How do we find #CFB quarterbacks who will turn into solid #NFL starters? By looking at 20+ production variables, then adjusting for both strength of schedule & schematic tendencies. Trying something new here with a video breakdown for you. Hope you enjoy!

@RiseNDraft Great stuff Travis. People don’t appreciate how poor the environment for Burrow has been his whole career. Having Chase and Higgins off-sets some, but man, he does it with no scheme help. None. Straight just “have to make the reads and throws” every game.

@RiseNDraft Thanks! And haha yeah, Burrow was such a beast of a prospect when he came out. Do wish he’d have a more consistent top tier schemer driving the show with him

A lot of good QBs in the yellow though 🫣

Historically speaking the green is the far superior performing cohort. But there will always be some hits in yellow for sure. That’s how modeling works in everything

Yes because having plus athleticism is always nice for a QB. And they’re just arbitrary cut offs anyway. Guys like Mahomes are on the border of yellow with Lawrence/Williams. Them and the guys in yellow or below are the top QBs right now. NFL already drafts mobile QB high

They’re not arbitrary cutoffs. Solved cutoffs. It’s a probabilities game. The Green cohort has essentially has 4x the AV/EPA expectations over yellow historically. Yellow has some hits. Just FAR fewer. Orange basically has 1-2 hits of ~200 in the group. Red: 0.

The model is describing how mobile athletic QBs get drafted higher. So yes, on avg they will produce more. But like many models, this is more of a benchmark than correlation. Below red = do not draft. Green/yellow needs much more evaluation given the current group

Yeah that last part is pretty much accurate. Orange and red are both essentially hard pass unless you have generational “something” like Josh Allen. But even when drafted the expected production in the pros is MUCH higher typically for green vs yellow

I wonder if it’s a situation where you don’t necessarily want your QB at the extremes. High greens may have better peaks but I’m curious on longevity for duel-threat QBs like that. I’m curious if there’s sweet spot. Side note, seems no one adjusts for production drop since NIL

This is actually true in a lot of football-based modeling. Most viable metrics are fancy threshold tests, and crazy alien one-off prospect scores are often indicative of something less quantifiable in the profile

So basically I’m right and the Dolphins should draft Mensah

If his play holds up down the stretch that may become a no-brainer

Just out of curiosity, how does your model view his performance against FSU?

It did actually bring his profile back down to human levels. It was too silly to even post prior to this week

@AquaOrangeBlood To silly to post even though he’s played horrible opposition? I like him and Carr but it’s facts that those two have played nobody. I love weigman tho

@AquaOrangeBlood Yeah so this model adjusts plenty for strength of schedule. But saying Carr hasn’t played anyone is just wrong (BYU’s a top 5 team right now)

@AquaOrangeBlood Clearly your model isn’t adjusting enough. When all the players with the easiest schedules rise to the top it’s indicative of error. The truth is this can’t be quantified through any formula. Also you mention byu lmfao…. they lost???

@AquaOrangeBlood Oh boy. There’s too many things you’re wrong about all at once. Enjoy this weekend’s CFB games!

Good work! However, in your model it has J. Allen in the red and M. Willis in the yellow. That alone tells me there a flaw.

That’s not how modeling works. There are outliers. Josh Allen is essentially the only one thet found great success with a profile like his in this century haha. But good for him!

It’s essentially this: What odds to hit would you prefer…>50%, 12-15%, 1%, or 0%.

@itscachi Obviously for your model to have any value at all it needs to be 100% accurate. That’s how X works.

@itscachi

This is a Mestemaker post

This is awesome. What did you use to build the video?

Opus 5.5 and some little assistive scripts of my own on the video. Nasally voice is my own (dad of toddlers, “always kind sick” life)!

Love the visual! But again - the long-standing scouting quest to shove numbers into a machine and pull a lever to tell us ‘good or bad’ ‘do or don’t’ Make it 200+ variables! Maybe 2,000! Plenty of data. Still can’t sort thru the human variable.

😆 thanks! The model is my own personal work. Yay math! 🤓 I did shove it thru a machine to help with the presentation though. Trying to bridge the gap between nerds & normies, film & analytics. Because there’s value in both!

Have you considered throwing a measurables layer as well?

Yeah there’s a version that lassos a few more layers on top for more signal but I’ll get to that. Athletic testing is just noise at QB though. Height matters for capital for sure, but not much else physically

Yeah I meant height/weight, should’ve been clear. Super interesting to hear that about testing. Thanks for posting this is great stuff

Yeah what’s funny with QBs is sometimes you get bonkers testing from a QB whose style doesn’t mesh with that at all. And often the true dual threats with little passing acumen impress in underwear and then disappoint immensely (AR most recently)

Mestemaker is a Dolphin

Check out who’s on this in the “green” @gloryugapodcast

@OmarKelly check this out

Is there a way to follow this data in real time week to week through the college season?

Not yet! But I’m working on some things to make that possible for every position. NFL & CFB tools both