Posted by thewizard17 on 8/30/2014 8:23:00 PM (view original):
I know this has been brought up briefly a short time ago, but wanted to go off on another tangent with it.
Often times I'll notice that RPI and the projection report are way off, sometimes by as much as 25-30 ranking points. For instance, a team with a 45 RPI but #75 on the projection report. It baffles me sometimes, because that same team appears to have a lot of similarities as the where they are with other teams in comparison to the #45 on the projection report.
What exactly is the result of such a huge discrepancy? Is it lack of wins vs Top 50 RPI? vs Top 100 RPI? From my observation, it appears like, what your team has done in the last 10 games is almost nonexistent or a very very small percentage of the equation. Having an excellent record 23-3 vs. 15-11 seems to make a difference as well. Any thoughts?
as fd said, there are a lot of different factors. one of the most common reasons i see for RPI being much higher than scheduling report, is that the typical RPI boosting conditions are in play... losses against high quality teams, and wins against high win teams that actually aren't very good. these things most commonly come out in the record vs top X rpi, as you suggested. i always start there, and almost all the time, that is the reason the projection report shows much lower.
in my opinion, RPI is a pretty crappy indicator of the quality of a team, and is easily fooled. the projection report is much better. if i wrote it myself, i'd be reasonably satisfied, which is pretty surprising to me, i had much lower expectations. the projection report is obviously not perfect - but creating a ranking system that correctly ranks teams across a wide spectrum is very difficult, and i think it does a pretty good job. the heavy weighting on record vs top X rpi is a decent way of ranking high end teams, i don't like using such wide, imprecise buckets (which is one area i'd target for improvement), but it does go a long way to make up for the weaknesses in both the ranking system (which over values record) and rpi. for rpi, keep in mind, it doesn't care who you beat! if you play the #1 and #100 teams, you have the same rpi if you go 1-1, regardless of which team you beat (assuming both games are home or away). to me, that is kind of stupid, its one of the major weaknesses in RPI. i do think RPI is better than record, however - and RPI gives credit for strength of schedule off opponent record - while the projection report gives it based on rpi, in the record vs top X rpi pieces, and also some against record, as i believe RPI is a direct factor in the projection report. there are other reasons i think the projection report is better than RPI, but hopefully that gives some insight into the big things...