2023 PB College Football Pick-em Rules and Regs

Not open for further replies.


Puritan Board Doctor
It is getting to that time of the year again! Rules and regs will be slightly different than last year.

1. Slate will normally be 13 games, including tie-breaker, which (contrary to last year) will be included in the total number of correct guesses.

2. Tie-breaking procedure will be three-fold: a. correct winner; b. closest aggregate score (Todd will be doing this part); c. first entry

3. Slate will be chosen on Monday mornings, as per usual, as will the winner of the previous week.

Other long-standing rules remain in place (entries received after kickoff of first game will not be eligible, no Sunday games, and announcements will include day of earliest game if it isn't Saturday. Week 0's slate will be announced this coming Monday (all 7 Saturday games will be on the slate), and week 1 will have Thursday games, so be on top of that!

I believe this is the first year the BCS will be doing a playoff. Am I correct? Never too early to be thinking of how that will work for our Bowl Bash.

I believe this is the first year the BCS will be doing a playoff. Am I correct? Never too early to be thinking of how that will work for our Bowl Bash.
I thought the playoff expansion was happening next season, not this season. To my current knowledge, this year's championship will be the winner of the Rose Bowl and the winner of the Sugar Bowl.
You can go ahead and put down my pick in the opener of Georgia over the University of Tennessee Martin,

Leaning toward Ole Miss being able to take Mercer in another big matchup.
For the newbies (or the old guard who need a refresher) who may be curious about how the tiebreaker will work if people have the same number of correct games picked, here is what we'll employ. It's an attempt to get the best overall score prediction by looking at a combination of how far the prediction is from both scores in the game, rather than other options that only consider part of the prediction.

Alternatives that sometimes have been used would consider A) weighing only the winning score first as a first tiebreaker, and then only considering the losing score in case of continued ties, for instance (which allows a prediction of 49-0 to be the tiebreak winner against a prediction of 48-47 when the game was actually 49-48); or B) weighing only the total points scored by the two teams first, and then the point spread (which would treat a prediction of 55 points to win over someone's prediction of 54 total points regardless of whether the final score was 55-0 or 28-27.

So, I've attempted to create a metric that works well taking into account both teams' efforts and rewards what is the prediction which best reflects the outcome of the game. It's a bit long-winded, but I have a graphic that helps :)

Tiebreaking procedures

The following rules apply if two or more contestants have picked the same number of winning games out of the full slate of games, including the one designated the tiebreak game.

The tiebreaking procedure follows these steps in order until there is one player remaining.

1) Correct winner chosen for the tiebreak game.

Example: If contestants A, B, and C are participating in the tiebreaking round, and only contestant B properly chose the winner of the tiebreak game, then B is the winner. If both A and B got the winner of the tiebreak game correct, then they move on to step 2, and C laments his or her prognosticating abilities.

2) Among those who correctly chose the winner of the tiebreak game, the following straightforward metric is used to choose the winner: the pick which is closest to the actual final score, taking into account both scores simultaneously, and reflects the most accurate prediction overall.

To determine this, we take the simple approach of measuring the “distance” between the prediction and the actual result in the plane of winner’s score vs. loser’s score. This has the virtue of I’ve included the following illustration to help visualize this:

3) In the unlikely event that there remain more than one with exactly the same “distance”, which will usually occur only if both contestants picked exactly the same score, though there are other possibilities, the winner will be the one whose entry was posted earliest. Again as above, if one edits their entry mid-week, the date of posting for tiebreaking purposes is the time of edit.

This procedure provides an unambiguous winner.


I used your formula for a few seasons before I understood how it worked. You finally explained it to me in a way I could understand a few years ago.
I believe that to be complete, I need to add one additional criterion to the above to choose the 'closest score prediction'. I just came up with this while thinking about scenarios. i'm surprised I hadn't considered this before, but I think you'll agree that this adjustment is best.

Distance in my previous proposal used

WSD = winner score difference = WS(pred) - WS(actual)
LSD = losing score difference = LS(pred) - LS(actual)

With the metric being

D = distance = square root of (WSD*WSD + LSD*LSD)

This almost always gives an absolutely clear minimum. One prediction in almost all cases will be the closest. It would be rare to have two predictions that gave the same D value, but it certainly can happen. In thinking through those rare cases, I found the following hypo:

Suppose as in the diagram, 30-24 is the actual score, and someone has predicted a score of 32-26. Winning score wrong by 2, losing score by 2.

This would yield a D value of the square root of 2*2 + 2*2, or sqrt (4+4) or 2.83

Problem is, a D value of 2.25 would also arise if the prediction was 32-22, since the winning score is wrong by 2 and losing score by 2, the other way.

I think all would agree that the prediction of 32-26, for a 6 point spread, just like the real game, is definitely better than 32-22, a 10 point spread, yet the score predictions for winner and loser would be just as close.

Others that would also give a tie with D = 2.25 would be 28-26 and 28-22. We'd surely judge 28-26 to be poorer than the 32-26 prediction, but we would judge 28-22 to be equally good as 32-26, I bet, because the errors made by the predictor were each the same, yielding the same point spread.

Given this... I am planning to evaluate the point spread too, in cases where D is the same between two predictions. That resolves the weirdness of those extreme cases where the errors in prediction are in the same direction (good point spread) vs. the errors being in opposite directions (making much too large or much too small point spreads).

The solution is simple. After D is calculated and is the same, we go to point spread. Closest to actual point spread is another step in deciding the best prediction that goes before the first-pick-in final tiebreak.


1) Was the correct winning team chosen
2) How far (in D as described) is the prediction from the actual score
3) How close was the prediction in point spread between the two teams
4) First pick in (with the understanding that the time of posting is the last edit of picks made)

This isn't a large change, but avoids some silly things that can occur.
For those who don't understand the formula - pretend it's a Baptist capital project, and "step out on faith"
After reading all that and with the addition of the chart, I’m ready for the rapture…
Not open for further replies.