## README.md
# Welcome to Sharps Research
**Note: We are rapidly pushing updates to the servers over the next couple of weeks.
We build AI models and datasets to predict the outcomes of sporting events.
Sharps Research is an internal research framework for sports prediction. It combines
data pipelines, feature engineering, anomaly detection, and interpretability at the
dataset, training, production, and market levels into a single system designed to build,
test, and refine predictive models in real markets.
Our thesis is that interpretability comes first. A system that can expose bias, drift,
and instability will adapt more reliably than one that cannot. This approach produces
models that are stable, comparable, and suited for the highest-volume betting environments.
We do not sell picks, track hit rates, or provide line-shopping tools. Our focus is on
research-grade predictive modeling and the infrastructure needed to sustain it.
Be curious. Bet early lines.
_
## TODO.md
Last Updated Nov 14, 2025
# TODO
[Backlog] zoomable graphs on match page.
[Backlog] kl-loss
[Backlog] Auth
[Backlog] Blog
[Backlog] shap visualization
[Backlog] Model card → /modelcards/
[Backlog] Local Visualizations.
---
1. [Bug/WIP] QA Inference data + elo
3. [Bug/WIP] Push update to model restructuring weighted RAPM's
2. [Bug/WIP] Fix cron
---
[Done] Rewrite DB repository
[Done] Rewrite AI repository
[Done] Nuke old web servers
[Done] cron jobs to update db from stats.nba
[Done] cron jobs for feature engineering
[Done] cron jobs for sharps_rapm, and binary classifier
[Done] man in the middle edits for inference
[Done] OOD inference filter
[Done] Finalize v0 moneyline model
[Done] Communicate site is buggy af until November
[Done] Build TODO for match page
[Done] /market-chart
[Done] push new email.
[Done] Add nba schedule to landing.
[Done] player name to model_inputs
[Done] editable player_name in model_inputs
[Done] /player-strength
[Done] cron for market odds.
[Done] /team-strength
--
x.c.l ensemble.
Upsize database connections/pools
Release to public
Advanced dataset/bias cleaning
Get back to new features.
_