09-05-2021, 03:32 PM
(This post was last modified: 10-07-2021, 09:55 AM by Tesla. Edited 2 times in total.)
Hello and welcome to my first shot at a statistical analysis.
After picking TE for my first sim league player ever, I wanted to see what made them tick in the sim and to see where I should be spending my TPE so I did this analysis.
TLDR; if you don't care about the journey and just want to see where I ended up, scroll to the very bottom and take a look at the attribute power rankings.
Limitations:
The main limitation of my analysis I would say is context. I'm not looking at the template of the player (vertical threat, possession, blocker), or if they have purchased traits. Also, I'm not taking into account the opposing defenses, individual defenders, who is throwing the ball to the TEs, or who is competing with the TEs for targets. A tight end with better attributes could have worse stats than a tight end with worse attributes due to his target share being lower or his quarterback being worse, etc. This is simply an analysis comparing attribute distribution and stats.
Methodology:
The 5 attributes I will be using in this analysis are speed, hands, strength, agility, intelligence. The 3 stats I will be using are yards, catches, and touchdowns.
So where did I start? I started by going through each box score of each game in S30 and getting the receiving stats for every tight end. Then I went through each tight end's update threads and recorded the attributes they had for each game of the season. This was pretty tedious so if you know of a better way I could do this for any future analysis, please let me know.
Findings:
So what did I find? Not much here to be honest. The highest r-square value I got for each stat are as follows: speed to yards 11%, agility to catches 12%, speed to touchdowns 1%. These are very low r-square values (50% r-square is a decent correlation, 70% good, 90% great) and we can't really assign any kind of correlation between attributes and stats based on these numbers; however, I do believe there is some information to be gathered here. Agility seemed to be the most beneficial for these stats and strength the least. Touchdowns being the stat that is the hardest to project (just like real life). Taking a look at the speed to yards distribution graph, it is pretty telling why these r-square values are pretty low:
So why did this particular analysis not give any kind of statistically significant data? These correlations are very low because I was looking at stats on an individual game basis. Each game, especially for tight end receiving stats, is highly variable. Maybe the tight end is playing against a very weak run defense so his team doesn't throw much and he has less stats for the game, but the next week the opponent's coverage linebackers are weak and he gets a lot of work. So while I think there is some information to be gathered from this initial analysis, I decided to take it one step further and get some less variable stats to look at.
Methodology part 2:
So what did I do next? I went through the past 4 seasons (S27 through S30) and took the receiving stats and attributes for each tight end that played all 16 games in the season.
Findings part 2:
Once again, this analysis wasn't super telling. For a lot of the attributes, most of the players had similar, round numbers not creating a good enough distribution to provide any conclusions. Here is an example of speed vs yards for the season.
So is there any takeaways from this one? I think there are a couple of minor takeaways. Looking at the distribution graphs, strength really doesn't have an effect on catches (3 of the lowest strength players had seasons in the top 25% of catches). The players with seasons with the lowest hands and agility respectively had a very low amount of catches and yards. But these aren't really earthshattering conclusions, just low attributes are bad. So I added another layer of complexity.
Methodology part 3:
So what did I change? Since the distribution graphs for the last method weren't really telling and players had similar, round numbers for their attributes (i.e. a lot of people had 85 or 90 speed, 80 or 85 strength), there wasn't really a good distribution of attributes for players to separate themselves from the pack. So I started added some of the attributes together to see if a combo of 2 attributes together was making an impact for these tight ends. Now I had 11 attribute categories to look at (strength, agility, intelligence, speed, hands, speed + hands, speed + int, speed + agi, hands + agi, hands + int, int + agi). I didn't consider adding strength to any other attributes due to how poorly that attribute performed in the previous methodologies.
Findings part 3:
So was this any better? YES THIS WAS FINALLY BETTER! The main takeaway from this method, quite surprisingly, is that intelligence is sneaky good for yards with agility being a close second. Touchdowns were still pretty volatile and I don't believe there are any attributes from this analysis that can lead to more touchdowns so I will only focus on yards and catches from here on out. While intelligence as a standalone attribute didn't pop in the graph, when added to the other 3 attributes it stood out a little more and all 3 graphs showed a distinguishable regression. Intelligence and agility combined had the highest r-square value out of all of the attribute combinations with a value of 46%. Hands was the least valuable attribute when it came to yards. On the flipside, agility was the most valuable attribute for catches and this became clear looking at the combined attributes. The regressions for catches weren't as solid as the regressions for yards, but the combination of hands and agility had the highest r-square value of 42%. Take a look at these 2 regressions:
Here are my post-analysis power rankings of these 5 attributes:
Yards
1. Intelligence
2. Agility
3. Speed
4. Hands
5. Strength
Catches
1. Agility
2. Hands
3. Speed
4. Intelligence
5. Strength
Overall
1. Agility
2. Intelligence
3. Hands
4. Speed
5. Strength
Weird right? Speed as the 4th best? One would think that speed would be important in getting open to catch passes and then racking up the YAC to get more yards. At the end of the day, speed still probably is the most important attribute. At the end of all of the analysis, I went back and looked at the distributions of these attributes across the recorded seasons for these tight ends. Of the 66 tight end seasons in this 4 year span, only 2 players had less than 80 speed, with the other 64 being between 80 and 90 speed. Each of the other 4 attributes had a much more normal distribution so that's why they popped more when doing this analysis. Basically, everyone had the same speed so it wasn't a good variable in the analysis. I think a more reasonable interpretation of this analysis is that you need to get your speed up to the level of the rest of the league (80-90) and then focus on agility and intelligence.
So here are my final power rankings of these 5 attributes after that caveat:
Overall
1. Speed
2. Agility
3. Intelligence
4. Hands
5. Strength
If you have read this far, thanks for sticking with it, please feel free to leave me feedback on the writing style, structure, etc. as I might do more of these in the future.
After picking TE for my first sim league player ever, I wanted to see what made them tick in the sim and to see where I should be spending my TPE so I did this analysis.
TLDR; if you don't care about the journey and just want to see where I ended up, scroll to the very bottom and take a look at the attribute power rankings.
Limitations:
The main limitation of my analysis I would say is context. I'm not looking at the template of the player (vertical threat, possession, blocker), or if they have purchased traits. Also, I'm not taking into account the opposing defenses, individual defenders, who is throwing the ball to the TEs, or who is competing with the TEs for targets. A tight end with better attributes could have worse stats than a tight end with worse attributes due to his target share being lower or his quarterback being worse, etc. This is simply an analysis comparing attribute distribution and stats.
Methodology:
The 5 attributes I will be using in this analysis are speed, hands, strength, agility, intelligence. The 3 stats I will be using are yards, catches, and touchdowns.
So where did I start? I started by going through each box score of each game in S30 and getting the receiving stats for every tight end. Then I went through each tight end's update threads and recorded the attributes they had for each game of the season. This was pretty tedious so if you know of a better way I could do this for any future analysis, please let me know.
Findings:
So what did I find? Not much here to be honest. The highest r-square value I got for each stat are as follows: speed to yards 11%, agility to catches 12%, speed to touchdowns 1%. These are very low r-square values (50% r-square is a decent correlation, 70% good, 90% great) and we can't really assign any kind of correlation between attributes and stats based on these numbers; however, I do believe there is some information to be gathered here. Agility seemed to be the most beneficial for these stats and strength the least. Touchdowns being the stat that is the hardest to project (just like real life). Taking a look at the speed to yards distribution graph, it is pretty telling why these r-square values are pretty low:
So why did this particular analysis not give any kind of statistically significant data? These correlations are very low because I was looking at stats on an individual game basis. Each game, especially for tight end receiving stats, is highly variable. Maybe the tight end is playing against a very weak run defense so his team doesn't throw much and he has less stats for the game, but the next week the opponent's coverage linebackers are weak and he gets a lot of work. So while I think there is some information to be gathered from this initial analysis, I decided to take it one step further and get some less variable stats to look at.
Methodology part 2:
So what did I do next? I went through the past 4 seasons (S27 through S30) and took the receiving stats and attributes for each tight end that played all 16 games in the season.
Findings part 2:
Once again, this analysis wasn't super telling. For a lot of the attributes, most of the players had similar, round numbers not creating a good enough distribution to provide any conclusions. Here is an example of speed vs yards for the season.
So is there any takeaways from this one? I think there are a couple of minor takeaways. Looking at the distribution graphs, strength really doesn't have an effect on catches (3 of the lowest strength players had seasons in the top 25% of catches). The players with seasons with the lowest hands and agility respectively had a very low amount of catches and yards. But these aren't really earthshattering conclusions, just low attributes are bad. So I added another layer of complexity.
Methodology part 3:
So what did I change? Since the distribution graphs for the last method weren't really telling and players had similar, round numbers for their attributes (i.e. a lot of people had 85 or 90 speed, 80 or 85 strength), there wasn't really a good distribution of attributes for players to separate themselves from the pack. So I started added some of the attributes together to see if a combo of 2 attributes together was making an impact for these tight ends. Now I had 11 attribute categories to look at (strength, agility, intelligence, speed, hands, speed + hands, speed + int, speed + agi, hands + agi, hands + int, int + agi). I didn't consider adding strength to any other attributes due to how poorly that attribute performed in the previous methodologies.
Findings part 3:
So was this any better? YES THIS WAS FINALLY BETTER! The main takeaway from this method, quite surprisingly, is that intelligence is sneaky good for yards with agility being a close second. Touchdowns were still pretty volatile and I don't believe there are any attributes from this analysis that can lead to more touchdowns so I will only focus on yards and catches from here on out. While intelligence as a standalone attribute didn't pop in the graph, when added to the other 3 attributes it stood out a little more and all 3 graphs showed a distinguishable regression. Intelligence and agility combined had the highest r-square value out of all of the attribute combinations with a value of 46%. Hands was the least valuable attribute when it came to yards. On the flipside, agility was the most valuable attribute for catches and this became clear looking at the combined attributes. The regressions for catches weren't as solid as the regressions for yards, but the combination of hands and agility had the highest r-square value of 42%. Take a look at these 2 regressions:
Here are my post-analysis power rankings of these 5 attributes:
Yards
1. Intelligence
2. Agility
3. Speed
4. Hands
5. Strength
Catches
1. Agility
2. Hands
3. Speed
4. Intelligence
5. Strength
Overall
1. Agility
2. Intelligence
3. Hands
4. Speed
5. Strength
Weird right? Speed as the 4th best? One would think that speed would be important in getting open to catch passes and then racking up the YAC to get more yards. At the end of the day, speed still probably is the most important attribute. At the end of all of the analysis, I went back and looked at the distributions of these attributes across the recorded seasons for these tight ends. Of the 66 tight end seasons in this 4 year span, only 2 players had less than 80 speed, with the other 64 being between 80 and 90 speed. Each of the other 4 attributes had a much more normal distribution so that's why they popped more when doing this analysis. Basically, everyone had the same speed so it wasn't a good variable in the analysis. I think a more reasonable interpretation of this analysis is that you need to get your speed up to the level of the rest of the league (80-90) and then focus on agility and intelligence.
So here are my final power rankings of these 5 attributes after that caveat:
Overall
1. Speed
2. Agility
3. Intelligence
4. Hands
5. Strength
If you have read this far, thanks for sticking with it, please feel free to leave me feedback on the writing style, structure, etc. as I might do more of these in the future.