I was on my treadmill this morning, simultaneously trying to avoid getting fatter while also watching DraftBreakdown cutups of Bruce Ellington, when I saw an amazing game that he had against Wisconsin. He had 6 catches for 140 yards and two touchdowns. It occurred to me that I could probably do an apples to apples post looking at Ellington and some of the Big Ten receivers. The first receiver that I looked at was Cody Latimer, who happened to share another common opponent with Ellington in that they both played Missouri last year as well. How did Ellington do against Missouri? He only went 10 for 136 and 2 touchdowns.
Here’s the matchup table that breaks down how Latimer and Ellington fared against these two common opponents:
*MSYDSWINS is the number of times the player won the battle in the MSYDS category versus the common opponents. WINSOS is the average SOS of the player’s wins in the Yards battle (lower is tougher competition). Age is the average age that the games were played at for all 12 matchups.
Here are the full game logs:
It’s probably worth noting that this is just two games. But it’s also probably worth noting that some media pundits, in their quest to grade every prospect, don’t watch every game of every player. In fact I think it’s fair to say that some might stop at 2-ish games. So on one hand this two game sample is totally ridiculous and then on the other hand.. ok, it’s still ridiculous.
There’s also the issue that Ellington is more than a year older than Latimer.
But what is undeniable is that against these two opponents Ellington topped Latimer barely in Market Share and whipped him in the raw stats.
After I looked at these game splits, which were unsettling to say the least, I wondered how Ellington might do against other prospects and whether his superiority would translate. So I looked at Bruce Ellington versus Kelvin Benjamin.
Here are the full game logs if you’re interested:
You can see that Benjamin comes out either narrowly ahead (by Market Share), or ahead by a wide margin (by raw stats). The other interesting thing is that Ellington is not quite, but almost, as old as Benjamin is. So he’s an undersized and somewhat old prospect.
So based on this exercise Benjamin > Ellington > Latimer.
I’m half joking here. OK, I’m actually like 90% joking.
But I do think there’s an important point to be made.
Context matters a lot. There is a ton of latent information available from watching games. That context is the same thing that I often look for when going through numbers.
In fact on the issue of whether it’s possible to play fantasy sports by not watching any games I have an answer for me at least – it’s a no. I know because I tried with NBA DFS where I didn’t have time to watch games. I didn’t have any of the latent information that you can learn from watching games like “Player X left with an ankle injury” and that latent information takes a ton of research to replace. But game watching is also extremely problematic as a sole evaluation framework for reasons that I hope I hinted at in this post. I could easily construct a narrative surrounding the outcome of this exercise. I would say that Bruce Ellington shows ability to get open deep, has a nose for the endzone unusual for a player of his size, and has an extra gear that most receivers don’t. I could say that because he was amazing in the two games he played against opponents he had in common with Cody Latimer. Then I could probably come up with a separate narrative about the Benjamin/Ellington matchup.
And yet we know this is a ridiculous exercise because we are looking at such a small sample. There’s a ton of information being left out.
I promise there’s a point here. A number of the posts that I’ve been writing lately have been reflective of that never ending search to compile as much context as possible for purposes of evaluating the prospects. That’s why I’ve been doing the apples to apples posts, the report cards, and the perhaps pointless article on arm length. There are 100+ college football programs and it’s impossible to really know every player on every team. I’m data mining in some sense in that I keep going through various numbers looking for things or hoping that we might find something.
What would be much better is if we had a perfect record of every game, but we don’t. We have some play by play data that records a few stats for each play. It’s hardly a perfect record and if it were ideal we’d have 20 years of that data instead of about 7 years of data.
But that’s the problem with any football analysis. How do you know what the observations look like when they’re missing from your data set? And that’s the question that should probably keep every football analyst up at night, whether their framework du jour is stats or game watching.