RPI Projections - MBB

Submitted by mgowill on January 31st, 2018 at 12:04 AM

I was messing around with the RPI Wizard found here -


Interesting tool as you can project wins and losses.  Just for fun I put Michigan in as winning out and it reported an RPI of 16 if that happened.  Sure would be nice, but it also predicts that chance at 3.54%.  One loss is a 15.27% probablity, two losses a 29.36% probability, three losses at 29.35% probability, and four losses a 16.15% probability.  The following are the remaining games followed by the percentage chance at a win.  I also ran the simulator for each single team as a loss so you could see which games have the biggest hit to RPI.  No surprise that the home losses were the biggest hit.  I did find it odd that the highest ranked team left on our schedule caused the biggest hit to our single loss RPI projection (OSU), but I can't really speak to how the wizard calculates its data.

Minnesota (H) 24 79%
Northwestern (A) 22 56%
Wisconsin (A) 21 62%
Iowa (H) 24 84%
Ohio State (H) 25 63%
Penn State (A) 21 53%
Maryland (A) 21 45%

Then I ran a couple two loss scenarios, and a handful of three loss scenarios.  When I get more time I will run some more scenarios to get a better picture.  What I came up with in the few I ran was pretty much defend home court, finish at least 2-2 on the road, and we're looking at a 7 or 8 seed in the tourney.



January 31st, 2018 at 12:53 AM ^

I would agree.  Many of the scenarios I ran for two losses had Michigan around the 28-32 range depending on opponent.  Most of what I ran didn't look bad for one of the losses being at Maryland - since the computers have us as a 1.5 point underdog.  FWIW - Michigan is favored in all remaining games except for that one. It would sure be nice to only drop one remaining game though.  That would most definately put us in the 6 seed range.


January 31st, 2018 at 2:02 AM ^

and get to the BTT semis (which would almost certainly require a win over Nebraska), we'd be at worst a 5, possibly even a 4 seed.

We'd be top 20 RPI and the committee doesn't just seed directly by RPI.  We'd have won 18 of our last 23 games with 3 (!!!) of the losses being to Purdue, and that's if we lose to them again in the BTT semis. Could certainly sqeak into a sweet 16 seed by going 6-1 the rest of the way, almost certainly if we go 7-0 to close.


Whole Milk

January 31st, 2018 at 9:52 AM ^

I don't know if this is necessarily true. I think everyone recognizes that the Big 10 is not deep, therefore creating the very realistic opinion that we probably will only get 4, maybe 5 teams, in. But I think there is a respect for the best teams in the B1G. Purdue and Michigan State get all kinds of national attention and love, and Michigan and OSU are certainly getting more respect on the national landscape. I would not be shocked to see all 4 as 5 seeds or better, without a single other B1G team being included.

Perkis-Size Me

January 31st, 2018 at 10:31 AM ^

Largely I agree with you, but MSU could still get a 1-2 seed depending on how they finish out the year. Outside of the Purdue game, which they get at home, and playing at IU, which is always a tough place to play, the rest of their schedule is a breeze.

If they win out, which is a very distinct possibility, that puts them at 28-3 for the year. And depending on what they do in the BTT they would be knocking on one seed territory. MSU has historically done very well in the BTT. If they win it and upset Purdue again along the way, they're definitely a one-seeder. 

But yeah, as far as everyone else goes, OSU may end up in 4-5 seed territory, we could go anywhere between 5-9, and then the rest of the conference is pretty much on life support as far as making it in. 


January 31st, 2018 at 4:10 PM ^

strength.  Not one bit.  And yeah, the big ten isn't great but we will end up playing a pretty difficult conference schedule: we play Purdue and OSU twice, which are the top two teams.  We play Maryland and Northwestern twice and they're both middle-of-the-pack teams.  We only play the bad teams once, except for Iowa.  So our SoS isn't hurt by the conference being bad, and if we go 13-5 against a tough conference schedule, that's going to look very good.

We will absolutely be a 5 seed at worst if the scenario plays out as I detailed.

And what I expect for other teams in the conference isn't great.  MSU for instance, I don't see how they're better than a 4 seed right now, but all the projections say 2.  They have played an awful schedule.  The unbalanced B1G schedule only gave them one game against Purdue, M, OSU, and Nebraska.  They got all the bad teams twice (Rutgers, Wisconsin, Illinois).

The played a weak non-conf schedule.  They lost most of their tough games (Duke, M, OSU), and their best win over UNC is looking worse and worse.  They'll be very interesting come selection time because the only way they get better than a 4-seed will be if the committee weights kenpom/talent/margin of victory against bad opponents higher than one would expect.

I also think Purdue has to win out to get a 1-seed because of their iffy non-conf losses and relatively easy schedule.

I think OSU is a 5 seed right now, at best.  They've been really good but don't have a decent road win.  They need to beat Purdue or M (and take care of business otherwise) to get into the top 4 seed line.

So no, I don't expect those teams to be seeded as highly as you might think, but Michigan will end up with a great resume (a 5 or 4 seed resume) if they finish 6-1 and get to the BTT semis.


January 31st, 2018 at 2:44 AM ^

I didn't neg you but what you're saying simply isn't true.  RPI, ESPN BPI, Kenpom, and Sagarin ratings are all used in comparison to the resume.  While it is true that RPI is not the sole determining factor, it is used in part of their formula.  It also gives you an in season barometer for where your team stands.  It seems to work in regards to what has been posted per computer projections so far.  It just makes for interesting banter with 7 games left to play.


January 31st, 2018 at 4:24 PM ^

because the team sheets the committee uses categorizes a team's games into four buckets of difficulty based on RPI (unfortunately, they're not bucketing them by a composite ranking). I think those records in games in each bucket are heavily used (that's what they've usually cited in the past to justify seedings).

So while they get info on the other ranking systems, RPI is still prominent. Also, 3 of the 6 rankings they use (RPI, KPI and one of the others) are simply W/L based, not margin of victory so the other two usually aren't much different than RPI.

That said, the new bucketing system should help us this year.

AA Forever

January 31st, 2018 at 6:59 AM ^

And we would probably have to close out at least 5-2, and win two games in the BT tournament.  We have the same handicap that we had last year, and in other past seasons...too many games against 300+ rated RPI teams.  There's no fixing that, and it will probably knock our seed down one from what the more optimistic folk here are projecting.


January 31st, 2018 at 12:50 AM ^

is essentially negligible, and my guess is that it's just a 3rd decimal quirk in the result of the formula.

This is actually an interesting example of how the RPI works.  It doesn't really care who you beat or lose to, it only compares your overall record to your opponents records and their opponents records with a different weight for road, neutral and home games.

That's why losing any of our home games results in the same 24 or 25 ranking, and losing any of our road games results in the same 21 or 22 rating, regardless of opponent (apparently the formula penalizes you more for losing a home game than it rewards you for winning a road game, which is actually surprising to me).


January 31st, 2018 at 12:59 AM ^

Other than Ohio State at home, there really aren't any more resume building wins left.  I would also guess that because the presumptive calculations are that tight, we are who we are.  Everyone else is pretty much the same.  This far into the season I would guess the computers have more than enough data to cause so little variation.  It was interesting to mess around with though.


January 31st, 2018 at 1:59 AM ^

of variation when you vary our record.  But my point is, given an assumed record, the only thing that matters is where you play your wins and losses, not who you play.  And that's because the formula inputs are an overall record weighted for location, opponents records and opponents opponents record.

The formula doesn't even know who you beat or lose to.  If we go 1-1 in any two game stretch, it doesn't matter who we beat to the RPI.  Our record is 1-1 and our opponents and opponents opponents records are the same regardless of which one you beat.

And I just realized the likely case for OSU is that we'd move ahead of them if we beat them, but we'd be behind them if we lose to them. Our RPI score (which is a decimal) would be identical regardless of which home game we lose, but under two of those scenarios you're assuming a loss for them (which would put them behind us) and under the scenario in which we lose to them, they'd remain ahead of us.

TL;DR if we are certain to lose one game, we don't want to lose to them because they'd stay ahead of us (assuming all their other expected outcomes occur).

For Northwestern, it's possible that by giving them a win against us that bumps one of their opponents up (Creighton or Texas Tech perhaps) just enough to move/stay ahead of us (since that would improve their opponents opponents record by a win more than the other scenarios).  Again the margins between spots in RPI in this range are very small and are affected by a lot of games since opponents record and opponents opponents record are all inputs (hence if you change one outcome it can slighly change a lot of teams RPI formula result which is why I previously referred to a 3rd decimal quirk).


January 31st, 2018 at 2:16 AM ^

This is correct.  Additionally, you're correct that it's more detrimental to lose a home game than it is to win a road game.  This is exactly the opposite of the result that they were attempting to induce, but they didn't think it all the way through.

Basically, all of the teams under consideration for the tournament are going to have records > .500.  If you have one home and one road game, and you're going to go 1-1 in the two of them, your record for RPI purposes will increase by either 0.6-0.6 (if you win at home and lose on the road) or 1.4-1.4 (if you win on the road and lose at home).

Since the team's record will be over .500, the 0.6-0.6 will drag it down (toward .500) less than the 1.4-1.4 would.

Whole Milk

January 31st, 2018 at 9:59 AM ^

I think @ Maryland would be a pleasant addition to a resume. I wouldn't be absolutely shocked to see them go 5-2 down the stretch here leading up to us and then we are playing a 20-10 team on the road to finish up. Other than that, you are right, not much else on the docket for a solid win.

Perkis-Size Me

January 31st, 2018 at 1:56 PM ^

Well we'd have to factor in the BTT as well, as there are definitely opportunities for resume-building wins if we make it to the semis or even the title game. I know that's more of an unknown since we don't know who we'd be facing, but if we drew, say, MSU or Purdue, and managed to beat them, that could definitely push the team up a seed. 


January 31st, 2018 at 3:08 AM ^

I see no scenario where that would be the case. We keep winning, we crawl to the 6 line (5 if we knock off Ohio St. and do decent in the B1G Tourney). We go only .500 and we stay an 8. Neither of those scenarios have us bumped up TWO seed-lines just because we played "Bama the VI".

Yes that game sucks to play but it isn't costing us at this point with the wins we have.


January 31st, 2018 at 5:28 PM ^

(per teamrankings.com). A BTT title is irrelevant beyond who you beat to get there, which is all baked into that 25 RPI.

A 25 RPI translates to a 7-seed.  Yes, the committee should have factored in the way were playing coming into the tournament, etc, etc, but a 7 seed wasn't egregious.

And for counter examples, we were 11th in RPI in 2014 but got a 2 seed. We were 19th in 2013 (with a very poor finish to league play and a poor showing in the BTT), but got a 4 seed.

It's recency bias to take last years underseeding and assume it's always the case for Michigan.  The committee has shown plenty of instances in which it seeded us better than our RPI would suggest.


January 31st, 2018 at 7:24 AM ^

Not sure why this year would be different.



January 31st, 2018 at 8:34 AM ^

And that is Brian's opinion.  I do think Michigan was one seed too low last year, however several Big Ten teams were underseeded despite impressive resumes due to the league being garbage.  Considering most projections only have four teams in (with Maryland and Nebraska on the cusp of the bubble) I would expect the conference to once again not get much love with respect to seeding.


January 31st, 2018 at 10:16 AM ^

Ah, there’s Maizen, everlasting concern troll. Make a point and pass it off as fact, reference one other person who agrees as if it validates your point, continue saying the point louder and louder, completely ignore any and all evidence you don’t like, also completely ignore things you predicted would happen when pointed out to you, find something new to concern troll about. What a sad troll

UM Fan from Sydney

January 31st, 2018 at 8:00 AM ^

I cannot stand the fact that OSU, with their first year head coach and after the last few shitty teams Matta had, is a good team and threat to win the Big Ten title. They sure as shit have great luck with new coaches.


January 31st, 2018 at 3:24 PM ^

That and they've had a pretty easy schedule so far.  Zero non-conference road games and only 8 out of 18 conference games on the road.


They still have to play at Purdue, at Michigan, at Penn State, and at Indiana.  I bet that'll knock them back down to earth.


January 31st, 2018 at 3:45 PM ^

they have like, what, eight scholarship players and are playing Andrew Dakich nearly half of their minutes!!

And yeah, KBD is back but Holtmann has to deserve a lot of credit for him playing at a BTPOY level.  He had shown potential in the past but not sure anyone thought he could put up the best numbers in his career by a huge margin while also using far more possessions.  That is really hard to do.

Wesson has been really good as a freshman. They're playing much, MUCH better defense than last year and that's not all KBD. Holtmann deserves a ton of credit.

Agree that the cupboard wasn't bare, but whatever he's getting out of his bench is all him.  He's shown excellent coaching chops this year, and he'll have to also show recruiting chops to stay at the top of the conference in the future. He has an ok class coming in but next year will almost certainly be a true "rebuilding" year.

Perkis-Size Me

January 31st, 2018 at 11:42 AM ^

I hope Beilein has had that game circled since December 4th. Even he's got to be sitting there wondering how that game ever got away from them. 

Sold out game on a Sunday afternoon, in a CBS nationally-televised matchup, in a game that could give Michigan a big seeding boost with a win. No reason for the place not to be packed to the gills and for Maize Rage to come out in full force. Need to take those punks down in SOMETHING this year. 


January 31st, 2018 at 2:59 PM ^

We'll lose 2 more games.  Just how she blows.  At Maryland I've thought all year long was a loss and unfortunately one more likely...vsOSU, @Wisconsin (yes they're bad), or @Penn St....are my guess for the 2 losses.  We still end up 23-8 heading into BTT and sitting at a 6/7-seed with a win over OSU.