This data converts Signals' data in classic format.

The only needed libraries are pandas and your favourite modelling library. The included script uses a sklearn to make predictions.

It combines a few paid data sources and it's pre-processed, feature names are hidden.

Combined stake on models using this data is 218 NMR (100% of my Signals stake)

The GDrive folder includes:

Training data: export_train_{timestamp}.csv (707 features)
Validation data: export_val_{timestamp}.csv (707 features)
Live data: export_live_{timestamp}.csv (707 features)
timestamp suggests last Friday date
Script: baseline_signals.py (~65 lines)
Correlation of features across dates on training data
Also the submission file from the script
covers ~95% of live universe

The model stays unlisted until my predictions are ready. Usually it will be active for listing within 5 hours of round opening. This is to avoid failed on-time submissions.

Thank you!

Diagnostics using the default script + per era correlation:

https://drive.google.com/drive/folders/1BtLxQTfcqN6PCCOoUdxMQCQR93BeFvQd?usp=sharing

1 Day Stake Return	-%
3 Month Stake Return	-%
1 Year Stake Return	-%
All Time Stake Return	697.54%
NMR Stake	183.40 NMR

FNCV4 Reputation	0.0001
CORRV4 Reputation	0.0001
ICV2 Reputation	0.0000
RIC Reputation	0.0000
TC Reputation	-

FNCV4 Rank	1456
CORRV4 Rank	1429
ICV2 Rank	1459
RIC Rank	1298
TC Rank	-

PS99

Description

Featured by seller

PS99

Model Metrics