The NSFL's inaugural season is under wraps. The Arizona Outlaws defeated the Colorado Yeti to obtain the first Ultimus Cup. While the Outlaws have reason to celebrate, it leaves the rest of the league with little to do but analyze our player and team statistics to try to determine the best route of hoisting the Ultimus Cup in season two.
With that in mind, I gathered season statistics for the first season and ran some analysis to attempt to determine which player attributes affect each statistic best. For the data, I used the League Defensive Stats page from the regular season index. I took only those with more than 30 tackles to ensure we had players with significant playing time. I also used the box scores from each week to determine the amount of time each team's defense spent on the field during the season. I felt this was an important thing to look at because there was a wide range.
Our champions, the Arizona Outlaws led the league in most defensive categories on a team-basis, including time on the field. The Outlaws defense spent 440 minutes and 32 seconds on the field through 14 games this past season. Following shortly behind were Baltimore and Yellowknife, at 439:59 and 439:37 respectively. San Jose came in at 428:44, Colorado at 398:53, and the Otters at a paltry 348:56 minutes on the field.
We'll start the analysis by position group, culminating in the full regression with every defensive position.
Starting in the defensive backfield, we have the defensive backs. I included CBs, SS and FS in this position group to allow for slightly more data points, to hopefully get a better picture.
This particular regression shows each player attribute and its relationship to total tackles on the defensive backfield, either CB, FS or SS. Speed is the the statistic that jumps out the most as important for DL and their tackling prowess. The coefficient of 1.33495 shows a strong correlation between speed and tackles, and is the only attribute with a p-value in the acceptable range for being called significant. Most statistics courses teach 95% confidence as the go-to for analysis. A p-value of less than 5% means that it is significant (I honestly can't remember why, but I do remember that it is).
Strength is the next-highest attribute in terms of coefficient at .578377. A statistician would not trust this number, as the p-value of .155723 is outside the 95% confidence interval. While we can't make definitive statements as to the level of importance on tackling, we can pretty safely say that it is important.
The last thing that jumps out to me are the negative coefficients for playing time, for agility and for tackling. The p-value's for these are atrocious, so we can at least safely assume that having points in agility and tackling, as well as playing more downs will NOT lead to less tackles.
The ANOVA table and Regression Statistics tables are included to show some important information as well. The Anova table shows an F-stat p-value of .068965. Going by the rule of wanting less than 5%, the regression as a whole is not significant, but is close. Basically what this tells us is that we cannot confidently say that these attributes lead to more tackles. It is quite close to the 95% confidence interval however, so there is some useful information we can find from this. Take this regression with a grain of salt. For those that don't know what regressions exactly tell you, this one for DL says:
Number of Tackles = -60 + -.016(playing time) + .578(strength) + -.63(agility) + .21(intelligence) + -.06(tackle) + 1.33(speed) + .31(hands) + .21(endurance)
Moving to the next level of the defense, we have the linebackers.
The regression results for linebackers leaves quite a bit to be desired. None of our coefficients can be described as statistically significant, as none have a p-value less than .05 (5%). Endurance is surprisingly high at 5.471742, showing a strong connection between high endurance and high number of tackles.
Tackling and Speed also show strong connections with number of tackles at 2.398985 and 1.378104 respectively. Again, none of the p-values are significant, so the relationship cannot be trusted and must be taken with a grain of salt. Playing Time shows the closest to significant of a relationship, and seems to have a fairly small relationship at .380659.
The strength and agility coefficients are both negative, which seems extremely silly. Luckily, the f-stat for the entire equation is not significant, so we can safely assume having points in strength and agility as a linebacker will NOT lead to less tackles.
Finally, we make our way to the trenches. The following regression includes only defensive linemen, both DTs and DEs and how their player attributes affect their number of tackles.
First off, the f-stat. It's REALLY high. Basically this tells us very little of this equation can be trusted to explain which attributes affect number of tackles along the defensive line.
None of our attributes have a coefficient that is significant, most aren't even close. Intelligence supposedly has the strongest relationship with number of tackles at a coefficient of 1.038118. Speed is the next highest at .652928. I believe this one a little bit more, but again, the p-value is not significant.
Lots of the coefficients for this one came out negative, which isn't a good sign. None of these attributes should hurt a defensive lineman's ability to make a tackle. I can certainly understand some not helping as much (if any, looking at you, hands) as others, but none should actually harm the statistic.
Basically, the DL regression tells us nothing, except that maybe speed is helpful for making tackles.
Finally, we have every defensive player above 30 tackles.
This one shows a very low F-stat p-value, which shows that these attributes can be trusted to lead to more tackles. I'm very loosely using the words, a real statistician would be able to tell you what they mean for sure, I'm just trying to remember from my week or two on this in Economics classes.
We have several coefficients that are actually significant, which is a very good sign. Speed and tackling show a pretty good relationship between the attribute and high tackle totals. Surprisingly enough, Intelligence also proved to be pretty high, and significant as well.
Agility and strength are a little lower, but still lead to more tackles, however the coefficients are non significant, meaning that the actual coefficient could be higher or lower (I'm guessing higher).
Playing Time is negative, which I found very interesting. Being on the field more should lead to more tackles, but that isn't what the regression has shown us. The only regression in which the coefficient for playing time was not negative was with linebackers. If you smell something fishy, it's okay, because I do too. None of the coefficients for playing time were significant by a statistical view.
In conclusion, these regressions didn't tell us all that much about each attribute and its effect on total number of tackles. Most of the coefficients, and even a few of the f-stats turned out to not be statistically significant. More data points would most certainly be helpful, only 63 players and their attributes were included in the overall regression. The position group regressions included only 26 data points (players) on the high end, with linebackers only have 16. Larger sample sizes will show us better information, and will be possible as we go through more seasons.
If I get the motivation, I will go through and run these same attributes and their effects on other defensive statistics, like interceptions, sacks and what-not. I've gathered most of the data already, so it shouldn't be difficult, I'm just lazy.
Thanks for reading!
GRADED
With that in mind, I gathered season statistics for the first season and ran some analysis to attempt to determine which player attributes affect each statistic best. For the data, I used the League Defensive Stats page from the regular season index. I took only those with more than 30 tackles to ensure we had players with significant playing time. I also used the box scores from each week to determine the amount of time each team's defense spent on the field during the season. I felt this was an important thing to look at because there was a wide range.
Our champions, the Arizona Outlaws led the league in most defensive categories on a team-basis, including time on the field. The Outlaws defense spent 440 minutes and 32 seconds on the field through 14 games this past season. Following shortly behind were Baltimore and Yellowknife, at 439:59 and 439:37 respectively. San Jose came in at 428:44, Colorado at 398:53, and the Otters at a paltry 348:56 minutes on the field.
We'll start the analysis by position group, culminating in the full regression with every defensive position.
Starting in the defensive backfield, we have the defensive backs. I included CBs, SS and FS in this position group to allow for slightly more data points, to hopefully get a better picture.
This particular regression shows each player attribute and its relationship to total tackles on the defensive backfield, either CB, FS or SS. Speed is the the statistic that jumps out the most as important for DL and their tackling prowess. The coefficient of 1.33495 shows a strong correlation between speed and tackles, and is the only attribute with a p-value in the acceptable range for being called significant. Most statistics courses teach 95% confidence as the go-to for analysis. A p-value of less than 5% means that it is significant (I honestly can't remember why, but I do remember that it is).
Strength is the next-highest attribute in terms of coefficient at .578377. A statistician would not trust this number, as the p-value of .155723 is outside the 95% confidence interval. While we can't make definitive statements as to the level of importance on tackling, we can pretty safely say that it is important.
The last thing that jumps out to me are the negative coefficients for playing time, for agility and for tackling. The p-value's for these are atrocious, so we can at least safely assume that having points in agility and tackling, as well as playing more downs will NOT lead to less tackles.
The ANOVA table and Regression Statistics tables are included to show some important information as well. The Anova table shows an F-stat p-value of .068965. Going by the rule of wanting less than 5%, the regression as a whole is not significant, but is close. Basically what this tells us is that we cannot confidently say that these attributes lead to more tackles. It is quite close to the 95% confidence interval however, so there is some useful information we can find from this. Take this regression with a grain of salt. For those that don't know what regressions exactly tell you, this one for DL says:
Number of Tackles = -60 + -.016(playing time) + .578(strength) + -.63(agility) + .21(intelligence) + -.06(tackle) + 1.33(speed) + .31(hands) + .21(endurance)
Moving to the next level of the defense, we have the linebackers.
The regression results for linebackers leaves quite a bit to be desired. None of our coefficients can be described as statistically significant, as none have a p-value less than .05 (5%). Endurance is surprisingly high at 5.471742, showing a strong connection between high endurance and high number of tackles.
Tackling and Speed also show strong connections with number of tackles at 2.398985 and 1.378104 respectively. Again, none of the p-values are significant, so the relationship cannot be trusted and must be taken with a grain of salt. Playing Time shows the closest to significant of a relationship, and seems to have a fairly small relationship at .380659.
The strength and agility coefficients are both negative, which seems extremely silly. Luckily, the f-stat for the entire equation is not significant, so we can safely assume having points in strength and agility as a linebacker will NOT lead to less tackles.
Finally, we make our way to the trenches. The following regression includes only defensive linemen, both DTs and DEs and how their player attributes affect their number of tackles.
First off, the f-stat. It's REALLY high. Basically this tells us very little of this equation can be trusted to explain which attributes affect number of tackles along the defensive line.
None of our attributes have a coefficient that is significant, most aren't even close. Intelligence supposedly has the strongest relationship with number of tackles at a coefficient of 1.038118. Speed is the next highest at .652928. I believe this one a little bit more, but again, the p-value is not significant.
Lots of the coefficients for this one came out negative, which isn't a good sign. None of these attributes should hurt a defensive lineman's ability to make a tackle. I can certainly understand some not helping as much (if any, looking at you, hands) as others, but none should actually harm the statistic.
Basically, the DL regression tells us nothing, except that maybe speed is helpful for making tackles.
Finally, we have every defensive player above 30 tackles.
This one shows a very low F-stat p-value, which shows that these attributes can be trusted to lead to more tackles. I'm very loosely using the words, a real statistician would be able to tell you what they mean for sure, I'm just trying to remember from my week or two on this in Economics classes.
We have several coefficients that are actually significant, which is a very good sign. Speed and tackling show a pretty good relationship between the attribute and high tackle totals. Surprisingly enough, Intelligence also proved to be pretty high, and significant as well.
Agility and strength are a little lower, but still lead to more tackles, however the coefficients are non significant, meaning that the actual coefficient could be higher or lower (I'm guessing higher).
Playing Time is negative, which I found very interesting. Being on the field more should lead to more tackles, but that isn't what the regression has shown us. The only regression in which the coefficient for playing time was not negative was with linebackers. If you smell something fishy, it's okay, because I do too. None of the coefficients for playing time were significant by a statistical view.
In conclusion, these regressions didn't tell us all that much about each attribute and its effect on total number of tackles. Most of the coefficients, and even a few of the f-stats turned out to not be statistically significant. More data points would most certainly be helpful, only 63 players and their attributes were included in the overall regression. The position group regressions included only 26 data points (players) on the high end, with linebackers only have 16. Larger sample sizes will show us better information, and will be possible as we go through more seasons.
If I get the motivation, I will go through and run these same attributes and their effects on other defensive statistics, like interceptions, sacks and what-not. I've gathered most of the data already, so it shouldn't be difficult, I'm just lazy.
Thanks for reading!
GRADED
[div align=center]
Gus T.T. Showbiz Player Page
Gus T.T. Showbiz Update Page
[div align=center]
[div align=center]
Bavitz Player Page Bavitz Update Page
Gus T.T. Showbiz Player Page
Gus T.T. Showbiz Update Page
[div align=center]
[div align=center]
Bavitz Player Page Bavitz Update Page