I want to address a topic debated every year about incoming rookies, usually with incredibly lazy analysis. Wide Receiver Height and Weight. It’s seemingly been researched thoroughly, but I’m not sure previous research has done the topic justice. The common, lazy “analysis” clickbait tweet goes along the lines of:
“We’ve NEVER seen a successful WR that’s less than *insert arbitrary threshold of height or weight here*.“
The problem with this analysis is A) arbitrary thresholds are chosen, and B) the pure volume of small WRs (determined by height, weight, or BMI) is smaller than the number of average-sized WRs. Funnily enough, if you go high enough, you can also say the opposite “We’ve NEVER seen a successful WR that’s bigger than *insert arbitrary*” Why is that? The same issue – it’s looking at the margins where the number of players on the other side of the threshold is less. If thresholds are going to be used, success should be judged by success rates, not the volume of success, and where possible, the groups to be roughly equal. Better yet, why distill a continuous variable (a list of player’s heights/weights/BMI) into a binary variable (did a player meet a threshold) or into a categorical variable (buckets of different heights/weights/BMI) if we don’t have to? (We definitely don’t have to, far from it).
You can see from the above three charts that most WRs who were drafted between 2010-2021 are between 5’10” and 6’3” tall, weigh between 185ish-220ish lbs, and have a BMI between 26.5-28.5. Therefore, it’s much more likely a WR who is successful in the NFL will land between those ranges, simply because it’s much more likely any randomly selected WR in the NFL will land between those ranges.
I then trimmed the sample for our analysis of WRs to the below:
- Drafted Rounds 1-3: This is where the majority of the successes come from and the biggest area in which player debates are centered around
- Draft years 2010 to 2019 – This allows the players to have had three seasons of play under their belts, so I could use Seasons 1-3 PPR points per game as an outcome measure.
- Had a combine height or weight – if they had both, their BMI was calculated too.
This resulted in a sample of 98 WRs. I then calculated the median (50th percentile) and interquartile range (amount between the 25th-75th percentile) of these WRs to put them into buckets, as the people seem to love themselves some buckets. I did round up or down to some whole-ish numbers, so the number of players in each bucket isn’t exactly equal.
Cross-bucket Comparisons
Our bucketed results:
Height
- Smallest PPR PPG for the 71-72 inch bucket
- Similar PPR PPG and similar Top 12 and 24 hit rates for the 70 inch and below bucket and the 73-74 inch bucket
- The shortest three buckets have similar hit rates for the Top 12 and Top 24 seasons.
- There appears to be a higher hit rate and a higher PPR PPG for 75 inches and above.
- It doesn’t seem to make sense to run with these results and recommend avoiding wide receivers who are 71 and 72 inches tall?
Weight
- Increase in PPR PPG as the weight buckets increase
- The lowest hit rate for the Top 12/24 seasons for the 190-204lbs bucket
- Again, appears to be a higher hit rate and a higher PPR PPG for the biggest bucket of 215lbs and above
BMI
- Lowest PPR PPG for the smallest bucket
- The other three buckets are relatively similar
Face value takeaways from this section: looks like the biggest cohort consistently had the higher Top 12 hit rates especially. On the flip side though, seems like the short king bashing might be unfounded so far. Maybe it’s a good thing Wan’Dale is 68 inches and not the perhaps doomed bucket of 71-72 inches? Should we be more worried that he’s two inches below 70 inches or would we be more worried if he was two inches above 70 inches? Also, he’s in the third BMI bucket, which has the highest Top 12 hit rate and second highest PPR PPG. Apart from the biggest Height and Weight buckets, it seems difficult to differentiate players based off their Height and Weight.
The WRs 70 inches and shorter who hit a Top 24 season in this sample?
- Brandin Cooks
- T.Y. Hilton
- Golden Tate
- Tyler Lockett
- Diontae Johnson
- Randall Cobb
- Marquise Brown
- Kendall Wright
Unfortunately, face-value takeaways, and small sample size traps like the tables shown above, are common issues the fantasy football community run into. Multiple pair-wise comparison is a statistical method to compare continuous variables between groups – in this case, Seasons 1-3 PPR points per game. In essence, it uses the same data in the above tables, and compares all the groups with each other to see whether there is a statistically significant difference between the groups, or if the presented means of the Seasons 1-3 PPR points per game potentially aren’t significantly difference from each other once you take into account the number of players in each group. The higher amount of players in each group, the more confident the mean is a true representation of that group. The results of this type of analysis? No significant difference between the groups for Height, Weight, or BMI.
Linear Regression Testing
The above plots are all those WR weights and heights plotted against their PPR points per game for their first three seasons. The yellow line indicates the trend, which we see slopes upwards. As the player’s weight or height has increased, so has their average points per game. Don’t pay too much attention to how sloped the lines are, though, as the visuals can be manipulated depending on the scale given to the chart. We can see that the height chart has an R^2 (R squared) value of 0.014, which means, in this sample height has accounted for 1.4% of the variance in S1-3 PPR PPG. The P value needs to be less than 0.05 for it to be considered a “significant” or “probably true in 95% of cases” finding, which it is not. For the weight chart, the R^2 is 0.032 (accounts for 3.2% of the variance), and the P value remains above 0.05. For BMI, the R^2 is 0.015 (accounts for 1.5% of the variance), and the P value remains above 0.05 still.
So what does this mean exactly? It means that weight, height, and BMI probably aren’t significant factors in determining a WR’s first three seasons of PPR PPG, at least when tested in a linear fashion.
Variable Importance Plots
We can look at the importance of Height/Weight/BMI by feeding them into a multivariate linear regression model alongside whether a wide receiver is an early declare or not and their draft position. Again we’ve stuck to Rounds 1-3 only for consistency’s sake. I haven’t included any college production metrics here as there can be debate about which college production metrics are most important. I’ve then charted their variable importance plots, which is, as the name suggests, plot the importance of each variable in the model onto the chart.
As you can see, for all three models, none of the Height/Weight/BMI appears to be very important.
So, does it matter if my favorite rookie WR is small?
Even on its own, size doesn’t appear to matter, and that amount is reduced even further once we also consider draft capital and whether the player was an early declare. It’s possible that with a larger sample size of similar results, then the biggest wide receivers (top 25th percentile) may reach a point where we can more confidently say they consistently return better results compared to the three-quarters of the rest of the cohort. So, if you’ve got two prospects you think are pretty even in all other areas, it’s reasonable to use your choice of height or weight as a tie-breaker. This does not mean, though, is discounting small prospects with solid draft capital just because they don’t meet a height/weight/BMI threshold set by someone on Twitter who thinks they know what they’re talking about but probably doesn’t.
It’s okay to consider height/weight/BMI as part of the jigsaw puzzle that is WR prospecting, but the amount it matters is relatively small. If you find the final piece that fits perfectly into your jigsaw puzzle, then great, the puzzle of Mt Fuji looks exactly like Mt Fuji! However, perhaps you lost a piece and had to jam a slightly off-colored or misshapen piece into the puzzle to finish – it’s not perfect, but it still looks like Mt Fuji, and we shouldn’t be now comparing it to the small hill at the end of your street that sometimes gets a little bit of snow on it.
Long live King Wan’Dale.
