Granger, when you have a game that goes into OT, do you enter the final end of game score or the tie before OT? You could have an eight point swing one way or the other. (would that even be a 16 point swing from the tie?) I was thinking that the tie score would better reflect the closeness of the two teams.
A question about the toy
- Thread starter Blue Bird
- Start date
Bulldog'88
Six-man fan
Is the drunken monkey one of the monkeys the ULI uses to realign?
Blue Bird":22addhpj said:
final score.
tie would be the best representation of the teams... if they didn't play the OT. But they do.
well that and drunken moneys, of course.
smokeyjoe53
Six-man pro
No..... granger uses his own specially trained monkeys to only pick his favorites...........Bulldog'88":2hvir2y4 said:
I agree with the weirdness if you look at just a few instances of how teams are ranked in the top-10, but if you look at the totality of the system, there comes a time when there are too few games yet some already conflicting results that really start to expand the system.
Here's an example of what I mean. One of the objectives is to recreate the already existing results with perfection. That could mean score and it also means who won. So far the season looks like this
(1, 78-33, '0.70', 58)
(2, 95-22, '0.81', 56)
(3, 83-29, '0.74', 69)
This is what the 'toy' predicted. (That last number if the number of '45' games)
Based on the new rankings, the 'toy' would go 332-8. Using the system, I could get it down to 335-5, but the expansion is too great. I use another error measurement as well that includes expected score vs. actual score. You try to minimize these together.
I think most people look at White Deer and freak out. Let's be honest here. That's what this is all about. White Deer too high and Richland Springs too low.
Let's see their body of work.
White Deer (3-0) is currently ranked #5 and has a rating of 342.03
def #56 Lefors (2-1) 49-0
def #90 Hart (1-2) 61-0
def #27 Motley County (1-2) 50-0
Best win is over #27
Worst loss is from #0
average opponent rating is 264.856666667
They haven't been scored on and they've 4'd everyone, including Motley County, who only lost to Crowell by 13. If Crowell is a top-10 team, why shouldn't White Deer be one?
Here's RS results
Richland Springs (2-1) is currently ranked #7 and has a rating of 334.30
lost to #2 Happy (2-0) 96-74
def #79 Waco Methodist Childrens Home (1-2) 52-6
def #85 Wellman-Union (1-2) 62-0
Best win is over #79
Worst loss is from #2
average opponent rating is 281.323333333
Other than a loss to Happy, they really do not have much helping their ranking.
I agree they should be higher, but not much there. The computer is not a human saying, hey, they were missing their best player. It just sees the score.
If you want to check the WHY's for more teams, do not forget this link
http://sixmanfootball.com/why_test.php
Here's an example of what I mean. One of the objectives is to recreate the already existing results with perfection. That could mean score and it also means who won. So far the season looks like this
(1, 78-33, '0.70', 58)
(2, 95-22, '0.81', 56)
(3, 83-29, '0.74', 69)
This is what the 'toy' predicted. (That last number if the number of '45' games)
Based on the new rankings, the 'toy' would go 332-8. Using the system, I could get it down to 335-5, but the expansion is too great. I use another error measurement as well that includes expected score vs. actual score. You try to minimize these together.
I think most people look at White Deer and freak out. Let's be honest here. That's what this is all about. White Deer too high and Richland Springs too low.
Let's see their body of work.
White Deer (3-0) is currently ranked #5 and has a rating of 342.03
def #56 Lefors (2-1) 49-0
def #90 Hart (1-2) 61-0
def #27 Motley County (1-2) 50-0
Best win is over #27
Worst loss is from #0
average opponent rating is 264.856666667
They haven't been scored on and they've 4'd everyone, including Motley County, who only lost to Crowell by 13. If Crowell is a top-10 team, why shouldn't White Deer be one?
Here's RS results
Richland Springs (2-1) is currently ranked #7 and has a rating of 334.30
lost to #2 Happy (2-0) 96-74
def #79 Waco Methodist Childrens Home (1-2) 52-6
def #85 Wellman-Union (1-2) 62-0
Best win is over #79
Worst loss is from #2
average opponent rating is 281.323333333
Other than a loss to Happy, they really do not have much helping their ranking.
I agree they should be higher, but not much there. The computer is not a human saying, hey, they were missing their best player. It just sees the score.
If you want to check the WHY's for more teams, do not forget this link
http://sixmanfootball.com/why_test.php
and please do not PM me. The questions I get asked are good and healthy discussions. I will not eviscerate you... probably.
nothing, really?
I'll add something here, if not. Richland Springs must be the biggest hot-button topic for people. I'd say it's a pretty even split of people I hear from who say How can RS be still ranked so high? vs. those who think they should be #1 in D2 no matter what...
The best part is that actual ppl from RS don't really care, from what I can tell. They seem to know what the end game is.
I'll add something here, if not. Richland Springs must be the biggest hot-button topic for people. I'd say it's a pretty even split of people I hear from who say How can RS be still ranked so high? vs. those who think they should be #1 in D2 no matter what...
The best part is that actual ppl from RS don't really care, from what I can tell. They seem to know what the end game is.
I'll chime in.. I think white deer being number 3 is not real accurate but only my opinion.. I get that it's numbers and why it does what it does. Now that being said it may be fun to look at but the teams who have what it takes to go on and play down the road in the playoffs could care less about rankings. The final score is decided by the boys on the field and coaches on the sideline and that is the best part as the Toy may not match a poll of coaches players and definitely not parents who's kid is the BEST LOL
CleverUsername1
11-man fan
I don't know if the data is still available but it would be interesting to see the percentage of games correctly predicted since the toy's creation broken down by weeks. One season worth of data isn't big enough to draw any conclusions, but a sample size thousands of games large taken over several years would be big enough to see how the toy improves over the course of a season. It'd also be cool to see the correlation between the predicted spread and the actual spread of every game ever predicted by the toy, but that would be an enormous undertaking that I doubt anybody would volunteer to take on.
