Prop Lab
Log InSign Up

The method

How Prop Lab Organizes NFL Prop Research

Every NFL player prop we look at goes through the same type of analysis.

  • What’s the line and what’s it cost on each side? How has that line moved since it opened?
  • What is the player’s participation, usage, and is that growing or shrinking?
  • Is the player hurt? What about notable injuries on his team or the opponent?
  • What are analysts and insiders saying?
  • Who’s calling the plays, and is that someone new?

Then we look at the matchup, the likely game script, and the weather. Everything gets compiled and analyzed side-by-side, so you can see what's driving a prop.

By Prop Lab

See how it works

The problem

What Makes This Hard

A garbage time reception, an injury on the fifth play of the game, or a stop at the goal line can sometimes mean the difference in hitting or not. The box score can miss the true story too. A running back who gets nine carries with little backfield competition in a 30-point loss probably didn't lose his job, his team fell behind and abandoned the run. A top receiver who gets zero catches on high route participation in a 20-point win is unlucky, but it doesn't mean he won't continue to be a focal point of the offense going forward.

The season is relatively short too. Seventeen games is small compared with other professional leagues. Player roles can shift week-to-week depending on gameplan, injuries, and coaching decisions.

Then there's the line. By the time a prop is posted, the sportsbooks have already priced in the matchup, the injury report, the weather, and the player's role, and the line changes quickly when there's new information.

So most of the work is taking a second look at that number. We're not necessarily going to find new information that the sportsbooks aren't aware of, but we can analyze the prop with our own data and conviction to see how confident we are in the outcome.

The AI part

Why Asking a Chatbot Isn't the Same Thing

Ask a general AI chatbot like ChatGPT or Claude whether Patrick Mahomes will clear 230.5 passing yards against the Chargers this week and you'll get an answer. It'll be well written and it'll sound like it thought about it, but you'd do about as well flipping a coin.

The chatbot isn't doing anything wrong. It's answering the question it was handed with what it has, and what it has is thinner than it looks. Its training data has a cutoff, so anything it already knows about Mahomes comes from a snapshot that's sometimes really old. To cover that, it searches the web, which sounds like the fix until you think about what a search returns. The pages that rank for a question like that are the pages built to rank for it. Picks posts, aggregators, previews written on Tuesday. What comes back is a well-organized summary of what other people published about the game. Those sources can easily disagree, and none of it is the same as analyzing the game.

Two things happen next that are easy to miss. The first is that a chatbot won't tell you what it couldn't find, it will just guess. If snap counts and route participation weren't in the pages it read, it doesn't stop and say the most important inputs are missing. It answers with what it has, in the same tone it would use if it had everything.

The second is that how good the answer sounds has nothing to do with how good the evidence was. Writing fluently is the thing these chatbots are best at. It'll say a player should comfortably clear his number whether the case behind it is strong or almost empty, and both answers sound the same. The phrasing of your question shapes it too. Ask whether the over looks good and you'll tend to get a case for the over. Ask about the under and you'll often get a case for the under. Same player, same game, same number.

That's where the actual work is. We hand our AI Analyst the same structured packet for every prop: the line and its price, the player's recent games, the matchup, advanced analytics, injuries on both sides, the week's news, who's calling plays, and the weather. The packet and instructions are always in the same shape, with the same questions, and in the same order. None of that guarantees the AI is right 100% of the time. But it means every prop was analyzed to the same, consistent standard, and you can be more certain about how every prop was analyzed and what was considered.

Why I built Prop Lab

How we think about it

Our Core Principles

  1. Step 1

    Analyze the market

    The line already has a lot priced in. We start by comparing what different sportsbooks are offering to see where the consensus is, and track movement over time to see how they react to news and money coming in.

  2. Step 2

    Model the outcome

    Opportunity tells you more than the box score

    You might get lucky every once in a while hitting the over on a boom-or-bust WR who goes 3 catches for 104 yds, but that’s hard to repeat and predict. We look at how much of the offense actually goes to a player over time. This includes participation and usage metrics that tell us how often he is on the field and how much opportunity he had to contribute. We look at how that opportunity is trending over previous weeks, how it lines up against the next opponent, the expected game script, and any weather impacts. Opportunity is more predictable than production.

    Availability beyond the player

    Obviously it makes a difference if the player is injured, but we look across the matchup at the injury reports throughout the week to see if other injuries on the player’s team or opponent could impact the game. Knowing that an RB1 is a game-time decision is tough, but we consider that information as early as possible and build it into our models so we aren’t chasing the market for the RB2 when the line starts to rise or fall.

  3. Step 3

    State our confidence

    Ranking into confidence tiers

    We don’t stop with just an Over or Under. Posting a list of hundreds of props with Over/Under still requires you to go through them and figure out which ones you want to take. So we sort the props into four tiers: Strong Take, Take, Lean, and Pass. Strong Take and Take we feel good about given all the information we have. Lean are the ones we are on the fence about. Pass are props that seem too uncertain or volatile to us. These tiers help you sort through all the noise and get to what we think the best props are for the week.

    Strong Take
    Take
    We feel good about these
    Lean
    On the fence
    Pass
    Too uncertain or volatile

The evidence

What Actually Moves a Prop Line

A prop line opens with everything already known priced in. After that, it only moves when something changes, and most of those changes arrive on a fairly predictable schedule. Here's what moves a line during the week, and roughly when.

Explore NFL player props

Signal 01

What’s Priced In at the Open

A line opens with the known stuff already baked in.

By the time a player prop posts, often a few days after the game line, the book has already accounted for the player’s season so far, the matchup, and how each team likes to call plays. That’s why a great matchup on paper rarely moves a line on its own. Everyone’s known about it since the schedule came out. What moves a line after the open is new information, and the other cards are where it usually comes from.

Reading the Signals

How to Read the Output

Open any prop and you'll see the same few pieces. Here's what each one tells you, and how much weight to give it. New to a term? The prop betting glossary covers them.

AI Pick
The side our AI Analyst lands on, over or under. Every prop gets one, even a Pass, because there’s almost always a side we’d take if we had to.
AI Confidence
The tier: Strong Take, Take, Lean, or Pass. It tells you how strong the case is for that side. It isn’t the chance the player clears his number. We don’t publish a win percentage, because a number like 63.4 percent claims a precision the data can’t back up.
Odds
The typical price the sportsbooks have posted for each side, with O for Over and U for Under. Open it and you’ll see every book’s line and prices, so you can spot the best number. It’s a price, not a prediction.
Model Proj and Proj Diff
What our model projects for the stat, and how far that sits from the line. A big gap is a reason to dig in. On its own, it isn’t a bet.
AI Analysis
The reasoning in plain English, with the time it was written. If the line has moved since, you’ll see both numbers, like “Written at 64.5. The line is now 67.5.” That read was made against the old number, so weigh it that way.
Supporting Evidence
The opponent, the player’s usage, the injury report, and his recent games, laid out so you can check the AI’s reasoning yourself instead of taking it on faith.
Floor
Strong TakeTakeLeanPass

Resulting read

Strong Take

Start here. Everything we look at points the same way. It’s still not a guarantee.

What we don't claim

Football is inherently tough to predict and many factors influence the outcome of a game (which is why we all love to watch it). Player availability, roles and usage, weather, market lines, and other factors can change at a moment's notice without warning, and any data source can be incomplete or delayed. For those reasons, our methodology and research framework doesn’t guarantee an outcome or replace independent judgment.

League sources

Availability

NFL injury report

The league’s own weekly report, with its practice-participation and game-status designations.

Player tracking

NFL Next Gen Stats

The league’s public player-tracking program, and the stats it publishes from it.

Player stats

NFL.com player stats

The league’s official player stats for passing, rushing, receiving, and more.

Methodology FAQ

The line and its price across sportsbooks, how the line has moved, the player’s usage and recent games, injuries on both teams, the week’s news, who’s calling plays, the matchup, the expected game script, and the weather. Every prop gets the same inputs in the same shape, so you can compare one read with the next.

A general chatbot like ChatGPT or Claude answers from old training data plus whatever a web search turns up, which is mostly picks and preview posts. It won’t tell you what it couldn’t find, and it sounds just as sure either way. Prop Lab’s AI Analyst gets a structured packet of current data for every prop and answers the same questions every time, so you can see what each read was based on.

They’re confidence tiers for the side the AI Analyst picks. Strong Take and Take are the props we feel good about. Lean is on the fence. Pass means the prop looks too uncertain or volatile to act on, even though we still name a side. None of them guarantees a result.

Because a number like 63.4 percent claims more precision than the data can support. A tier tells you how strong the case for a side is, and that’s what you need to decide whether a prop deserves a closer look.

The board refreshes three times a day during the season. Reads change when the inputs do: a line moves, an injury designation comes out, a depth chart shifts, or the forecast firms up. Each write-up shows when it was written, and if the line has moved since, it shows both numbers.

Passing, rushing, and receiving props: passing yards, passing TDs, pass attempts, completions, interceptions, rushing yards, rush attempts, receiving yards, receptions, and the combined pass + rush and rush + rec yards markets. Which ones show up each week depends on what the sportsbooks post.

See the full board

Put the research to work every week

Create a free account to see this week’s full board, with the matchup, injuries, and line movement for every prop in one place.

Explore PropsSee how Prop Lab works