Road Victories – Home Losses

General NCS Basketball16

Last night I was wearing out the remote watching GW vs Xavier, UNC vs UMd, and Smallville. The Xavier game was crucial to keep GW in the top-50 and protect one of State’s two wins vs top-50 opponents. The ACC match-up featured State’s next opponent…and two of State’s most hated rivals. (And you never know if Superman is going to pull through or not. ;-))

The evening ended with two road victories for GW and UNC. Here is a quick lesson in the effect of road wins and home losses on the RPI ranking (from kenpom.com):

Team

2/2/2006

2/3/2006

r

Maryland

25

38

-13

UNC

36

23

+13

G. Washington

45

35

+10

Xavier

49

55

-6

Frankly, I expected Maryland and Xavier to drop farther than their opponents rose….but things are not always as simple as we expect. I think the relative movement of UMD/UNC shows that the actual RPI values of the teams in the top 40 or so must be tightly grouped….such that home losses (counting as 1.4 games) are absolute killers to a team’s ranking. With one home loss, Maryland has essentially moved onto the bubble.

Which is the main thing I wanted to point out. If you have to lose…do it on the road and not at home. It is absolutely essential to pound the Turtles on Super Bowl Sunday!

EDIT

Chief93’s comments spurred me into looking for something official on the weighting factors. I found the following article on ncaa.org

….”The committee adopted a formula that more accurately reflects the historical data regarding a team’s performance at home,” said Bob Bowlsby, director of athletics at the University of Iowa and chair of the committee. Over the past 20 years, home teams have won about two-thirds of all games….

…The mathematical components of the RPI will continue to be 25 percent winning percentage, 50 percent opponents’ winning percentage and 25 percent opponents’ opponents’ winning percentage. The new formula will weigh each road victory at 1.4, each road loss at .6, each home victory at .6 and each home loss at 1.4 in the win or loss column in the RPI. Neutral-site games will be valued at 1.0 in the win or loss column….

About VaWolf82

Engineer living in Central Va. and senior curmudgeon amongst SFN authors One wife, two kids, one dog, four vehicles on insurance, and four phones on cell plan...looking forward to empty nest status. Graduated 1982

General NCS Basketball

16 Responses to Road Victories – Home Losses

  1. Mr. O 02/03/2006 at 9:27 AM #

    We should beat Maryland without Chris McCray. They will be desperate for a win and we haven’t been playing our best basketball.

    Again, you have really contributed some interesting information to chew on.

  2. Chief93 02/03/2006 at 10:23 AM #

    This reminds me now why I think the weighting is too heavy in the “new” RPI. I believe they stated that they chose the weighting based on home teams historically winning about 2/3 of the games played. Therefore, they go with the 0.6 weighting (nevermind that 0.7 is closer to 0.667). However, they also add the an equal offset the other way (1.4 per loss). Now your disparity between a win and a loss is greater than the credit you get for winning. There’s very little credit for winning at home; the only reward for winning is that you didn’t lose–which would be devastating.

  3. RickJ 02/03/2006 at 11:03 AM #

    Amazing effect! I also noticed that Seton Hall has vaulted all the way to 33 with three straight road wins against us, Syracuse & Providence. Weren’t they like 85 when we played them? You can really get well on the road if you win.

  4. VaWolf82 02/03/2006 at 11:07 AM #

    I should have documented GW’s SOS before and after playing Xavier as well. GW’s SOS ranks 213 even after playing a top-50 team. GW’s RPI probably got a boost from X’s record as well as from the road win.

    UNC’s SOS on the other hand probably didn’t change that much after playing UMD…but I can’t say for sure. I still can’t get over how much UNC’s ranking jumped. State doesn’t have any more road games left against teams near the top…so it won’t have another opportunity to make that same kind of jump.

    I haven’t seen any discussion on how the weighting factors were chosen. I’ve always assumed that the Committee had the calculations done and then picked the weighting factor that gave them the effect they were looking for. Maybe I’m giving them too much or too little credit….but it would be interesting to know how the factors were chosen.

    After seeing how large the jumps were, I wish that I had put off the RPI update until today. Oh well, I’ll have to check the Thursday schedule closer in the future.

  5. VaWolf82 02/03/2006 at 11:08 AM #

    I’m positive that Seton Hall was in the 80’s before they came to Raleigh.

  6. Trout 02/03/2006 at 11:12 AM #

    ^ VaWolf: do you have the NCAA’s “offical’ RPI results? They released them yesterday.

  7. VaWolf82 02/03/2006 at 11:25 AM #

    I saw that they were out and read on Ken Pomeray’s blog about some of the differences between his calcs and theirs….but didn’t go through them extensively. Why?

    The RPI changes daily for nearly every team. Unless they are going to do daily updates like kenpom.com…..then their numbers are out of date almost as soon as they are released.

  8. Trout 02/03/2006 at 11:31 AM #

    ^ Well, I’d be interested to see where NC State is at this point. I dont know how often the NCAA will release their RPI. But it is the official one.

  9. VaWolf82 02/03/2006 at 11:41 AM #

    Based on Chief’s question, I added a small blurb and a link to the ncaa release on the weighting change. Chief’s memory was very accurate.

    For the games through 1/30, State ranked #30 on yesterday’s official release. (Their computers must be slower than everyone else’s.)
    http://web1.ncaa.org/app_data/weeklyrpi/rpi1.html

  10. Chief93 02/03/2006 at 12:03 PM #

    Hmm. Thinking about it a different way than I have in the past, maybe this can make sense. If the goal is to give you credit in proportion to your chance of losing the game (i.e. more credit for winning a game that you should have lost–a.k.a. a road game), then maybe it works like this:

    Your chance of losing a neutral-site game is 50-50, so the winner gets 1 of the 2 points “available” for the game in his win column and the loser gets the other point in his loss column. 1/2 = 50%.

    Your chance of losing a road game is apparently about 0.667, so maybe their adjustment of 1.4 of the two points for winning it is close to that: 1.4/2 = 0.7.

    So, I can hand-wave it into making sense from that perspective, but from a risk/reward perspective, the 0.6 credit or 1.4 penalty still sounds extreme for the home team.

  11. Chief93 02/03/2006 at 12:07 PM #

    BTW, that’s exactly where I got the 2/3 number from except I got it indirectly via Pomeroy:
    http://kenpom.com/blog/index.php/weblog/an_rpi_with_road_rage/

    I need to dig up my RPI program and we would be able to revert back to a certain date to get Seton Hall’s ranking prior to our game (the 85 number does sound familiar to me too, though) as well as doing rough projections of the effects of future games.

  12. Mr. O 02/03/2006 at 12:15 PM #

    “I need to dig up my RPI program and we would be able to revert back to a certain date to get Seton Hall’s ranking prior to our game (the 85 number does sound familiar to me too, though) as well as doing rough projections of the effects of future games.”

    Wow. He has his own RPI program. 🙂

  13. VaWolf82 02/03/2006 at 1:23 PM #

    but from a risk/reward perspective, the 0.6 credit or 1.4 penalty still sounds extreme for the home team

    This may or may not be true….I literally haven’t even tried to think about it. It is what it is….I’m just trying to understand what it is. Here’s something to think about it while we’re smaking the RPI formula around:

    The Dance Card using the new RPI formula (that weights road and home performance differently in its first component), correctly predicted 32 of the 34 at-large selections in 2005 (missing on North Carolina State and Iowa State), or 94%. The fact that the original computation of the Dance Card (using the old RPI) was more accurate last year than the formula that used the new RPI as input suggests that the committee stuck closely to past patterns.

    http://www.unf.edu/~jcoleman/dance.htm#SAS.

    The main advantage that I see in the RPI formula is that the results are not blindly followed for either at-large bids or seeding. Now you can argue about the Selection Committee’s decisions, but I find it comforting that someone is at least attempting to do an overcheck of the results. It doesn’t hurt my opinion of the Selection Committee that State has fallen on the right side of the bubble twice. 🙂

  14. Chief93 02/03/2006 at 2:04 PM #

    Those committee dudes watch a LOT of games, and I really think they look past the numbers for the most part (And us showing up on their TV’s on Saturday and/or Sunday of the ACCT can’t have hurt in pushing us to the happy side of the bubble). Maybe the general populace should start believing them when they say the RPI is just one of many tools they use and even then they only use it to compare small groups of two or three teams. Further, even then I would expect they get more useful information from those “Nitty Gritty” reports than they do from the raw RPI listing.

  15. JeremyHyatt 02/03/2006 at 2:41 PM #

    Let’s have some turtle for dinner! YEE HAA!

  16. Jeff 02/04/2006 at 9:16 AM #

    For the record…Seton Hall wsa #85 in the RPI the night that they entered the RBC.

Leave a Reply