@Sangre Naranjada — I’m working on the spreadsheet and reading your proposed method, the actual final score (31-27) spread came to 4 points. So contestant A would calculate at 3+6+4 rather than 3+6+3 since the spread is 4, no? What am I missing?
Contestant A and B guessed the spread would be 7 points. The actual spread was 4 points. 7-4=3 Contestant C guessed the spread would be -4 points (from the perspective that Texas would lose). The actual spread was +4 (Texas won), thus a total of 8 points assigned to Contestant C
I think we have a good prediction model now. Thanks to @Sangre Naranjada for the method and answering questions while I worked on it. The Delta column is the calculated number based on how close the pick was for three data points: 1. Texas’ predicted score and reality 2. Opponent’s predicted score and reality 3. Difference between predicted and actual spread Lower delta indicates the more accurate prediction. Picking the final score exactly would result in a delta of zero. If someone picks the wrong winner the Delta column just gets a “-” rather than a score Example 1: TEXAS wins 48-17 Dukesteer wins here with a delta of 2 HFTEXASLA TECHDelta Driver 8522414dukesteer48182FWHORN452726gahornphan4044-Garmel382128George Bailey372334 Example 2: TEXAS loses 20-28 Dukesteer wins here with a delta of 16 HFTEXASLA TECHDeltadukesteer172016easy3417-EDT173520FWHORN4527-gahornphan404440Garmel3821-George Bailey284126
Assume you get the winner right. If W is how far off you are from the winner's score, and L is how far off you are from the winners score, it looks like "delta" is 2W + 2L if you were off in opposite directions (one too high, one too low), and W + L + |W-L| if you were off in the same direction (both high or both low). That seems reasonable. In other words, you always have "bad points" equal to the sum of W and L for the individual teams. For the spread, you either gain "bad points" either to the sum of W and L again, or to difference between W and L. Harder to say than it is to do, but the mathematical logic of that is fairly simple and elegant. This way, if the real score is 31-10 and the predictions are 34-13 and 34-7, both guesses were off by 3 for both, but 34-13 has a better delta because his spread was correct and so he has a delta of '6', whereas 34-7 has '12'. So the tipping point where getting the spread exactly right wouldn't be enough due to going way over the total points is that 37-16 would be an equivalent to 34-7, but 38-17 would be worse. TL/DR formula seems reasonable.
I just realize I can state that a simpler way. If you were off in opposite directions, your delta is 2W + 2L. If you were off in the same direction, it's either 2W or 2L, whichever is larger. The result is you end up with everyone having an even number for a score, since everything is getting doubled (because every point off from reality affects two things, both that team's score and the overall point spread).