Top AI Repos — open-source AI, indexed and scored
Top AI Repos tracks AI repositories on GitHub and answers two different questions about each one: is it moving right now, and would you bet a product on it.
Top AI Repos tracks AI repositories on GitHub and answers two different questions about each one: is it moving right now, and would you bet a product on it.
Machine Learning Model for Sport Predictions (Football, Basketball, Baseball, Hockey, Soccer & Tennis)
| Date | Stars |
|---|---|
| 2026-07-24 | 282 |
| 2026-07-25 | 282 |
| 2026-07-28 | 283 |
| 2026-07-30 | 283 |
| 2026-08-06 | 283 |
Today
— stars today
This week
— stars this week
This month
— stars this month
Momentum
0.0
growth rate 0.00%/day
============================================
Sport Game Outcome Prediction Project - Bet on Sibyl
======================
### Bet on Sibyl in a nutshell
BetonSibyl is a platform controlled by a set of algorithmic models (a model defined for each sport)
that projects accurately estimated results (predictions of upcoming games) from a multitude of statistical variables.
At launch, the platform will cover the four major US sports (Football, Basketball, Baseball, Hockey), Soccer and Tennis.
Moreover, the models provide stats to measure the
performance of the algorithm for the current season (for each sport) along with bankroll comparaison stats with bookmaker
(scraping from the [oddsportal's website](http://www.oddsportal.com/) to do so). Here's below an image from the mobile app
propotype based on the platform.
[](https://marvelapp.com/31dgaj4/screen/17384930)
## Table of content
- [Data collection](#data-collection)
- [Decomposition - Data Design Decisions](#data-design-decisions)
- [Web Scraping - Selenium/Beautiful Soup](#web-scraping)
- [Predictions](#predictions)
- [Data Preprocessing](#data-preprocessing)
- [Algorithm Tuning and Running](#algorithm-tuning-and-runing)
- [Results Presentation](#results-presentation)
- [Bookmakers comparaison over the year](#bookmakers-comparaison-over-the-year)
- [Model Performance Metrics](#model-performance-metrics)
- [ML one-sport process in a nutshell](#ml-one-sport-process-in-a-nutshell)
- [License](#license)
- [Links](#links)
- [Notes](#notes)
## Data collection
### Data Design Decisions
1. It is decided that the two participating teams/players in each matchup are respectively represented by visitor team and home team (player A/player B in case of Tennis).
This is contrary to another popular method of representing the teams as the favorite and underdog.
2. The point differential is chosen to be positive when the home team scores more points than the away team.
3. To represent the difference between the two teams that are playing in the matchup, the ratio or the difference between
the same attributes are taken between the two teams. That is, the home team’s statistic is divided by the visitor team’s
statistic. Therefore, when attributes are a positive indicator of performance, a value greater than 1 indicates that
the home team performs better for that particular attribute. When the attributes are integers (not statistics), then
the difference between the home team’s attribute and the away team’s attribute is taken. Examples of this are win
streak, compared to statistics such as points per game that would be compared by taking the ratio.
### Web Scraping
The data is scraped from several websites according to each sport using Python and the Selenium and BeautifulSoup (only for MLB data)
packages for Python. Data sources for each sports are described in the "Link" section.
According to each sport/league, the script goes through each summary season page and writes season team stats and season game stats
(e.g. date, the home team, away team, home team points, away team points etc.) to a csv file.
Game stat data, team stat data, and datetime data are merge later into a feature file (.npz) for the ML algorithm (Lasso Logistic Regression)
Here's below an example team stats data and game stats data for the NFL league.
In the same scripts, After having obtained the raw data set, data is cleaned throughout the script. The script
checked the completeness and validity of all the data files, and eliminated any CSV parsing errors
or erroneous data values.
ex of NFL team stat data from the 2000 season to the 2015 season
ex of NFL game stat data from the 2000 season to the 2015 season
## Predictions
### Data preprocesExcerpt of 13,020 characters
Read on GitHubWould you bet a product on this? Bounded 0–100 and slow moving.
matched fp:71715a1feacb6ef4, topic:web-scraping, readme:web scraping
matched fp:71715a1feacb6ef4, topic:selenium