I wanted to take a more structured look at a common question: what matters more for range—sling length or projectile weight? Instead of focusing on absolute performance (which is heavily influenced by skill), I compared the relative importance of these two variables using existing throw data.

The dataset comes from the forum-compiled records here:
https://www.slinging.org/ranges.html

This page aggregates user-submitted throws, including range, sling length, and projectile weight. It’s not a controlled dataset—entries vary in skill level, style, and measurement conditions—but it provides a useful real-world snapshot of performance across many slingers.

I used a random forest model with permutation importance and bootstrapping to compare how much predictive signal each variable contributes. The model itself isn’t especially strong (OOB R² ≈ 0.186), which is expected given the variability in human performance, but it’s still useful for comparing variables against each other.

In the first figure, each point represents a bootstrap resample comparing the importance of projectile weight (x-axis) and sling length (y-axis). The dashed line shows where they would be equal. Most points fall above that line, meaning sling length is more important in most cases. Quantitatively, projectile weight was more important only about 20.5% of the time, so sling length dominates roughly 80% of the time in this comparison.

The second figure shows the partial dependence of range on projectile weight. This relationship is irregular and nonlinear, with a sharp spike at lower weights and then a broad plateau beyond ~150 g. There’s no consistent trend, which suggests that weight has a weak and unstable relationship with range in this dataset. Most of the variation here likely reflects noise or sparse data rather than a strong underlying effect.

The third figure shows the partial dependence of range on sling length. This is much more structured: range increases with sling length up to around ~120–130 cm, followed by a noticeable jump and then a plateau around ~215–220 m. This suggests a threshold effect, where increasing sling length improves performance up to a point, after which returns diminish.

Taken together, the results suggest that sling length carries a stronger and more consistent signal than projectile weight when predicting range. That doesn’t mean weight doesn’t matter—it clearly does—but within this dataset, its effect is less stable and less predictive compared to length.

Because this is forum data, it reflects real-world variability rather than controlled conditions. So the takeaway isn’t a strict rule, but a tendency: across many slingers and setups, sling length shows up as the more reliable driver of range.

Curious how this lines up with people’s experience, especially across different styles and projectile types.

(See attached figures: 1 = bootstrap importance comparison, 2 = weight PDP, 3 = length PDP)