32 53s: Projecting Every 53-Man Roster
No one has ever dared to try it (as far as I know)
Over the course of this week, I’m going to be unveiling a big new project that I’ve been chipping away at over the course of the summer. This piece will serve to provide some background on this work.
I’ll update the list below with links as my projections for individual teams go live:
What exactly am I doing?
I’m going to estimate each active roster NFL player’s probability of landing on a 53-man roster on the day of cutdowns. I’m going to be ignoring the possibility that a presently healthy player could land on PUP, IR, or an exempt list that wouldn’t count them against the 53-man total. I’m also making my prediction irrespective of the team the player ultimately ends up on (if they make it through final cutdowns), so trades are irrelevant for my purposes. In essence, this reduces my prediction to two outcomes: either a player makes the team, or they are cut.
What will my predictions look like?
I’m going to place each of the 2,902 players currently on active rosters (the Rams have one open roster slot and 23 teams are utilizing a bonus roster spot from the International Player Pathway Program) into one of six buckets based on about a dozen models plus a few of my own heuristics:
Lock (>99% chance to make an initial 53 man roster)
Near lock (90% to 99%)
Strong (65% to 90%)
Bubble (35% to 65%)
Dark Horse (15% to 35%)
Longshot (<15%)
As I mentioned above, I’m not predicting for outcomes where a currently healthy player lands on an injured or exempt list through final cuts. Historically, teams will eventually send an average of 6.5 players from their July 90-man roster to those lists, so my probabilities tend to add up to ~57 to cover for those players who will succumb to injuries through the preseason. It looks a little weird, but it works.
While I think the probabilistic estimates are the most interesting part of this project, I’m also going to rank the top 53 players (along with the next 5 players out) for a more traditional roster projection.
What data am I using?
I’ve collected over a decade worth of data about the 90-man rosters that teams have taken into training camp. This data includes:
Transactions - from The Football Database & Spotrac
Contracts - from OverTheCap & Spotrac
Depth Charts - from OurLads
Injury Reports - NFL (via nflverse R package)
I won’t go into great detail about my entire process, but once I’ve separated players into categories, I apply the information above in different ways. The player categories are as follows:
Vested players who remain on same contract as last season
Players remaining on draft contracts that have never been waived
Non-vested players who remain on same contract as last season
Futures contract players
Vested players on newly signed contracts
Non-vested players on newly signed contracts
Restricted Free Agents
Players acquired in trades during this offseason who have not signed a new contract
Rookie draft picks
Rookie UDFAs
Will my predictions actually be good?
I don’t expect my model to perform better than an experienced, well-positioned beat writer. That being said, my model may have a more reliable perspective on the outlook for players in more standard situations. My model will struggle most with players who have something that makes them unique, whether that is their contractual situation, their on-field utility, their recent injury history, or the pathway they took to their current team. For now, I’m not going to be putting a thumb on the scale when I disagree with the model. Ultimately, a major benefit in creating a formalized model is it can control for the things we think we know, freeing us to spend more effort on understanding edge cases.
Why should you care about rosters at the 53-man cutdown?
It’s a common refrain among beat writers that the first 53-man roster is not the complete 53-man roster. While this is true, more than half of teams will swap two or fewer roster spots in the lead up to week one, including moving players to IR. In 2024, organizations had an average of 49 players from their initial 53-man roster on active or reserve lists following the final game of the season. The draft and opening of free agency are indisputably more important days in team construction, but final cuts force teams to finally take shape.
I haven’t yet decided if I’ll revisit my predictions later in the preseason, but at the very least, I’ll be doing a recap after cutdowns of how my model ended up doing.