Statistical Analysis: Correlation between Combine Scores and Madden 17 Rookie Ratings

[zacification1]

Madden NFL 17 Statistical Analysis
Correlation Between Combine Scores and In-Game Ratings
Version 2.0
---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Idea Origin, Fundamental Questions, and Methodology

Spoiler

While creating the incoming rookie class on Madden NFL 16, I came into quite the dillemma when trying to rate players on a consistent basis. My philosophy was that if Player A and Player B ran the 40 Yard Dash in 4.44 seconds, their speed rating should reflect that, and be exactly the same. So, realizing how deep the operationsports.com community was when it came to roster creation, I figured I'd provide them and myself a tool to rate players fairly based on combine numbers. This would be done with a linear regression analysis, with the following attributes being listed with their real-life combine counterparts:

Speed: 40 Yard Dash (time in seconds)
Strength: Bench Press Drill (quantity of repetitions of 225 lbs.)
Jumping: Vertical Jump (height in inches)
Agility: 3-Cone Drill (time in Seconds)
Acceleration: 20 Yard Shuttle (time in seconds)

In order to calculate the correlation between these ratings, I needed a spreadsheet with the 2016 Rookie Class's NFL Scouting Combine scores on the individual drills, which can be found here. After using the Madden NFL 17 Player Ratings Database graciously provided to us by EA to input the correlating in-game attributes, I would use THIS online graphing calculator to prepare and present the mathematical correlations in graphic form after the insertion of the five bivariate sets of data gathered. In other words, da graphs, yo! Sections on all five ratings can be found below.

Speed

Spoiler

In Madden, a player's top speed is determined by their speed rating. The players with the fastest and slowest speed ratings in Madden NFL 17's base roster will be presented below in a graphic that also includes their 40 Yard Dash time.






As quite evident, the fastest player's all had the same speed rating, but ran different 40 Yard Dash times. What is similar about these group of players, is that their 40 Yard Dash times are considered outstanding compared to a typical 40 Yard Dash for their position, HB/WR. In Madden, speed tends to be valued quite a lot, since the ability to stretch the field is quite an advantage in a game that is somewhat far from simulation.

On the topic of mathematical correlations, here is the graph that represents thre correlation between the 2016 NFL Rookie's 40 Yard Dash scores and their Speed Ratings in Madden NFL 17.



Quite basic analysis of the graph will reveal a quite strong correlation between Speed and the 40 Yard Dash time. This is evident in the regression line's r value, which is approx. -.96. It seems that a majority of the data points are above the regression line in the left half of the graph, with the opposite being true for the right half. The image has some handy information in the top left corner, which will be broken down here:

-228 Rookies in Madden NFL 17 ran the 40 Yard Dash at the NFL Combine
-The average 40 Yard Dash for a rookie in Madden 1 time was a 4.81
-The average speed for a rookie in Madden 17 is 80
-The equation of the regression line is: f(x)=-32.16x+234.33

For those looking for charted data, here you are:

Strength

Spoiler

Strength, in Madden, is determined mainly off of the Bench Press drill, repetitions of 225 lbs. Strength, as a raw attribute, should theoretically factor into Pancake Blocks, run blocking, driving players forward, stiff arms, trucking, etc. As well as what scheme a player fits into and what type of player they are. For example, Madden may classify a Tight End with high strength as a "run blocking" Tight End. Here are the strongest and weakest players in Madden:




After analyzing the images, it appears that some more attributing factors to EA's rating of strength are things that usually require raw power, such as pocket disruption, stuffs, pushing back/off OLman, etc.

Let's look at the statistics:



After analyzing that graph, it appears that the data has a quite passable correlation. There is an obvious outlier, which is Daryl Worley, but other than that, the spread seems satisfactory. This is also indicated in the r value, which is .86. The equation for the regression line is quite simple, f(x) = x + 55. Very intuitive system for rating rookies on Strength, EA.

Charted data:

Jumping

Spoiler

Jumping, seemingly only inportant for how high a player can hurdle/jump, is determined by the NFL Scouting Combine's Vertical Jumping drill. Here are the jumping graphics:




Byron Jones was an absolute freak of nature in the 2015 NFL Scouting Combine.

Here is the raw data:



The data for Jumping here seems quite spread. In the left 1/2 of the graph, it seems that most data points are below the regression line. However, the right half displays just the opposite. The two x's at the bottom left of the graph displays some of the inconsistencies on EA's part with rating Rookies based on the vertical jump. The equation for the regression line goes as follows: y = 1.74x + 23.61. The correlation coefficient here is .824, which is close to the previous value, that of strength.

Charted Data:

Agility

Spoiler

Agility is defined by the quickness in which the body changes direction. The 3-Cone drill is a measurable of agility and is used as well in Madden when scouting players. The correlation is simple, the quicker the 3-cone, the higher the agility. Players with higher agility will be able to change direction quicker, wheher they are a ballcarrier or a defender. Understanding the difference between speed, agility, and acceleration is key in understand player ratings. Here are the graphics:




Here is the raw data:



Yuck. There appears to be A correlation, but that's it. In the left half of the graph, there is an abundance of data points above the regression line, the opposite it true for the other side. However, there may be a rationale behind this. EA could add/decrease from the Agility rating based on other change of direction and speed skills, such as Juking and Spinning. The correlation coefficient is -.69, which is quite. underwhelming

Charted data:

Acceleration

Spoiler

Acceleration is defined by the rate of change of velocity. Negative acceleration would indicate slowing down, positive acceleration indeicates speeding up. Madden players obviously accelerate, however, usually the objective is not to decelerate. Acceleration therefore in Madden terms is how fast it takes a player to reach their "top speed". Acceleration, as a measurable, is determined by the "Short Shuttle" or the 20 yard shuttle. Here are the graphics:




It is noteworthy that Marquise Goodwin is an Olympian long jumper.

Here is the raw data:



It appears that the data here is quite spread, but is definitely more consistent than the regression line was for Agility. Correlation coefficient here is -0.75, and it shows. The equation goes as follows:

(approx) y = -13.8x + 146.18 In which x = 20 yard shuttle score and
y= Madden rating for Acceleration

Not much else to say here. If you'd like tabled ratings for acceleration, they are provided below:

Conclusion

Spoiler

So, after finishing this statistical analysis, it was quite evident that when EA rates players, specifically rookies, they take into account more than just combine numbers. However, my goal was to provide a consistent ratings scale so that roster creators could rate players more fairly and consistently based on actual combine numbers. So, salute.

Possible Future Works

Spoiler

-Find Correlation between broad jump and jumping/strength, find a system to attribute this in rating players

-Find Correlation between 60 yard shuttle and acceleration, factor this into rating players on acceleration

-Correlate Wonderlic scores to awareness

-Correlate 10 yard split into a percentage of the Speed rating

-Test to see if stride length affects straightline speed and agility

[manEATgod]

Re: Statistical Analysis: Correlation between Combine Scores and Madden 17 Rookie Rat

I still don't understand why people want players based solely off of combine numbers when play speed is more important. Who cares about 40 time that much?

[Hooe]

Re: Statistical Analysis: Correlation between Combine Scores and Madden 17 Rookie Rat

This is interesting, thanks for doing this.

Out of curiosity - could you perform a regression analysis between ACC ratings and prospect 10-yard split times?

[T4VERTS]

Re: Statistical Analysis: Correlation between Combine Scores and Madden 17 Rookie Rat

I still don't understand why people want players based solely off of combine numbers when play speed is more important. Who cares about 40 time that much?

Fun fact, In talking with John White recently he told me players won't be a one for one in CFM draft classes this year. While the combine results will give you a general area the points will differ a couple points either way. His reasoning was there is a difference in track speed and football speed and they wanted to allow for that a bit. It is the same with other attributes, but that was the example he used.

[Hooe]

Re: Statistical Analysis: Correlation between Combine Scores and Madden 17 Rookie Rat

I still don't understand why people want players based solely off of combine numbers when play speed is more important. Who cares about 40 time that much?

"Game speed" isn't a real thing. A player doesn't have slower straight-line speed by some arbitrary amount or cut more sluggishly solely because he puts on a set of pads and a helmet.

What people refer to when they say "game speed" is technique and mental abilities. How well does a receiver disguise the stem of his route to make a great cut and break open? How well does a cornerback flip his hips from a backpedal to man coverage? How quickly does a free safety react to the quarterback tipping his shoulder up to launch a pass? None of these things affect how fast a player runs or how high he can jump, but they do affect the extent to which he can leverage his athletic abilities to succeed on the field.

A player with great athletic traits can afford to be a little slower mentally or a little bit less refined speaking to technique because he can make up for it with raw athleticism. Similarly, a player with less athletic ability can "play faster" than his measurable by having great technique and cognitive abilities; he recognizes and reacts to developing plays faster, so he can begin to put himself in position to make a play before his opponent does. In fact, a slower player with elite technical abilities can often beat the more athletic but less refined player; see Jerry Rice and his 4.71 forty as a prime example of this.

To that end, the physical ratings in Madden should be determined by measurable, and the technical ratings - Man Coverage, Block Shedding, Play Recognition, etc. etc. - should be used to determine "game speed".

[T4VERTS]

Re: Statistical Analysis: Correlation between Combine Scores and Madden 17 Rookie Rat

"Game speed" isn't a real thing. A player doesn't have slower straight-line speed by some arbitrary amount or cut more sluggishly solely because he puts on a set of pads and a helmet.

What people refer to when they say "game speed" is technique and mental abilities. How well does a receiver disguise the stem of his route to make a great cut and break open? How well does a cornerback flip his hips from a backpedal to man coverage? How quickly does a free safety react to the quarterback tipping his shoulder up to launch a pass? None of these things affect how fast a player runs or how high he can jump, but they do affect the extent to which he can leverage his athletic abilities to succeed on the field.

A player with great athletic traits can afford to be a little slower mentally or a little bit less refined speaking to technique because he can make up for it with raw athleticism. Similarly, a player with less athletic ability can "play faster" than his measurable by having great technique and cognitive abilities; he recognizes and reacts to developing plays faster, so he can begin to put himself in position to make a play before his opponent does. In fact, a slower player with elite technical abilities can often beat the more athletic but less refined player; see Jerry Rice and his 4.71 forty as a prime example of this.

To that end, the physical ratings in Madden should be determined by measurable, and the technical ratings - Man Coverage, Block Shedding, Play Recognition, etc. etc. - should be used to determine "game speed".

I think the argument falls apart because they start in sprinter stances and train for month with sprinting coaches on how to get out of the blocks. If they did it from the stance you start a normal play in for each position I'd agree with you.

[jfsolo]

Re: Statistical Analysis: Correlation between Combine Scores and Madden 17 Rookie Rat

Fun fact, In talking with John White recently he told me players won't be a one for one in CFM draft classes this year. While the combine results will give you a general area the points will differ a couple points either way. His reasoning was there is a difference in track speed and football speed and they wanted to allow for that a bit. It is the same with other attributes, but that was the example he used.

I know that some people are going to melt down when they discover this in the game, but I like this decision.

[Hooe]

Re: Statistical Analysis: Correlation between Combine Scores and Madden 17 Rookie Rat

I think the argument falls apart because they start in sprinter stances and train for month with sprinting coaches on how to get out of the blocks. If they did it from the stance you start a normal play in for each position I'd agree with you.

I don't see how it makes a difference. If anything, the initial burst for every player is going to be proportionally slower, since no one in the NFL starts in a sprinter's stance. Once players get up to speed, however - admittedly this rarely happens, given how few busted plays there are in the NFL - they are going to run however fast they timed at. If a player can't react to a developing play fast enough or doesn't have good enough technique to beat an opponent, however, 4.3 speed doesn't matter; he gets slowed down because of things other than his ability to run.

Example: Vikings WR Cordarrelle Patterson has struggled in the NFL despite his athleticism because he can't beat a press or make a clean route cut to get open. He's still very obviously very fast and has great vision and elusiveness as a ballcarrier - look at his success on kick returns and his 4.42 40 as very real proof of this - but his great athleticism doesn't matter when he plays receiver because his technique is so poor. He is every bit as fast as his times say he is and demonstrates that effortlessly when returning kicks, but his isn't a good enough football player to take advantage of his athletic gifts as a receiver.

[T4VERTS]

Re: Statistical Analysis: Correlation between Combine Scores and Madden 17 Rookie Rat

I don't see how it makes a difference. If anything, the initial burst for every player is going to be proportionally slower, since no one in the NFL starts in a sprinter's stance. Once players get up to speed, however - admittedly this rarely happens, given how few busted plays there are in the NFL - they are going to run however fast they timed at. If a player can't react to a developing play fast enough or doesn't have good enough technique to beat an opponent, however, 4.3 speed doesn't matter; he gets slowed down because of things other than his ability to run.

Example: Vikings WR Cordarrelle Patterson has struggled in the NFL despite his athleticism because he can't beat a press or make a clean route cut to get open. He's still very obviously very fast and has great vision and elusiveness as a ballcarrier - look at his success on kick returns and his 4.42 40 as very real proof of this - but his great athleticism doesn't matter when he plays receiver because his technique is so poor. He is every bit as fast as his times say he is and demonstrates that effortlessly when returning kicks, but his isn't a good enough football player to take advantage of his athletic gifts as a receiver.

There is a reason they pay sprinting coaches big money to teach them how to sprint at the combine. Guys who are better trained sprinters will look better at the combine than on a football field. Even in off coverage getting to speed from a sprinter stance and football stance are totally different things. What that means is a 4.3 guy at the combine may not be very different than a 4.4 guy except he had a better sprint coach.

[manEATgod]

Re: Statistical Analysis: Correlation between Combine Scores and Madden 17 Rookie Rat

I'd reply in depth but these guys are getting the job done instead. I definitely disagree with you and your opinion on game speed. Other guys in the thread have pretty much laid it all out though. Agree to disagree.

[zacification1]

Re: Statistical Analysis: Correlation between Combine Scores and Madden 17 Rookie Rat

I still don't understand why people want players based solely off of combine numbers when play speed is more important. Who cares about 40 time that much?

The main reason in me wanting combine-to-Madden consistency is so that when creating a roster, I can glance at a combine number, knowing it's straight-line speed, and give a fair rating. For more accurate rosters, yes, speed would differ slightly with players who ran the same 40 time, but still, it's the same 40 time regardless.

[zacification1]

Re: Statistical Analysis: Correlation between Combine Scores and Madden 17 Rookie Rat

This is interesting, thanks for doing this.

Out of curiosity - could you perform a regression analysis between ACC ratings and prospect 10-yard split times?

I could do it if you'd like. I let Google sheets do most of the work, it's the final production that takes the most time. I'll PM you if I get up to it. I'll probably take a few days, but will definitely look into doing it.

[zacification1]

Re: Statistical Analysis: Correlation between Combine Scores and Madden 17 Rookie Rat

Fun fact, In talking with John White recently he told me players won't be a one for one in CFM draft classes this year. While the combine results will give you a general area the points will differ a couple points either way. His reasoning was there is a difference in track speed and football speed and they wanted to allow for that a bit. It is the same with other attributes, but that was the example he used.

That actually excites me. I have screenshots in my arena (cheap plug) of two of my draft classes, and I got a ridiculous load of talent one year just based on decent ratings and good combines. Variability and unpredictability make franchise that much more exciting.

[Hooe]

Re: Statistical Analysis: Correlation between Combine Scores and Madden 17 Rookie Rat

There is a reason they pay sprinting coaches big money to teach them how to sprint at the combine. Guys who are better trained sprinters will look better at the combine than on a football field. Even in off coverage getting to speed from a sprinter stance and football stance are totally different things. What that means is a 4.3 guy at the combine may not be very different than a 4.4 guy except he had a better sprint coach.

Right; the reason a combine invite pays sprint coaches for training is to maximize his timed straight-line speed and initial burst. In turn this allows him to quantify himself as a better athlete than his peers and be seen as having a higher peak. Because, again, athleticism + technique = on-field speed.

Technique is what separates Harrison Smith from J.J. Wilcox. Both ran 4.57 40s with 10- and 20-yard splits within 0.01 seconds of each other at their respective combines. However, Smith can play up to his timed speed more often because he's mentally sharper, he reacts to plays faster, and he has more refined football technical skills. Meanwhile Wilcox just got torched again.

If technique isn't game speed, what do you think it is? Are you suggesting that players play faster than their peak times in gym shorts with all that athletic training?

[T4VERTS]

Re: Statistical Analysis: Correlation between Combine Scores and Madden 17 Rookie Rat

Right; the reason a combine invite pays sprint coaches for training is to maximize his timed straight-line speed and initial burst. In turn this allows him to quantify himself as a better athlete than his peers and be seen as having a higher peak. Because, again, athleticism + technique = on-field speed.

Technique is what separates Harrison Smith from J.J. Wilcox. Both ran 4.57 40s with 10- and 20-yard splits within 0.01 seconds of each other at their respective combines. However, Smith can play up to his timed speed more often because he's mentally sharper, he reacts to plays faster, and he has more refined football technical skills. Meanwhile Wilcox just got torched again.

If technique isn't game speed, what do you think it is? Are you suggesting that players play faster than their peak times in gym shorts with all that athletic training?

Technique is part of game speed, yes I agree with this. Part of your formula however, is not. Getting out of the blocks is one of the most important part of a 40 yet has no place in football. By using a 40 as gospel (40=x rating) you are starting with bad information as some guys can't figure out the start as well as others. So if you take two guys who are dead even standing up then put them in blocks one may have a decided advantage, however on a football field it doesn't exist.

We have kind of derailed the thread, so I'm going to leave it at that. I will say there is a reason going forward though draft classes generated in CFM won't have one for one speed to combine ratings.