Normal Distribution for Player Props: Calculate Over/Under Probability | DMP Learn

The Normal Distribution for Player Props

What You'll Learn

The normal distribution (bell curve) is the most important statistical tool in prop betting. It tells you the probability that a player goes over or under any line — not as a guess, but as a calculation. By the end of this lesson, you'll know when the normal distribution applies, how to calculate over/under probabilities yourself, and why this matters for finding +EV props.

---

Why This Matters for Prop Betting

Every sportsbook sets player prop lines based on projections. But projections alone don't tell you whether a line is worth betting. You need to know the probability of clearing that line — and that requires understanding how a stat is distributed.

For many player stats, performance follows a bell curve: most games cluster around the average, with fewer games at the extremes. When a stat fits this pattern, you can use the normal distribution to calculate exact probabilities for any line.

This is how sharp bettors move beyond "I think he'll go over" to "there's a 62% chance he clears this line, and the book is pricing it at 52%."

---

When Does the Normal Distribution Apply?

The normal distribution works best for continuous stats with moderate variance — stats that can take many possible values and tend to cluster predictably around a player's average.

Excellent fit (use confidently):

• NFL passing yards — Quarterback passing totals are among the most normally distributed stats in sports. Large sample sizes per game (30+ attempts), continuous outcomes, and relatively stable game-to-game performance make this a textbook application.
• NBA assists — Playmakers tend to produce assists in a tight, predictable range. A point guard averaging 8.5 assists might range from 5-12 on a typical night, following a clean bell curve.
• NBA rebounds — Similar to assists. Consistent minutes and role produce a tight distribution around the mean.

Decent fit (use with caution):

• NBA points — Points have higher variance than assists or rebounds because scoring depends on shot selection, foul trouble, and game flow. The distribution is roughly normal but with fatter tails (more blowup and bust games).
• MLB strikeouts (high-K pitchers) — For pitchers averaging 7+ strikeouts, the normal distribution works reasonably well. Below that threshold, the Poisson distribution is often a better fit. (We cover this in the distribution selection lesson.)

Poor fit (don't use):

• Home runs — Too rare and binary. A player either hits 0, 1, or occasionally 2. This isn't a bell curve.
• Anytime touchdowns — Same problem. Discrete, rare events with heavy zero-inflation.
• MLB hits — Low counts (0-4 typical range) with a skewed distribution.

---

The Core Concepts

Mean (Average)

The player's average performance over a relevant sample. For props, you want a recent, context-adjusted mean — not just the season average. Factors that shift the mean include matchup difficulty, home/away splits, minutes or pitch count expectations, and recent form.

Example: A quarterback averages 265 passing yards on the season, but against top-10 pass defenses he averages 238. If he's facing a top-10 defense this week, 238 is your adjusted mean.

Standard Deviation (SD)

Standard deviation measures how spread out a player's performances are around their mean. A small SD means the player is consistent; a large SD means volatile.

Why this matters: Two players can have the same average but very different betting profiles.

• Player A: Averages 24.5 points, SD = 3.2 → Most games fall between 21.3 and 27.7. Very predictable.
• Player B: Averages 24.5 points, SD = 8.1 → Games range from 16.4 to 32.6 regularly. Volatile.

If the line is 24.5, both players have the same "average" relationship to the line — but the probability distributions are completely different. Player A is a much more reliable over/under candidate. Player B's wider spread means the line matters less because outcomes swing wildly either way.

Z-Score

The Z-score tells you how many standard deviations a line is from the player's mean. It standardizes the question so you can look up exact probabilities.

Formula:

Z = (Line - Mean) / Standard Deviation

A Z-score of 0 means the line equals the mean (roughly 50/50). A positive Z-score means the line is above the mean (under is more likely). A negative Z-score means the line is below the mean (over is more likely).

Probability Lookup

Once you have a Z-score, you convert it to a probability using the standard normal distribution table (or a spreadsheet function like NORM.DIST). This gives you the probability of going under the line. Subtract from 1 to get the over probability.

---

Step-by-Step: Calculating Over/Under Probability

Let's walk through a real example.

Scenario: Josh Allen's passing yards line is set at 275.5. Based on your research, his adjusted mean for this matchup is 285 yards with a standard deviation of 42 yards.

Step 1: Calculate the adjusted mean.

Season average: 272 yards. But this week he's at home (add ~8 yards based on home splits) against a bottom-5 pass defense (add ~5 yards from matchup adjustment). Adjusted mean: 285.

Step 2: Estimate standard deviation.

Look at his game log. Calculate the SD of his passing yards over the last 10-15 games in similar contexts. In this example: SD = 42.

Step 3: Calculate the Z-score.

Z = (275.5 - 285) / 42 = -9.5 / 42 = -0.226

Step 4: Convert Z-score to probability.

Using a Z-table or NORM.DIST(-0.226):

• P(Under 275.5) = 41.1%
• P(Over 275.5) = 58.9%

Step 5: Compare to the sportsbook's implied probability.

If the over is priced at -115 (implied probability ~53.5%), and your model says 58.9%, you've found a potential edge of ~5.4 percentage points.

---

How to Calculate SD from a Game Log

You can calculate standard deviation in any spreadsheet:

1. Collect the player's last 12-16 game results for the stat
2. Remove outlier games caused by injury exits (partial games distort the data)
3. Use =STDEV() in Excel/Google Sheets on the remaining values

Tip: Don't use the full season if conditions have changed significantly (new team, injury return, role change). A rolling 12-16 game window is usually better than a full season average for capturing current form.

---

Common Mistakes

Using season-long averages without adjustment. A player's true mean for any given game depends on context. Season averages wash out matchup effects, venue effects, and role changes.

Ignoring standard deviation entirely. Most casual bettors only look at averages. But as the Player A vs Player B example shows, SD determines whether a line is tight or loose — and that drives the actual probability.

Applying normal distribution to rare events. Home runs, touchdowns, and other low-count stats don't follow a bell curve. Using the normal distribution here gives misleading probabilities. Use Poisson or Negative Binomial instead (covered in the next lessons).

Using too small a sample. Five games isn't enough to estimate a reliable mean or SD. Aim for 12-16 games minimum, and weight recent games more heavily if the player's role has changed.

---

How DMP Handles This

DMP calculates fair probability for every prop using consensus devigged odds from sharp sportsbooks. The platform's projections account for matchup context, and the EV calculations reflect the true probability implied by the sharpest lines in the market.

When you see a prop's fair probability on DMP, the underlying math draws on these same distribution concepts — converting projections and variance into probabilities, then comparing them to what the sportsbook is offering.

---

Key Takeaways

• The normal distribution converts a player's mean and standard deviation into exact over/under probabilities for any line.
• It works best for continuous, moderate-variance stats: NFL passing yards, NBA assists, NBA rebounds.
• Two players with the same average can have completely different probability profiles depending on their standard deviation.
• The Z-score formula (Line - Mean) / SD is the bridge between raw stats and probability.
• Always adjust the mean for matchup context rather than using raw season averages.
• Don't apply the normal distribution to rare, discrete events like home runs or touchdowns.

---

Next lesson: The Poisson Distribution for Count Props →