Warrior General Games

OK, guys, not very scientific, but I plotted a few variables vs. season average punting distance and looked at the Rsquared for each of them.

I used attributes with equipment.
Some of my attributes were pretty clustered (e.g. at first cap 48-50 for strength) so may not be very good predictors

Guess what the biggest 2 predictors of punting average were, by a long shot...

n=18 punters. All season 8 numbers.

Using what Octowned found out, i'd say...

Conf and Agi

(Conf was a safe pick, Agi was more of a shot in the dark)

The Rsquared for 1 was 0.3882 and the other was 0.4074. Highly predictive. The next highest was only 0.1371 while the 4th was all the way down to 0.0500

Hint... I didn't look only at attributes

Originally posted by Firenze
Using what Octowned found out, i'd say...

Conf and Agi

(Conf was a safe pick, Agi was more of a shot in the dark)

Actually, didn't even look at Agility, Jumping or Speed. Will do those another day...

Originally posted by toppshelff
Originally posted by Firenze

Using what Octowned found out, i'd say...

Conf and Agi

(Conf was a safe pick, Agi was more of a shot in the dark)

Actually, didn't even look at Agility, Jumping or Speed. Will do those another day...

Lol, ok. So you looked at Punt, Conf, Vis, Str and SA's?

Didn't look at SAs, but did look at a couple different numbers that ended up being the biggest predictors.

field position?

your Rsquared values were very very low...there is tons of variation in your statistics...

Originally posted by caesarrodneybengal
your Rsquared values were very very low...there is tons of variation in your statistics...

Huh? 0.4 is low, since when? There are so many vaiables, I would expect a good prediction to be about 0.1 to 0.2. To get a 0.4 is very high, IMHO. If I got a 0.7 or 0.8 I wouldn't believe the data.

Originally posted by zomgmike
field position?

No that's a great suggestion. I didn't look at this, would be a lot of work, but I believe it is true, not just because of short coffin corners and TBs, but I think Bort increases the chance of long punts when you are deeper in your end of the field.

Would be very interesting to look at.

So the last factor was number of punts. There was a very strong negative correlation with the number of punts and the average distance of those punts. This is a little different than Octowned's analysis and was what surprised me. I believe that Octowned found that punt distance didn't change over the length of a game, suggesting that punting isn't impacted from energy level.

I've wondered how much of this is "pre-determined" before each game. So if my punter's team is really overmatched, does the AI calculate what his average will be before running the game model, knowing that the punter will punt much more.

I don't know if there is a better reason, I'd love to hear thoughts on this.

For information, also very interesting, the punting attribute was perfectly flat, not a predictor at all of punting distance. Now, I will add that all my punters are over 100 with equipment. My first assumption was that punting over 100 has no value (ugh), but thinking about the cost of adding a punting punt above 100, I realize that the cost to other attributes would maybe make up for the increase. Still, concerning, would appreciate your thoughts.

Vision was the 3rd best predictor at 0.137. Strength was a predictor with an R squared of 0.05.

So this was a quick analysis. All these variables were looked at independently. I'm not smart enough to do a regression analysis to really understand these data. If I were to do something like that, I'd want a lot more data points, and not "clustered" like some of my guys are.

Thoughts?

The problem with this is correlation. If you want to PM me your data I can run a quick correlation test to see what predictors even have the potential to mean something.


Stamina may play an issue, but possibly not at such a low level as my analysis, so it never showed up in the puny punts anyway?

R squared is considered about average at .6, low at .4 and high at .8

An R squared of .4 means that only 40% of your variation in sample data was accounted for by the regression curve. That says to me that your regression was very weak

Originally posted by caesarrodneybengal
R squared is considered about average at .6, low at .4 and high at .8

An R squared of .4 means that only 40% of your variation in sample data was accounted for by the regression curve. That says to me that your regression was very weak

If you have only 2 variables, 0.4 would be low. When you have many variables involved, 0.4 could be strong. Actually, I've seen values of 0.2 be considered pretty strong in a very complex scenario.

I would hope that there is an equation somewhere that is very highly predictive of punting distance. Probably the equation that Bort uses, for instance. And even it will have some random variables in it to ensure a significant variation.

What about a punter's weight? Do fat punters blast it farther than skinny ones? A quick run on the stat re-roller gave me a low of 160 and a high of 203. How hard would it be to get 10 punters with identical stats with weights throughout this range? Put them all on 0-16 teams and start collecting stats for punts not affected by going OOB or TB.