How the Lyzos Score works (and why most picks score below 64)

A walkthrough of the 6-factor model, what each factor measures, and why the scoring scale is intentionally stingy.

If you've spent any time in prop markets, you've noticed something: every picks tool says everything is a "strong play." Every pundit on social media has 4–6 "locks of the day." The math doesn't add up. If everything's a lock, nothing is.

The Lyzos Score is built on the opposite premise. Most picks aren't actually edges — they're coin flips with thin margins, which is exactly how sportsbooks design their lines. A scoring system that recognizes this and tells you to pass is more valuable than one that tells you to play everything.

The factors, in order

Six factors go into every score. Each contributes a weighted sub-score; the weights vary by sport and stat type because what predicts NBA points is different from what predicts MLB strikeouts.

1. Recent form (L5 / L10)

Recent games matter more than the season average. A player averaging 22 PPG who has put up 28, 31, 25, 30, and 27 in the last five is not the same as one averaging 22 PPG who has put up 18, 20, 22, 19, and 21. Recency carries information about minutes, role changes, and form that the season number washes out.

2. Hit rate vs. this exact line

The most underrated factor. Knowing a player averages 24.5 points isn't useful if the line is 22.5 — what matters is how often they specifically clear that number. A player who has gone over 22.5 in 7 of his last 10 games has a meaningfully different signal than one who has gone over in 5 of 10, even if their averages are identical.

3. Head-to-head matchup

Some teams are nightmare matchups for specific player archetypes. A point guard who feasts on switching defenses isn't the same play against a drop-coverage team. We pull H2H splits going back two seasons and weight them based on the opponent's defensive identity.

4. Home / away splits

Home/away differences are real but inconsistent — some players have huge splits, others don't. Vegas mostly accounts for this in line-setting, but not always perfectly, especially for younger players whose splits are still developing.

5. Rest days

Back-to-backs and short rest reduce expected output for most players, especially older ones and high-minute starters. The fatigue factor scales with player age and recent minutes load.

6. Injury status

This is the one most public picks tools get wrong because their data is stale. We pull live injury reports on every analysis. A "questionable" tag from yesterday is not the same signal as a "probable" tag from this morning. Teammate injuries also matter — a star's absence changes usage rates for everyone else on the floor.

Why most picks score below 64

Sportsbook lines are not random. They're set by people whose job is to make every prop close to a coin flip after juice. Most lines do their job — they sit close to the player's true expected value, which means most picks should not have a meaningful edge.

The Lyzos Score reflects this. The distribution looks something like:

• 0–49 (Skip): About 50% of picks. Data does not support the play.
• 50–63 (Lean): About 30%. Coin flip with a slight tilt.
• 64–77 (Good Lean): About 15%. Real edge worth a measured play.
• 78–100 (Strong Play): About 5%. All factors aligned. Rare.

If you analyze 100 props in a night, you'll see maybe 5 strong plays and 15 good leans. The other 80 are passes. That's not the model failing — that's the model working.

The hardest skill in prop betting

It's saying no. Every play you don't make is a play you can't lose. If the score says 47, listen to it. The next strong play is coming — there's another night, another slate, another spot. Discipline outperforms hot streaks every time.

That's the whole pitch. We built the score to be hard to please because the markets are hard to beat. If a tool tells you everything is a winner, it's selling vibes, not analysis.

Want to see the methodology in more depth? Full breakdown here.