(verbose=True)   # blocks until done
print("Alpha ID:", sim.alpha_id)

Retrieve Results

alpha = sim.get_alpha()
print("Sharpe:", alpha["is"]["sharpe"])
print("Fitness:", alpha["is"]["fitness"])

pnl = sim.get_pnl()
print(pnl)

Retrieve Yearly Results

yearly = sim.get_yearly()
for row in yearly:
    print(f"{row['year']}  PnL: {row['pnl']}  Sharpe: {row['sharpe']}  Days: {row['days']}")

Or directly from the client using an alpha ID:

yearly = client.get_yearly("YOUR_ALPHA_ID")

Each entry contains: year, pnl, sharpe, days.

Run Multiple Alphas in Parallel

expressions = [
    "ts_mean(returns, 5)",
    "ts_sum(volume, 10)",
    "-ts_corr(close, volume, 20)",
]

results = client.simulate_parallel(expressions, workers=3)

for r in results:
    alpha = r.get_alpha()
    print(f"{r.alpha_id}  Sharpe: {alpha['is']['sharpe']}")

You can also pass per-expression settings overrides:

results = client.simulate_parallel([
    {"expression": "ts_mean(returns, 5)", "settings": {"universe": "TOP500"}},
    {"expression": "ts_sum(volume, 10)",  "settings": {"universe": "TOP3000"}},
], workers=2)

---

API Reference

BrainClient

Method	Description
__init__(email, password, credentials_file)	Initialize client with credentials
BrainClient.login(email, password)	Create client and authenticate in one step
authenticate()	Sign in and obtain a session token
simulate(expression, settings, ...)	Submit an alpha for simulation; returns SimulationResult
get_alpha(alpha_id)	Fetch alpha details by ID
get_pnl(alpha_id)	Fetch PnL record set for an alpha
get_yearly(alpha_id)	Fetch PnL grouped by year for an alpha
get_recordset(alpha_id, record_set_name)	Fetch any named record set
simulate_parallel(expressions, settings, workers, verbose)	Submit multiple alphas in parallel; returns list of SimulationResult

Default Simulation Settings

{
    "instrumentType": "EQUITY",
    "region": "USA",
    "universe": "TOP3000",
    "delay": 1,
    "decay": 15,
    "neutralization": "SUBINDUSTRY",
    "truncation": 0.08,
    "maxTrade": "ON",
    "pasteurization": "ON",
    "testPeriod": "P1Y6M",
    "unitHandling": "VERIFY",
    "nanHandling": "OFF",
    "language": "FASTEXPR",
}

Any key passed via settings overrides the default.

---

SimulationResult

Returned by client.simulate().

Method	Description
wait(verbose=True)	Poll until simulation completes; returns result JSON
get_alpha()	Fetch full alpha details (call after wait())
get_pnl(poll_interval)	Fetch PnL record set (call after wait())
get_yearly(poll_interval)	Fetch PnL grouped by year (call after wait())

Attribute	Description
alpha_id	Alpha ID string (available after wait())
progress_url	URL used to poll simulation progress

---

Example

from brain_client import BrainClient

client = BrainClient.login()  # interactive prompt

sim = client.simulate("close / ts_mean(close, 20) - 1")
result = sim.wait(verbose=True)

alpha = sim.get_alpha()
print(f"Alpha ID : {sim.alpha_id}")
print(f"Sharpe   : {alpha['is']['sharpe']}")
print(f"Fitness  : {alpha['is']['fitness']}")

Notes

• The client uses HTTP Basic Auth to obtain a session token from the /authentication endpoint.
• Polling respects the Retry-After response header returned by the Brain API.
• Never commit your credentials to version control. Use ~/.brain_credentials or environment variables instead.