Quick RPI Tutorial

[ValpoHoops]

Quick RPI Tutorial

The RPI is a tool that the NCAA Selection Committee uses to determine at-large bids and seeding for the NCAA Tournament. It is not the only tool they use, and many believe that it is being used less and less as years pass and more analytics are available.

It is a measure that is based simply on winning percentages of your team, your team's opponents and your team's opponents' opponents. The last two of these is where the strength of schedule is calculated.

Only games against Division 1 teams are included in the calculations.


The Math (Simple Version) – See below for why some numbers may not always match:
25% of team winning percentage
50% of opponents winning percentage (games against your team are excluded in this area only)
25% of opponents' opponents winning percentage

Thus, if your team is 8-2, their winning percentage is .800. Multiply that by 25% and you get .200

If your opponents are all 7-3, their combined record would be 70-30. Subtract 8 losses and 2 wins because you take out results against your team, and that record changes to 68-22, good for a winning percentage of .756. After multiplying this number by 50%, you get .378.

If all of your opponents' opponents are 4-6, this is a total of 400-600 (ten games, ten teams, ten opponents). The winning percentage of these teams is .400, of which you take 25%, good for .100.

In this scenario, .200 + .378 + .100 would give your team an RPI rating of .678.

Notice in this example (and any you could create), that it DOES NOT MATTER which teams your team beats or their opponents beat, only the winning percentages. So, it's not a matter of who you beat for the RPI, just how many. (The who comes into play when you get their winning percentages factored in). Thus, there is no such thing as a "win that will really help our RPI".

Understand that my saying "does not matter which teams" only applies to their own schedule (ie: 15-15 is the same winning percentage, regardless of which of their games they win and lose). It ALWAYS matters how good the teams you play are.


The Math (Complicated Version Details):
The formula that the NCAA uses has a few tweaks to it. Over time, research has shown that the home team wins about 70% of the time, so the NCAA chose to adjust their actual formula.

Home Wins: .6 wins
Home Losses: 1.4 losses

Road Wins: 1.4 wins
Road Losses: .6 losses

Neutral site games did not change.

I won't go back through an entire example, but you can see how it would affect a team's record. To the NCAA RPI formula, Valpo has a 23.4-3.8 record this season.

Edit: Added phrase in bold type

[wh]

Notice in this example (and any you could create), that it DOES NOT MATTER which teams your team beats or their opponents beat, only the winning percentages. So, it's not a matter of who you beat for the RPI, just how many. (The who comes into play when you get their winning percentages factored in). Thus, there is no such thing as a "win that will really help our RPI".

If we beat Auburn early in the year and Auburn beats a 30-0 Kentucky team, that's 30 more wins and only 1 loss that goes into the opponents-opponents equation. You're saying that's "not a win that really helps our RPI" as compared to beating a 30 loss Maine team?

[ValpoHoops]

Notice in this example (and any you could create), that it DOES NOT MATTER which teams your team beats or their opponents beat, only the winning percentages. So, it's not a matter of who you beat for the RPI, just how many. (The who comes into play when you get their winning percentages factored in). Thus, there is no such thing as a "win that will really help our RPI".

If we beat Auburn early in the year and Auburn beats a 30-0 Kentucky team, that's 30 more wins and only 1 loss that goes into the opponents-opponents equation. You're saying that's "not a win that really helps our RPI" as compared to beating a 30 loss Maine team?

That's exactly what I'm saying. Had we played Auburn, those 30 wins still show up on our opponents record, so it doesn't matter whether we had beaten them or not.

Yes, exchanging Auburn for Maine would help, but part of who you play, but not who you beat. You're confusing two similar statements.

It doesn't matter who you beat = your winning percentage is the same, no matter which of the teams on your schedule you beat and lose to (home/road games obviously affect this, but pretend it doesn't). Do not confuse this with the next one.

It doesn't matter who you play = your winning percentage is the same regardless of who you play if your record is the same...the teams you play WILL ALWAYS affect parts 2 and 3 of the formula. Thus, this is not an accurate statement.

[valpo4life]

So, instead of saying "this will be a good win for our RPI" we should be saying "this will be a good game for our RPI". Is that a safe to say?

[ValpoHoops]

So, instead of saying "this will be a good win for our RPI" we should be saying "this will be a good game for our RPI". Is that a safe to say?

That would be correct. The strength of schedule bump for playing a fictional game against Kentucky tonight would show up win or lose, while the SOS drop from playing 2-27 Grambling would hurt either way.

[sliman]

And so, our loss at Missouri "helps" us because they also lost to Kentucky (and nearly everyone else in the SEC).  And all this time we've been upset by it.   

[valpo4life]

And so, our loss at Missouri "helps" us because they also lost to Kentucky (and nearly everyone else in the SEC).  And all this time we've been upset by it.   

No.

[ValpoHoops]

And so, our loss at Missouri "helps" us because they also lost to Kentucky (and nearly everyone else in the SEC).  And all this time we've been upset by it.   

No.

No, indeed. Missouri losing hurts our RPI far more than the win their opponents pick up.

Our opponents have played about 900 games. Our opponents' opponents have played about 27,000. Plus, part 2 of the RPI is worth twice as much, which means every Missouri loss hurts part 2 about 60 times as much as it helps part 3.

[wh]

And so, our loss at Missouri "helps" us because they also lost to Kentucky (and nearly everyone else in the SEC).  And all this time we've been upset by it.   

No.

No, indeed. Missouri losing hurts our RPI far more than the win their opponents pick up.

Our opponents have played about 900 games. Our opponents' opponents have played about 27,000. Plus, part 2 of the RPI is worth twice as much, which means every Missouri loss hurts part 2 about 60 times as much as it helps part 3.

Now substitute Oakland for Missouri and you can see what an RPI cancer they are to the rest of the HL by playing teams in OOC play that they can't possibly beat:

Oakland losing hurts our RPI far more than the win their opponents pick up.

Our opponents have played about 900 games. Our opponents' opponents have played about 27,000. Plus, part 2 of the RPI is worth twice as much, which means every Oakland loss hurts part 2 about 60 times as much as it helps part 3.

Now consider that unlike Missouri, Oakland plays each HL team not once, but twice, and 3 times in in the case of UIC. They are a self-serving league RPI killer.

[justducky]

This tutorial has brought my thinking into a sharper focus. No longer is geographical convenience, rivalry continuation, and maybe even maximum ticket sales the prime motivator. Now everything takes a back seat to the resulting RPI considerations.

So if Ball St suddenly signs 5 top shelf recruits (and expect to leap to a top 80 program) will we suddenly get a phone call from them trying to get out of our game at Muncie next year? We already know the Mid-American Conference advises against OOC games that are difficult to win.

We have witnessed an evolutionary process that now routinely features the tail wagging the dog.

[sliman]

Just joking about Missouri, thus the smiley face.  That loss still annoys me as much as the loss at home to New Mexico.

As WH suggests, the goal if we want to increase the RPI is play non-conference opponents we can beat that we expect to have a good record.  If everyone in the league did that, our RPIs would not be as heavily affected by games against each other.  I presume the Mac is doing this and that could be a philosophy in both teh A-10 and Big East although I haven't studied their schedules; just know that the RPIs of those teams, like the Big 10, don't change dramatically when they get into conference play unless they have a significantly poor record.

This RPI discussion has been great but I still wonder if it's any more than a starting place for the NCAA selection/seeding committee.  Seems they often have their work done by Friday night.  Anyone willing to take the NCAA RPI rankings, if available, and compare them to the final bracket?

[agibson]

As WH suggests, the goal if we want to increase the RPI is play non-conference opponents we can beat that we expect to have a good record.  If everyone in the league did that, our RPIs would not be as heavily affected by games against each other.

RPI conversations around the board lately have taught me a few things, confirmed some others, and me me wonder about yet others.

One is that if we just want to improve _our_ RPI then scheduling a couple of top-RPI losses is a fine idea. 

If we want to improve the _conference_ RPI then it's maybe not such a great strategy.  In that's our goal, I wonder if the best strategy's not quite about scheduling "winnable" games against decent opponents.  I wonder if it's about padding the non-conference schedule.  If the whole conference was 10-2 even against weak opponents we'd all benefit from the 50% of our RPI that comes from opponents' win-loss records.  We play our conference opponents more than anybody else.  Sort of a "big conference" strategy, writ small.   I wonder if that can work for a mid-major conference?

[a3uge]

As WH suggests, the goal if we want to increase the RPI is play non-conference opponents we can beat that we expect to have a good record.  If everyone in the league did that, our RPIs would not be as heavily affected by games against each other.

RPI conversations around the board lately have taught me a few things, confirmed some others, and me me wonder about yet others.

One is that if we just want to improve _our_ RPI then scheduling a couple of top-RPI losses is a fine idea. 

If we want to improve the _conference_ RPI then it's maybe not such a great strategy.  In that's our goal, I wonder if the best strategy's not quite about scheduling "winnable" games against decent opponents.  I wonder if it's about padding the non-conference schedule.  If the whole conference was 10-2 even against weak opponents we'd all benefit from the 50% of our RPI that comes from opponents' win-loss records.  We play our conference opponents more than anybody else.  Sort of a "big conference" strategy, writ small.   I wonder if that can work for a mid-major conference?

Look up the MAC - joke of an OOC schedule. Worst in the country. If the Horizon scheduled like that, our RPI would be like 25 right now. They only had like 3 top 100 RPI wins out of conference. Buffalo's RPI is tremendously inflated due to losing to Wisconsin and  Kentucky.