DOES SIZE REALLY MATTER?
That’s the question on the minds of many. Some argue that bigger is better, while others believe that size is irrelevant, with other attributes being more important. Then, there are those who sit in the middle, convinced that as long as it’s not too small, it surely has its uses.
Today, we’re here to make an initial analysis of this age-old debate. I’ve gathered data from across the league—though let’s drop the double entendre here, LOL. What I’m really talking about is the weight of players in Madden NFL 08 for PC. Specifically, in this first phase, I’m analyzing the effect of weight on Offensive Linemen (OL) in pass protection. I hope I have the energy to explore other aspects and positions in the future.
To start, I looked at the distribution of sacks allowed by starters on the OL this season, categorized by their weight, both in general and by specific positions.
What did I find?
Most starting OL players weigh between 300 and 320 pounds and have allowed between 0 and 5 sacks so far this season. The trend line in the graph represents a linear approximation of the average sacks allowed per weight value. If the line has a noticeable slope, it indicates that weight may influence the result (weight on sacks allowed). The equation above the graph defines this line. The R² value measures how well the line fits the data, with values closer to 1 indicating a better fit. The equation has two components: the multiplier for the x-value (weight) and a constant. If the line is a good estimator, the multiplier can be interpreted as the number of additional sacks allowed per pound added to the player’s weight.
In this initial plot, the line is quite horizontal, suggesting that regardless of weight, the number of sacks allowed remains relatively consistent on average. This observation is backed by the numbers—the multiplier is very close to zero, meaning adding a pound doesn’t significantly impact the number of sacks allowed. Additionally, the R² value is close to zero, indicating a poor fit between the line and the data.
Next, I examined each OL position individually to see if any specific position showed a significant difference.
    
As we can see, the results are quite similar across positions. None of them have a high R², meaning the data doesn’t fit well with any of the trend lines. The lines are mostly horizontal, with two minor exceptions: Centers, who show a slight declining trend (indicating that each pound added results in 0.05 fewer sacks allowed, or 1 fewer sack for every 20 pounds), and Right Guards, which unexpectedly show an opposite relationship (suggesting that each pound added results in 0.04 more sacks allowed, or 1 additional sack for every 25 pounds). However, these conclusions are not recommended due to the low R² value for both.
I also analyzed the Pancakes/Sacks Allowed ratio, but since this involves a running block stat combined with pass protection, it’s not the focus of this article. I can, however, provide a sneak peek at those results at the end.
Conclusion? Not quite. Player stats aren’t everything—they’re just the numbers Madden generates in the game logs based on its internal calculations. Who really knows which OL player was responsible for a sack? Madden assigns blame, but it could have been a quarterback’s bad movement, multiple OLs, or even the wide receivers never getting open. There are many factors to consider, and while averaging helps account for some of these cases, I wanted to conduct a more controlled experiment. Instead of using sacks allowed, I wanted to measure how long the OL could hold off defenders until they sacked a non-throwing quarterback. This “Time to Sack” is our new focus.
Methodology
To carry out this experiment, I created a test roster using our league’s official roster, editing it so that every player had the average attributes for their position. For example, every quarterback had 94thp, 91tha, 89awr, and so on—the league averages for starting quarterbacks. For this study, I focused on the average weight for OL positions: LT = 314 lbs, LG = 318 lbs, C = 304 lbs, RG = 313 lbs, and RT = 314 lbs. Every player was identical, except I modified the OLs on two teams—making one team significantly heavier (400 pounds per OL) and the other much lighter (161 pounds per OL). This allowed me to compare these two teams against a third with average-weight OLs to see if there was any significant difference in their time to allow a sack. The teams used were the Redskins with 400-pound OLs and the Cowboys with 161-pound OLs, randomly selected.
To run these tests and measure the time to allow a sack, I used Practice mode. This way, I could repeat the same play over and over until I had enough repetitions to analyze the results. I controlled the offense for all three teams (49ers, Redskins, and Cowboys), keeping the quarterback stationary until a defender sacked him. The offensive play selected was SHOTGUN-5WRS SLANTS to ensure only the five OL were blocking, while the defense ran 3-4-OVER DBL X BRACKET with four defenders rushing. Half the repetitions were run normally, and half with the defensive play flipped. The entire process was recorded, and afterward, I timed each play from the snap (when the QB says “hut”) to the sack whistle by the referee.
Importantly, regardless of the team—whether the Redskins with 400-pound OLs, the 49ers with average-weight OLs, or the Cowboys with 161-pound OLs—they all faced the exact same defense, as all players were edited to have the same size and attributes (league starter average).
Results
So, what would we expect from the data? Intuition might suggest that the Redskins’ OL would last longer before allowing a sack compared to the 49ers and Cowboys, with the 49ers performing better than the Cowboys due to their heavier OL.
Here’s what the results showed:
These histograms of all three teams combined have the bell-shaped curve characteristic of a normal distribution, with very close averages, medians, and variances.
At first glance, the time for the Redskins (400 lbs) and 49ers (average) are almost identical, with just 0.01 seconds difference in the average. However, the Cowboys’ time is slightly lower. The variance and standard deviation are smaller for the Cowboys and 49ers, meaning their times were more consistent. The maximum time recorded for the Cowboys stands out as being lower. Interestingly, the number of strip sacks or fumbles followed the weight trend, with the lighter OL leading to more fumbles than the average-weight OL, and the 400-pound OL leading to the fewest.
However, these numbers alone can’t confirm or deny anything—we need to test them statistically to determine if the differences observed (0.01 or 0.14 seconds) are significant or just random variations. The easiest way to do this is by creating confidence intervals for each team’s average time. This uses the standard deviation of each sample and, based on a 90% confidence level, gives us a range for the inferred parameter. Essentially, this shows us where 90% of all results (not just the sampled ones) would likely fall. If two confidence intervals overlap, there is no significant difference between the two groups compared.
The confidence intervals for the three teams, with a confidence level of 90%, show overlap between 4.01 and 4.13 seconds. This means that, based on these results, there is no significant difference in the time to allow a sack between OLs of different weights.
Final Thoughts
I know this might be disappointing for those who believe size matters—and I’m one of them. I still believe size plays a role, but this isn’t definitive proof. We’re only testing one aspect of the game across five positions. Also, the testing conditions weren’t perfect—I had to control the offense manually rather than relying on CPU vs. CPU. The time from the start of the sack to the referee’s whistle (when I stopped the clock) might introduce some randomness that obscures true differences. The same goes for the sack animations—randomness in animation selection could add fractions of a second to some plays.
Nevertheless, I wanted to share the results I obtained, and more importantly, I’d love to hear your thoughts on what I could do differently to achieve more consistent results. Of course, any suggestions on making the testing and measurement process easier would also be appreciated.
As mentioned earlier, I plan to continue testing OLs, and the next step might involve exploring the importance of size for running plays. I haven’t finalized the methodology yet, so I’m open to suggestions. But as a teaser, here’s what I noticed while analyzing our league’s Player Stats for the season so far:
The ratio of Pancakes/Sacks Allowed among starting Offensive Linemen in the 2019 SFL season until Week 11.
The graph follows the same format as the ones shown earlier for sacks allowed alone. Each point represents a starting OL and their Pancakes/Sacks Allowed ratio according to their weight. For all positions combined, the trend line is nearly horizontal with a low R², indicating that weight doesn’t significantly impact the results.
    
For Left Tackles, the line has almost the same inclination, but with a higher avg, still, a lower R². For Left Guards and Centers it's possible to see a steeper line, with a multiplier of 0.358 and 0.339, respectively, that means that for each pound added the LG would get 0.358 more Pancakes/Sack Allowed and 0.339 for a Center, or for each 3 pounds added, a LG or a C would have one more PNK/SCK, but it still has a low R². For the Right Guards the line is odd, though, being inverted in relation to the rest of the OL. It shows that the heavier the RG, less Pancakes/Sacks Allowed he will have. And for Right Tackles the line is almost completely horizontal, meaning WGT doesn't affect at all on the PNK/SCK ratio.
Thanks for the reading, and I'll be waiting for suggestions.
|
