I dig into the topic of draft capital on today’s (very nerdy) episode of the Late-Round Podcast. open.spotify.com/episode/4wyq...
10.03.2026 11:32
👍 5
🔁 2
💬 1
📌 0
I dig into the topic of draft capital on today’s (very nerdy) episode of the Late-Round Podcast. open.spotify.com/episode/4wyq...
When I was in on Kenneth Walker last year, I really meant this year.
11 days until the Late-Round Prospect Guide drops. Profiles are written. Editing is happening. We're getting close.
- Info on the ZAP Models
- Profiles for every WR/RB/TE at the Combine
- Year 2 player profiles
LateRound.com before the price increases by 5 bucks.
We'll be live in a half hour to talk about some high-level NFL Combine takeaways, news items, and more. www.youtube.com/watch?v=YcoE...
I don’t know for sure if it’s a thing or not — if we can confidently say a good draft would drive down costs, since it’s deep. Too much noise to something like that to actually model for IMO.
If you have any questions for this week's mailbag episode, send them my way.
Yes, players are getting older. I think we still have a little bit of leftover from the COVID season, though, too. Likely will regress a tad.
9. Adjust for program strength in some way. A wide receiver playing at Ohio U is going to have an easier time compiling compared to one playing at Ohio State.
Thanks buddy
Oh, one final thought: I do not have all the answers. I'm just a guy in mom's basement trying to do my best to help fantasy football teams. Be open-minded about this stuff, because no one has this shit solved.
I think those are my thoughts for now. I probably have more. I've just been really deep into this prospect guide and thought, "You know what? Maybe I'll put a thread together that 14 people care about."
8. To this point: Have a mission. If you're testing your data's predictiveness, see if it's more predictive than something that's readily available, like draft capital. Otherwise, what's the point?
7. R squared analysis is almost always lacking context. I've had people ask me about my model's R^2, as if there's a point of comparison. You can make R^2 say what you want it to say based on the sample you're analyzing. You need a reference point.
(Continued...) This isn't to say that you couldn't have spotted Nacua. He had good analytical traits. This is saying that the smaller your sample, the more a player like Nacua will influence how your model behaves.
6. It's easy to overfit. You should have robust samples to model off of. When you don't, it's very easy for those models to overfit — to see a hit like Puka Nacua and then build the foundation around that archetype. That hurts in the long run, since he's an outlier.
5. Beware of threshold analysis. Using thresholds to bucket players can be helpful for explaining your stance on a player, but it shouldn't be treated as gospel. And those thresholds are often driven by a particular player's numbers, so the sample captures players who are worse.
4. One data point isn't everything. It's not very scientific to totally fade a player because he misses on one singular metric. Conversely, you shouldn't target a guy because of one. Using analytics to scout is like putting together a puzzle — there are tons of pieces.
3. Age is not just about being old in the NFL. Players who leave college early are generally doing so because they're more talented. Players who stick around for a while typically do so because they're not good enough to leave early. That's a talent signal.
2. Age adjustments are everything. I've seen tweets saying someone like KC Concepcion must be hated by analytics folks because his career yards per route run isn't fantastic. That's not true. He produced immediately as a Freshman and should look fine in most tested models.
1. Raw statistics lack meaning. Collegiate offenses see a wider range of style, which means there's a wider range of pass rates and personnel deployment. Raw receptions, raw yards — they need context. Converting to shares is almost always more valuable.
I've seen a lot of analytical draft analysis on the timeline over the last few weeks, which is great and awesome. It's a fun time of year. But having done this for some time (I am unc), I want to call out some things to watch out for on the numbers-driven side.
I talk about NFL Combine winners and losers on today’s episode of the Late-Round Podcast. open.spotify.com/episode/3AUM...
I use speed score in the RB model, yeah
Speed Scores for the running backs who ran their 40s. These are official times via NFL.com.
Mike Washington has the second-highest Speed Score in the ZAP Model database, which goes back to 2011. Keith Marshall only beat him by 0.05.
That unofficial 4.33 for Mike Washington = 126.88 Speed Score. That's just behind Keith Marshall (126.93) for the best in the ZAP Model database. (Again, it's not official)
Loved chatting with @dynogametheory.bsky.social on today's Late-Round Perspectives episode. Scott gives his take on where dynasty managers go wrong, and we talk a whole lot about the RBs and WRs in this year's class.
Listen: podcasts.apple.com/us/podcast/t...
Watch: www.youtube.com/watch?v=c8zO...
Now the Arian Smith pick makes sense bsky.app/profile/awfu...
We'll be talking news, Year 2 wide receiver comps, and more on today's Late-Round Show. You can tune in live at noon Eastern right here (and I won't talk about greek yogurt this week): www.youtube.com/watch?v=OC0D...
I don’t remember a season in recent history where there’s been such a disconnect between what the numbers are saying and what draft analysts and the media are saying about a wide receiver class. There are some guys who are being talked about as top-50 picks and I’m like “Him?”