32 53s: Most Surprising Cuts of 2025

A look at the players my 53-man prediction model had the most confidence in that ultimately failed to make an initial roster and whether my model should have seen it coming

Yesterday, I shared a summary of my inaugural 53-man prediction model’s performance in 2025. Overall, I was pleased with how accurate the model was. Somewhat surprisingly, my modeled predictions did about as well as ESPN’s team of 32 NFL Nation beat writers.

Despite that, there is still room to improve the model. In service of that idea, I did a deep dive into the 26 players who my model tagged as “locks” (>99%) or “near locks” (90% to 99%) who were not on a roster on the day of final cuts.

This doesn’t constitute a wholistic view of potential model improvements, but given that we often have a lot of information about the kinds players that I placed in these categories (as opposed to the kinds of players with <35% to make a roster who are frequently practice squaders or UDFAs), this may represent a larger share of the low hanging fruit I can use to better calibrate the model. On top of that, categorizing this group of players is key to inspiring confidence in the model. Nearly two-thirds of players on initial 53-man rosters received an estimate of >90% from my model, so these tend to be key players that set the stage for the larger pool of remaining players who are legitimately fighting for the final third of roster spots.

With that out of the way, here are the cuts my model deemed most surprising (and whether or not they actually should have been).

Christian Wilkins (LV, IDL)

Wilkins was shockingly released by the Raiders on July 24th, despite the fact that he’d already earned $20m in the offseason. The Raiders are currently attempting to claw back that money, so this is an extremely rare case.

Josey Jewell (CAR, LB) & George Odum (SF, ST/DB)

Both Jewell and Odum were released in the lead up to training camp due to concerns about injuries. My understanding at the time of my prediction was that Jewell had $2.1m in salary guarantees and that Odum had $1m vest to a full guarantee in the spring. Odum is expected to eventually sign with a team, so it is possible the 49ers could benefit from an offset. Without considering Odum’s vesting, my model likely would have place him in the “Strong” group (65% to 90%).

Morgan Fox (ATL, IDL), Nick Niemann (HOU, LB), Marquez Valdes-Scantling (SEA, WR)

These three players each took on a $1.5m signing bonus on top of some salary guarantees, pushing them into the “lock” category for my model. These are rare cases and I’m not sure if there is anything that distinguishes them.

Shy Tuttle (CAR, IDL)

I beefed this one. Tuttle signed a renegotiated contract in the spring, but I modeled him as if he collected a $1m roster bonus in March. He probably would have been around the bubble if I’d identified him correctly.

Kyle Trask (TB, QB) & Taylor Heinicke (LAC, QB)

Trask was my last of eight misses on locks and Heinicke was my highest miss in the “near lock” category. Trask is notable in that he did not receive a signing bonus, even though he did have $1.33m in salary guarantees. If those salary guarantees offset, Trask will ultimately leave almost no dead cap if he’s picked up by another team on the active roster for all of 2025. Perhaps my model should distinguish guaranteed salary below the threshold of a player’s minimum salary as less meaningful that the dollars above.

Bradyn Swinson (NE, ED) & Chris Paul (LAR, LB)

These two were the only players from the 2025 draft that failed to make a roster after I gave them a 90% chance to do so. Paul was right on the cusp of that 90% threshold, but the Rams lack of off-ball LB depth pushed him over the top.

Cole Strange (NE, IOL) & Layden Robinson (NE, IOL)

My model gave modest downward adjustments to both players due to log jammed roster in New England, but it wasn’t enough. Not a surprise, but teams undergoing leadership changes are more likely to bring in more players. That may be something I could pick up in the model on the front end.

Carson Steele (KC, HB)

My model credited Steele as a starter because OurLads.com had him listed as the starting fullback. Had he been properly credited as a third-stringer or worse as a running back, he likely would have been on the bubble.

Mike Pennel (KC, IDL) & Otis Reese (TEN, LB)

Both players earned “near lock” status by credited as starters before camp. Pennel went on to start KC’s first two preseason games as well.

Malachi Corley (NYJ, WR) & Hendon Hooker (DET, QB)

Corley barely played as rookie and recorded five healthy scratches to boot. He was grouped as a “near lock” exclusively because of his being drafted early in the third round a year and a half ago. Like Corley, Hooker’s prediction was similar based on his status as a former early third round pick.

Zack Moss (HB, CIN) & Cordell Volson (IOL, CIN)

My model was deferential here to Moss & Volson being a returning starters from last season and receiving a renegotiated signing bonuses of $375k & $500k respectively.

Robert Woods (PIT, WR)

Woods signed a deal with a non-guaranteed league minimum salary with a $745k signing bonus on top of it. He was also listed as a starter at the time of my prediction.

Matt Hayball (NO, P) & Cal Adomitis (LS, CIN)

Hayball was considered the favorite in his position battle, but was released before the preseason opener. Adomitis lost his position battle in final cuts.

Aaron Brewer (ARI, LS), Charley Hughlett (PHI, LS), Thomas Morstead (SF, P)

These three specialists hardly count as misses, as they were each cut and re-signed the following day.