- This topic has 11 replies, 5 voices, and was last updated 7 years, 10 months ago by
.
-
Posts
-
The week we’re going to focus on the teams clustered on the NCAAT Bubble. If you missed last night’s game in Blacks[See the full post at: ACC BB Update]
Thanks again for the terrific work, VaWolf.
Are we completely sure the RPI is used by the committee as the important tool that it once was? And if so, why? It is quite apparent to most that the RPI has quite a few inherent problems and at the very least should be modified to gain a more complete understanding of an actual power ranking. And this is not directed at you, VaWolf, as you have adequately stated your standards for these entries. My beef is mostly with the use of RPI by the selection committee as opposed to an improved tool.
The seeding is gonna be SUPER interesting. My guess is they will avoid the 8/9 line with our sides, consistent with VA’s demarcation points, but so much can still change. I hate that I forgot to put on the Castrated Turkey game when I got back from yoga class (SHUT UP – yes, I am taking yoga with my high school-age twins, talk about a life experience).
Are we completely sure the RPI is used by the committee as the important tool that it once was? And if so, why?
We know that it is used because it is on the one-page summary that is prepared for each team. The committee has attempted to separate its use of the RPI from the football calculation that used to be used in the BCS. So when I say that it is used, I mean that it is one of a number of factors that are being considered.
That does not mean that if Team A has a RPI ranking of 50 and Team B has a RPI ranking of 60, that Team A is a step ahead in the selection process. When you get to the point of evaluating bubble teams, the RPI ranking of the opponents that Team A and Team B beat is much more important than their individual ranking.
Here’s a quote that I put in last year’s Selection Sunday entry and it came up again in the Dance Card entry here last week.
When Joe Castiglione [Selection Committee Chariman] made his media rounds after the Selection Show to explain the committee’s decisions, he repeatedly mentioned the number of RPI top-50 wins that bubble teams did and did not have. He didn’t mention the records that bubble teams had against RPI top-50 competition. Just the number of wins. And, when you look at the bubble teams that did and didn’t make the cut, it becomes apparent how much of an impact the metric had.
Bubble Teams’ RPI Top-50 Wins Bubble Teams in the FieldNumber of RPI Top-50 Wins
Syracuse 5
Temple 5
Michigan 4
Tulsa 4
Vanderbilt 2
Wichita State 1Bubble Teams Out of the Field Number of RPI Top-50 Wins
St. Bonaventure 3
St. Mary’s 2
Monmouth 1
South Carolina 1
San Diego State 1
Valparaiso 1Another way that the RPI is used is in the seeding of the tournament. Specifically, about 75% of the teams are seeded within one line of what you would predict solely by its RPI ranking. We’ve discussed this before and did so last week on the Dance Card entry.
One last thing, the professors behind the Dance Card have found statistical evidence demonstrating that the old RPI formula (pre-2005) is an important factor in the Selection Committee’s decisions. So it’s not just what the Selection Committee says, we can show that it affects their decisions.
Good summary and explanation, VaWolf. My biggest concern, although perhaps that is not the correct term, is that the RPI continues to be used instead of a better tool. Just looking at the RPI rankings that you provide, I have serious doubts about a number of those. I believe too much weight is given to perceived strength of schedule (again, based on RPI based ratings, which may be flawed to begin with), and no weight given to margin of victory and defeat. The RPI, at least in my opinion, is a very flawed rankings tool. That being stated, there is NO DOUBT of its current and past use by the committee.
respect, BJD.. yoga isn’t as easy as it looks
My biggest concern, although perhaps that is not the correct term, is that the RPI continues to be used instead of a better tool.
This claim is often stated as fact, never proven, and mostly irrelevant for several reasons.
1) If you would back up and read my recent entry on the Dance Card and the comments, both Pomeroy and Sagarin’s rankings are provided to the Selection Committee and evidently Pomeroy’s ranking was used last year in several instances.
2) Since the RPI is not directly used to select the teams, a comparison between the RPI and any other system is pretty much meaningless. You have to compare whatever system you love to the selections made by the Committee, not the RPI since the RPI isn’t solely used to pick the field.
3) The NCAAT selections are reflective of what has happened during the year and not intended to be predictions of what would happen in the tournament. A team earns its way into the field by beating other good teams. In interviews with the Selection Committee, they often quote wins against other teams in the field and/or losses to teams not in the field.
Yes, SOS can skew the RPI calculations…see this year’s WF’s ranking. But point spread can skew the rankings as well. According to foxsports.com, Pomeroy was used last year to pick Wichita St. The problem I have with Wichita St last year is they didn’t play many Top50 teams and only won one of those games. I don’t think that their winning margin against a bunch of nobodies should push them into the NCAAT field over a team that actually beat good teams.
But point spread can skew the rankings as well. According to foxsports.com, Pomeroy was used last year to pick Wichita St. The problem I have with Wichita St last year is they didn’t play many Top50 teams and only won one of those games. I don’t think that their winning margin against a bunch of nobodies should push them into the NCAAT field over a team that actually beat good teams.
Perhaps the right thing would be to incorporate margin of victory/loss but in a manner weighted by strength of opponent. Is there a system that does that?
Or maybe do something like the breakdown of wins/losses vs. RPI (e.g., 1-25, 26-50, 51-100) with margin of victory. In other words, show average margin of victory against top 50 RPI opponents vs. top 100 RPI opponents vs. rest of opponents, or something like that. More data couldn’t hurt, right?
Is there a system that does that?
The short answer is that I don’t know. Sagarin doesn’t discuss the dirty details of his system, so only he could answer that question for his system. I don’t follow either Pomeroy’s rankings or ESPN’s BPI; so I have no idea what goes into either system…but I doubt that they go that far.
Margin of victory matters. It matters a whole lot if you’re trying to make predictions. It could even matter when you are trying to rank teams. But the weighting factor you apply to point spreads, or to SOS, are open to discussion/dispute/disagreement.
What you don’t want is blind allegiance to any one formula. Speaking only for myself, I don’t want a team to get an at-large bid for scheduling a bunch of tough opponents, but never winning any of those games (see WF). But I also don’t want someone to get a bid because they had a impressive point spread against a bunch of teams that were not included in the field (see Wichita St last year).
In the comments to the Dance Card entry linked in this thread, foxsports.com claimed that the Selection Committee chairman said that Pomery was used for Wichita St last year. I didn’t see that mentioned in any other bubble article last year, so I’m a little leery of accepting that as a fact. So I’m going to withhold judgement until I find out for sure that the selection committee is using Pomeroy/Sagarin and how/when they are being used.
If a decision is needed between a bunch of really similar teams, then another ranking system would probably be better than a coin flip. But if people are using a different ranking system without understanding its weakness and strengths…then that’s not too good either.
- You must be logged in to reply to this topic.
