We are grateful to have received a lot of positive feedback from the release of Episode 83 in which we unpacked NPI. We are equally grateful to have had the opportunity to answer some questions and make some clarifications. NPI is a big topic that can be confusing so we wanted to centralize all the questions we’ve received from Division III coaches and responses we’ve given for the benefit of anyone interested.
The initial release of this blog post contains the questions received to date but it will be updated if/when additional questions are received. We can be reached through social media, email or our site contact form.
Question 1: You shared about the dials and the different values between home/away outcomes. Am I right to assume a game played on a neutral floor (in season tournament, conf tournament, etc) would count as an away contest outcome?
Answer 1: Neutral court games would have a 1.0 weight applied. It’s a good question that isn’t clear in the way the dial settings are presented and we should have made more clear. It’s really a 1.2/1.0/0.8 for non conference and 1.1/1.0/0.9 for conference.
You can see it show up on the example of Calvin’s schedule that we used in the show with games against Marietta and La Roche (at the Great Lakes Invitational at Wittenberg), Berry and Platteville (played at Trine holiday classic) and one of the Hope losses (MIAA tournament at Trine).
Question 2a: Only part that confused me was the weighted NPI. Does that even matter? You guys sorted by regular NPI? And then used the multiplier for minimum wins calculation only. Just wanted to see if the NPI is the weighted average or not.
Answer 2a: I kind of glossed over that quickly, but the final NPI is the sum of the weighted NPIs divided by the sum total of the net games played (multiplier wins plus multiplier losses).
Question 2b: So if you use weighted NPI for final numbers do you scratch the weighted NPIs that don’t help you once you get to 12.5 wins? I believe you guys were scratching regular NPI.
Answer 2b: You scratch regular NPI, since for any single game, the weighted NPI divided by the weighted wins just gets you back to regular NPI. In a way, the multiplier just makes the game “more or less of a game” rather than a harder or easier game.
Question 3: Every team needs an NPI to calculate strength of schedule & quality wins but how does that work on the first iteration when none of the teams are given an initial NPI?
Answer 3: This is an issue with NPI in that it cannot be calculated directly all at once since, as you point out, the outputs of one are inputs to all the others. In order to do this you must iterate the process multiple, multiple times. That still leaves the problem of needing a initial outputs. The solution is to substitute opponents’ winning percentage (OWP) for ONPI and to exclude the quality of win bonus for the first run of the calculations then iterate from there. A more robust discussion of the formula iteration process starts at 35:55 of the video.
Question 4: related to the multipliers when I was looking at your breakdown of the Calvin game by game and Defiance game by game to get their NPI scores. I’ll use the 1.1 / 0.9 men’s basketball conference dials for this example. Should the 0.9 multiplier (conference) be applied to the home team regardless of win or loss? If you look at Calvin’s schedule around the 31:19 timestamp of your video. They lost to Trine at home (column 24) and got a 1.1 multiplier but lost on the road to Hope (column 25) and got a 0.9 multiplier.
What really made me look more closely at this was Defiance and maybe it will make it easier looking at the same opponent (Earlham in this example around the 32:25 mark in the last two columns). If I am reading your data sheet correctly. Defiance gets a 1.1 multiplier (35.026 final weighted NPI score) for their home loss against Earlham and gets a 0.9 multiplier for their road loss (28.659 final weighted NPI score). You seem to have a much better understanding of the system and maybe I am misunderstanding this (which is very possible), Defiance gets a higher weighted NPI score for a home loss compared to a road loss against the same opponent.
I think if I understand their logic that they want the weights to come out to average 1.0 for the game between the two opponents. (1.1 vs. 0.9 dial averages out to 1.0). So every home win for a team has to correspond to a road loss for their opponent and vice versa.
Team A wins at home (0.9 multiplier) / Team B losses on the road (1.1 multiplier?)
Team B wins on the road (1.1 multiplier) / Team A losses at home (0.9 multiplier?)
I actually just looked at the NCAA cheat sheet/reference guide and their reference sheet backs up what you did mathematically and you followed it how it was written. But it just seems backwards to me in my mind and I am curious if you thought the same thing.
Answer 4: The thing to clarify is when the multipliers are applied because they don’t always average out to 1.0. The “weightier result” (1.1 in conference or 1.2 in non-conference for men’s basketball) is applied for an away win or a home loss whereas the “lighter result” (0.9 in conference, 0.8 non-conference) is applied for a road loss or a home win.
For example, when Calvin won at St. Norbert last season, they got “credited” with a 1.2 multiplier while St. Norbert got “dinged” with a 1.2 multiplier. Winning on the road is good (more credit), losing at home is bad (more penalty). Winning at home is more expected (less credit), losing on the road is more expected (less penalty). It can be confusing and isn’t the natural way my mind thinks about how a multiplier would be applied because it scales the game, not the opponent.
In regards to the Defiance vs. Earlham example, they both show up as a counted game when dividing out by the adjusted games. With it being two losses to a conference opponent you get a scenario where you’re basically multiplying then dividing by the multiplier and ending up where you started.
Question 5: In soccer, how will draws work? I would assume they count as half a win and half a loss. Thus, are you able to drop just one side of the draw if the win part hurts you or the loss part helps you? Or do you have to drop the entire match? Or can you even drop a draw?
Answer 5: [Note: a previous version of this post incorrectly described how ties would function; this post has been updated.] You are correct that ties are counted as half a win and half a loss. These two halves of the result — the win half and the loss half — are essentially handled separately. If the “win half” of the result is one that would hurt NPI (barring minimum wins threshold), it can be dropped from the calculation.
As a (hopefully) clarifying example, one win and one loss versus the same opponent would result in the same NPI as two ties versus that opponent. If you would drop the ‘true’ win versus that team, you would also drop the two ‘half wins’ from the ties.
D3 Datacast Links:
YouTube | Spotify | Twitter | Facebook | Patreon | Merch Shop
It is our commitment to the Division III basketball community to keep this site free of ads or paywalls. We are grateful for the teammates we have in that goal who have chosen to support us through Patreon.
D3 Datacast 
Leave a comment