Set a player's game median (points, rebounds, or assists) and see the Over/Under probability at each half-point line around it.
| Offset |
Line |
P(Over) |
Over Odds |
P(Under) |
Under Odds |
|
-5
|
15.5 |
79.19% |
-381 |
20.81% |
+381 |
|
-4
|
16.5 |
74.23% |
-288 |
25.77% |
+288 |
|
-3
|
17.5 |
68.72% |
-220 |
31.28% |
+220 |
|
-2
|
18.5 |
62.75% |
-168 |
37.25% |
+168 |
|
-1
|
19.5 |
56.46% |
-130 |
43.54% |
+130 |
|
median
|
20.5 |
50.00% |
-100 |
50.00% |
-100 |
|
+1
|
21.5 |
43.54% |
+130 |
56.46% |
-130 |
|
+2
|
22.5 |
37.25% |
+168 |
62.75% |
-168 |
|
+3
|
23.5 |
31.28% |
+220 |
68.72% |
-220 |
|
+4
|
24.5 |
25.77% |
+288 |
74.23% |
-288 |
|
+5
|
25.5 |
20.81% |
+381 |
79.19% |
-381 |
Normal N(median, σ²): symmetric, best for points (a sum over many possessions is ≈ normal) and fine for higher-count rebounds. Auto-SD scales with the median: points ≈ max(4.0, 0.30×median), rebounds ≈ max(1.8, 0.38×median), assists ≈ max(1.6, 0.42×median).
Negative binomial: a right-skewed count model for rebounds / assists. The mean is solved so the median line sits at 50/50, and the spread comes from the variance/mean ratio φ (φ = 1 would be Poisson; defaults ≈ 1.30 rebounds, 1.35 assists). The skew mainly shifts lines far from the median — a bit more Over on high lines, less Under on low lines — which a symmetric normal misses. Its effect is largest for low counts and shrinks as the count grows (assists > rebounds > points ≈ none), so it's offered only for rebounds and assists.
Lines are half-integers, so no pushes. Odds shown are fair (no-vig) American odds derived directly from the probabilities.