drewscottt/cfb-rankings

Computer ranking system for NCAAF FBS teams. Only final score and home field is considered. Teams are ranked using a score based on comparative margin of victory.

cfb-rankings's Issues

Feature interaction: ignoring top/bottom n game results and recency bias

When top/bottom n games are chosen to be ignored, should the recency bias be applied before or after filtering out these games?

"Predictions" based solely on relative ranks; should incorporate home field

Difficult task because even if I determine a reasonable home field advantage score, I have to apply it to something. While the rankings are obviously based on a numeric metric, I'm doubtful that applying the home field advantage to that metric would be "correct". The ranking metric, while it does account for point differential, also account for wins/losses.

Is it problematic that the rankings treat James Madison and Blue Bloods (Alabama, OSU, Georgia) similarly?

If JMU beats down teams just like Alabama does, then how can we know that JMU shouldn't be ranked similarly to Alabama?

There's already some notion of opponent strength being used, so I'm not sure what could/should change.

Parsing games from ESPN team page issues

1. In neutral site games, both teams appear as the home team on their respective team page
2. If two teams play twice to the same result, it will only be treated as one game. Should apply a game identifier like date or ESPN game id.

All FCS opponents viewed equally

Should incorporate some sort of methodology to treat FCS teams different (e.g. NDSU is different than Idaho State...)

• Include FCS teams in ranking
• Incorporate previous season ratings to weigh FCS teams
• Incorporate more recent ratings (current FCS top 25) to weight FCS teams