Question on Crystal Ballz Scoring
Been wondering about this for awhile. On the 247 site, their crystal ball predictors are given a score along with their percent correct. What's odd is that there doesn't seem to be much correlation between the two; i.e., some predictors with high scores have low percentages. What goes into determining the score, and why would it be seen as an indicator of merit if it doesn't require accuracy? A brief Google search turned up nothing, hoping someone in the know on the board will respond.
January 18th, 2016 at 8:50 AM ^
Crystal Ballz is the new "Flight Tracker"
January 18th, 2016 at 8:54 AM ^
Higher score for how early you predict.
So if you put one in for Brandon Peters a day before he commits, you'd get a much lower score than someone who put it in weeks prior.
Which makes sense for the %-score correlation, the high % pickers may have a lower score because they've waited for substantial evidence before they made their pick, causing them to lose points compared to someone who threw in a guess.
Personally, I don't pay attention to the score, I think the % helps gauge better. If I see someone with a high % putting in a CB, then that is more than likely a quality pick.
January 18th, 2016 at 8:56 AM ^
January 18th, 2016 at 9:00 AM ^
I agree, there doesn't seem to be much point to the score, which makes me wonder why they bother to calculate it.
January 18th, 2016 at 9:02 AM ^
The point of the score is to encourage pickers to pick early if they feel like they have insider information, and not just put in Ballz immediately before a guy is going to commit.
January 18th, 2016 at 9:07 AM ^
if the score isn't a good gauge for how accurate a predictor is, to the point where people ignore it entirely in favor of the percentage, why would predictors want to try and get a good score?
January 18th, 2016 at 11:28 AM ^
The point is that both the score and percentage are relevant. If a guy has a good score but a bad percentage then he isn't that accurate, and if he has a bad score but a good percentage he's just making picks once they're obvious. Only pickers with both a good score and a good percentage are accurately making picks before they're known.
January 18th, 2016 at 12:42 PM ^
It just seems to me that if you're going to have a "score", that you would want a single metric that captures everything rather than two metrics which capture different traits. Otherwise, how do you know who wins?
A suggestion to the 247 mods (who I'm sure are following this thread closely): give extra points to the 1st person to correctly make a prediction.
January 19th, 2016 at 10:43 AM ^
Providing two scores give a more complete picture of the kind of predictions a picker makes, though. I don't think the point of the score is to determine who "wins," because there can't be that many people who care whether Wiltfong or Lorenz is better. I think the point of the score is to give context to the crystal ball picks.
January 18th, 2016 at 5:45 PM ^
A high score but low percentage means that person is relatively good at early guessing.
A low score but high percentage means that person is very good at late guessing, but their early guesses should be ignored.
January 18th, 2016 at 5:45 PM ^
A high score but low percentage means that person is relatively good at early guessing.
A low score but high percentage means that person is very good at late guessing, but their early guesses should be ignored.
January 18th, 2016 at 10:39 AM ^
January 18th, 2016 at 9:01 AM ^
Follow-up question: do pickers lose points for switching their Ballz? Because if not, it seems like the way to go would be to make an early pick for each recruit (going for points), then if that pick appears to be wrong, switch picks when it becomes clear the recruit is about to commit (preserving a good percentage).
If that's the case, then a pick by a guy with a high percentage wouldn't necessarily mean much, because he might just switch that pick later if it looks like he's going to be wrong.
January 18th, 2016 at 10:10 AM ^
January 18th, 2016 at 9:28 AM ^
The one problem with looking at % only is that a lot of people will put their pick or switch their pick in the last 24 hours before an announcment after someone else has broken the story. The record book shows they got the pick right, but there really was no value in the pick though. I do think they need a better way to present the score though points per crystal ball prediction or something like that.
January 18th, 2016 at 10:06 AM ^
January 18th, 2016 at 11:29 AM ^
Accuracy percentage is not meaningful though if it is gained through changing picks at the last moment. How can you have confidence in someone's pick if they are likely to change it later?
January 18th, 2016 at 1:13 PM ^
January 18th, 2016 at 4:11 PM ^
January 18th, 2016 at 5:49 PM ^
Agreed that a month ago that you were more interested in score than percentage, but a high percentage person's changes in the last 2 weeks would have predictive value.
January 18th, 2016 at 9:53 AM ^
January 18th, 2016 at 9:00 AM ^
It's like that trivia game you play at B-Dubs. The quicker you guess an answer after the question is posted, the more points your correct answer is worth. If you wait till after 3 clues, and then put in your answer, you are worthless and embarrassing to your family.
January 18th, 2016 at 9:17 AM ^
This, pretty much.
January 18th, 2016 at 9:37 AM ^
Nice to see you here again, Steve. How often can someone change their pick and how long is it locked? Also, do you guys have a lockout time before a scheduled announcement?
January 18th, 2016 at 1:15 PM ^
We can change our pick frequently. Normally you don't see more than one change or so, but with some up and down recruitments I've changed mine a few times.
If a player makes an announcement date clear, it's locked usually about a week before the commitment is made.
The thing about the CB is that all the standard complaints people have about it are things that we (I) are more than cognizant of. I do what I can to avoid the flipping and flopping of picks, but in some recruitments it's difficult because you want to paint an accurate picture for readers.
January 18th, 2016 at 9:30 AM ^
That is a good analogy
January 18th, 2016 at 10:07 AM ^
January 18th, 2016 at 12:54 PM ^
Meta note: this post os on topic, and neither OT nor Meta.
OT posts are obvious: dog's UTI problems, the impending birth of a child, gas prices in Michigan, pro teams in Michigan, Flint water problems, death of David Bowie, tour dates of favorite musicians, what you're drinking at a particular time, day of the week posbang threads, restaurant or food questions, girlfriend or wife questions, CJK5H, etc.
META posts, on the other hand, are questions about the site: who designed it, why it crashed, why the links don't work, what's going on with the server, suggestions for the site banner, who is on the staff, what's happening with HTTV, why you can't page down to see threads from earlier in the day, how to search for something, how to embed something, how to format something.
OT posts are often interesting and informative. Well, except for the endless, interminable, mind-numbing Friday posbang and what-are-you-drinking threads.
META posts are kind of pointless. Brian and Seth are very aware of the site inadequacies. Because of Harbaugh, the busy season, wives and children, and the position of the moon, you will very rarely get an answer to a META question. It is what it is, and will maybe be fixed or improved some day.
January 18th, 2016 at 1:05 PM ^
Thought "meta" extended to the entire industry, not just MGoBlog.