This is my first attempt at an experiment using the simulator. I am particularly interested in the effect of player attributes conditional on other attributes. I will begin with a little background into how I approached my analysis followed with my results and implications. I am very interested in hearing feedback from those that have experience completing these types of statistical analyses.
I will begin by saying that the sample size for this study is 72. Though I have noticed that some people complete hundreds of simulations in their analyses I was unable to for this project. The problem that this could cause is imprecision. A larger sample size would likely lead to more significant results and less “noise” in the data. However, a general rule of thumb is that the necessary sample size to satisfy a large enough sample is 30. That being said, let’s dive into my study.
I am interested in how the attributes intelligence (intel) and strength affect two different dependent variables for a linebacker. The first I call “make a play”. Make a play is the combined number of tackles and pass deflections that a player completes in a game. For example, if make a play was 7 for a player, that player had any combination of tackles and pass deflections that added up to 7. The second dependent variable is “big play”. Big play is whether or not the player completes a big play in a game. A forced fumble, fumble recovery, interception, sack, block, or tackle for loss are considered big plays. If the player completed any or multiple of these in a game it is coded a 1, otherwise it is 0.
Basically, I am looking at how intel and strength affect the number of plays made by a linebacker, and whether or not a big play is made by a linebacker, in a game, holding all else constant. Holding everything else constant is important. In this experiment the only adjustments made throughout the simulations are the values of intel and strength.
Below I have a table and four graphs that show my results. Here I explain my interpretation of the results, but I welcome everyone’s thoughts. The effect of strength on the number of plays made increases as intel increases (and vice-versa). This means that to make strength or intel more effective in terms of the player making more tackles/pds in a game, then the other should be improved as well. This is shown in graphs 1 and 2. It is important to note that these results are not statistically significant; however, I tally this up to a small sample size. I am confident that these results would be more significant otherwise.
The more interesting result is the effect of strength and intel on whether or not a big play is made. Independently, an increase in either strength or intelligence positively affects the likelihood of the player making a big play in a game. Interestingly, as the other attribute increases, the effect of the other decreases. As seen is graph 3, as intelligence increases, strength has less of an impact on whether the player will make a big play. In graph 4, as strength increases, intelligence has less of an impact on whether the player will make a big play.
The implications are as follows. The impact (marginal effect) of strength/intel on the likelihood of making a big play in a game decreases as the other attribute increases (graphs 3 and 4). Alternatively, as the impact (marginal effect) of strength/intel on the number of plays (tackles/pds) made by a player increases as the other attribute increases; however, these results are not statistically significant.
I have future statistical analyses planned, but first I need to figure out how to create larger-n studies as I expect that will be the biggest flaw that people will point out. When I have fixed this problem I plan to post another study with the new numbers. Additionally, this will allow me to look at the necessary number of simulations for an adequate study. Down the road I want to look at other conditional relationships and the effects of attributes on individual player/position performance. I am open to all feedback, suggestions, and question on this article.
I will begin by saying that the sample size for this study is 72. Though I have noticed that some people complete hundreds of simulations in their analyses I was unable to for this project. The problem that this could cause is imprecision. A larger sample size would likely lead to more significant results and less “noise” in the data. However, a general rule of thumb is that the necessary sample size to satisfy a large enough sample is 30. That being said, let’s dive into my study.
I am interested in how the attributes intelligence (intel) and strength affect two different dependent variables for a linebacker. The first I call “make a play”. Make a play is the combined number of tackles and pass deflections that a player completes in a game. For example, if make a play was 7 for a player, that player had any combination of tackles and pass deflections that added up to 7. The second dependent variable is “big play”. Big play is whether or not the player completes a big play in a game. A forced fumble, fumble recovery, interception, sack, block, or tackle for loss are considered big plays. If the player completed any or multiple of these in a game it is coded a 1, otherwise it is 0.
Basically, I am looking at how intel and strength affect the number of plays made by a linebacker, and whether or not a big play is made by a linebacker, in a game, holding all else constant. Holding everything else constant is important. In this experiment the only adjustments made throughout the simulations are the values of intel and strength.
Below I have a table and four graphs that show my results. Here I explain my interpretation of the results, but I welcome everyone’s thoughts. The effect of strength on the number of plays made increases as intel increases (and vice-versa). This means that to make strength or intel more effective in terms of the player making more tackles/pds in a game, then the other should be improved as well. This is shown in graphs 1 and 2. It is important to note that these results are not statistically significant; however, I tally this up to a small sample size. I am confident that these results would be more significant otherwise.
The more interesting result is the effect of strength and intel on whether or not a big play is made. Independently, an increase in either strength or intelligence positively affects the likelihood of the player making a big play in a game. Interestingly, as the other attribute increases, the effect of the other decreases. As seen is graph 3, as intelligence increases, strength has less of an impact on whether the player will make a big play. In graph 4, as strength increases, intelligence has less of an impact on whether the player will make a big play.
The implications are as follows. The impact (marginal effect) of strength/intel on the likelihood of making a big play in a game decreases as the other attribute increases (graphs 3 and 4). Alternatively, as the impact (marginal effect) of strength/intel on the number of plays (tackles/pds) made by a player increases as the other attribute increases; however, these results are not statistically significant.
I have future statistical analyses planned, but first I need to figure out how to create larger-n studies as I expect that will be the biggest flaw that people will point out. When I have fixed this problem I plan to post another study with the new numbers. Additionally, this will allow me to look at the necessary number of simulations for an adequate study. Down the road I want to look at other conditional relationships and the effects of attributes on individual player/position performance. I am open to all feedback, suggestions, and question on this article.