How good is Stingtalk at predicting the outcome of GT football games?

MtownJacket

MtownJacket

Dodd-Like
Joined
Jun 2, 2009
Messages
6,698
I have compiled the past three years of predictions and compared them to the actual results. Here is what I found (note: the obvious outlier MTSU 2012 has been disregarded in these plots).

marginVSprob.png


fractionVSprob.png


Some interesting facts:

1. Every game Stingtalk has predicted as a loss has resulted in a loss. (Miami 2011, VPISU 2011, UGA 2011, UGA 2012)

2. Every game stingtalk predicted would have a win probability of less than 0.59 resulted in a loss.

3. The lowest probability Stingtalk predicted that resulted in a win was Clemson 2011 (0.598).

4. The games where Stingtalk was the most confident but GT still lost were:
MTSU 2012 (0.942)
NCST 2010 (0.851)
Kansas 2010 (0.836)

5. Georgia Tech wins 69% of the games Stingtalk predicts to have a win probability above 0.6.
 
Pretty mediocre, IMO.


Sent from my iPhone using Tapatalk
 
I think that all means that we are a pretty optimistic bunch.
 
MtownJacket said:
I have compiled the past three years of predictions and compared them to the actual results. Here is what I found (note: the obvious outlier MTSU 2012 has been disregarded in these plots).

marginVSprob.png


fractionVSprob.png


Some interesting facts:

1. Every game Stingtalk has predicted as a loss has resulted in a loss. (Miami 2011, VPISU 2011, UGA 2011, UGA 2012)

2. Every game stingtalk predicted would have a win probability of less than 0.59 resulted in a loss.

3. The lowest probability Stingtalk predicted that resulted in a win was Clemson 2011 (0.598).

4. The games where Stingtalk was the most confident but GT still lost were:
MTSU 2012 (0.942)
NCST 2010 (0.851)
Kansas 2010 (0.836)

5. Georgia Tech wins 69% of the games Stingtalk predicts to have a win probability above 0.6.
Click to expand...

So we're fantastic barometers of failure, but unreliable in predicting success. This result definitely explains observations. Good science, MtownJacket. :lol:

I think I'll do a similar analysis on my new predictor version and put it ITT.
 
I'm just here for the inevitable nerd battle between gtphd and buzzczar about what R^2 really means
 
I might do a realist version of my final predictions where I subtract .1 from all probabilities and see what comes out.
 
clapper said:
Seems like this would be a more useful tool if you added the spread to the margin of victory. At least for the gamblers of the group.
Click to expand...
Yeah, I can add that part if mtown puts all this data in a google spreadsheet and shares it online.
 
neat. thanks for posting. we definitely need to continue this... perhaps make it an informal competition.
 
Any way to get you to limit the x-axis to 1.0? Seeing a winning probability going up to 1.2 creeps me out.
 
cyptomcat said:
put in Vegas line but I have several missing that I can figure out later:
https://docs.google.com/spreadsheet/ccc?key=0Ar8cqnkh36RRdFhqQlpUdjE1bHV3X1BUdUJ1X28wZlE&usp=sharing
Click to expand...

https://docs.google.com/spreadsheet/ccc?key=0AgAWEinzf84adGloVTNwTjk5QWczVU5QYUM5YnVhNkE&usp=sharing

Next season's version of the predictor in the same span would have been 28-5-3 (84.8% on games post week-1) on the money line for GT using only the data available at the time, missing only Kansas in 2010, UVA and Clemson 2011, and Miami and UNC in 2012. Beginners luck, IMO.

Notably, it correctly predicts us to lose to MTSU in 2012. Also notably, it picked us to beat Presby 184-7. :lol:
 
ThisIsAtlanta said:
https://docs.google.com/spreadsheet/ccc?key=0AgAWEinzf84adGloVTNwTjk5QWczVU5QYUM5YnVhNkE&usp=sharing

Next season's version of the predictor in the same span would have been 28-5-3 (84.8% on games post week-1) on the money line for GT using only the data available at the time, missing only Kansas in 2010, UVA and Clemson 2011, and Miami and UNC in 2012. Beginners luck, IMO.

Notably, it correctly predicts us to lose to MTSU in 2012. Also notably, it picked us to beat Presby 184-7. :lol:
Click to expand...
solid, but what's the '3', predicted as tie? Which games?

Btw you need to make it public for others to be able to view it.
 
cyptomcat said:
solid, but what's the '3', predicted as tie? Which games?

Btw you need to make it public for others to be able to view it.
Click to expand...

Derp. Public'd.

The 3 were week-1 games where it wasn't able to issue a prediction. Still working on that part.
 
You must log in or register to reply here.
Back
Top