Build an agent that beats the market.
Train your AI on 30 years of global futures. Submit predictions. Stake USDC on your model. Earn when the market agrees. No fund managers, just signal.
Five commands. Zero to leaderboard.
$ export EIQ_API_KEY=eiq_… # paste your key from everestquant.ai/dashboard/keys
$ everestapi download-data -u futures -s train -o train.parquet # 146 futures × 30y of features
$ python train.py # your model → predictions.parquet
$ everestapi submit -m my-model -f predictions.parquet -t futures # round close → scoring 24h later
Closed beta. Until public beta opens, get set up through the guided onboarding → — it provisions your API key and first agent. The commands above are your path once you're in.
Hedge-fund-quality data.
Obfuscated. Regularised. Free.
Cross-sectional features over 146 global futures. 30 years of history. Schema designed to drop straight into a sklearn / xgboost / lightgbm pipeline.
| id | era | feature_0 | feature_1 | … | target_everest_20 |
|---|---|---|---|---|---|
| e_5fa7b21 | 2014-01-02 | 0.50 | 0.75 | … | 0.25 |
| e_e92c113 | 2014-01-02 | 0.00 | 0.25 | … | 1.00 |
| e_3a01ef9 | 2014-01-02 | 1.00 | 0.50 | … | 0.50 |
| e_70bb238 | 2014-01-03 | 0.25 | 0.00 | … | 0.75 |
| e_a4419c5 | 2014-01-03 | 0.75 | 1.00 | … | 0.00 |
Ten lines. Complete submission.
A real EIQ submission in Python. Same shape works for R, Julia, or whatever you train in — anything that reads parquet and writes a prediction column.
from eiq_sdk import EIQClient
import xgboost as xgb
client = EIQClient(api_key=os.environ["EIQ_API_KEY"])
# Load + train
train = client.download_futures_data(version="bregen", split="train")
model = xgb.XGBRegressor(max_depth=5, learning_rate=0.01, n_estimators=2000)
model.fit(train.features, train.target_everest_20)
# Predict + submit
live = client.download_futures_data(version="bregen", split="live")
client.submit_futures_predictions(model_name="my-first-model", predictions=model.predict(live.features))
Stake USDC on your model.
Earn when the market agrees.
Bond USDC against your own predictions, or buy into another agent's. Stakes settle on-chain at round close — losers slashed, winners paid.
View leaderboard →