Recommendations are a transparent, deterministic nearest-neighbour search — no black-box ML. Here's the exact pipeline.
1
Every shaft becomes a fingerprint
Each shaft is reduced to 8 characteristics on a 1–5 scale: launch, spin, tip stiffness, mid stiffness, butt stiffness, balance (counterbalance), feel (smoothness) and stability. That ordered list of 8 numbers is the “DNA” — a point in 8-dimensional fitting space.
2
Distance = how far two fingerprints sit apart
A weighted Euclidean distance — squared per-axis differences, scaled by importance, summed, then square-rooted:
d = √( Σ wᵢ · (aᵢ − bᵢ)² )
| Weight | Axes | Why it's weighted that way |
| 1.0 | launch · spin · tip · stability | The outcomes plus the main dispersion/feel drivers a fitter chases first |
| 0.8 | balance | Counterbalancing changes how the whole shaft loads |
| 0.7 | feel | Matters, but subjective |
| 0.6 | mid · butt | Finer structural detail, secondary to the tip |
3
Two real-world penalties
A perfect profile in the wrong build isn't a real alternative, so distance grows for weight and flex gaps:
d += (|Δgrams| ÷ 10) × 0.55 ← weight gap
d += 0.4 if your flex isn't offered
4
Distance → a 0–100% score
match% = clamp( 100 · e^(−d ÷ 3), 35, 99 )
Exponential decay: an identical shaft scores ~100%, and the score falls off smoothly with diminishing returns rather than linearly. Capped at 99 (nothing's a literal clone), floored at 35.
5
Rank & surface the trade-offs
Score every other shaft, sort descending, take the top 6. The “Other makers only” toggle keeps true alternatives. Each card's delta pills (▲ softer tip, ▼ lower spin) are literally the axes with the largest gaps — so you see what you'd be trading, not just the number.
The honest caveat
The math is exact and reproducible — but the 1–5 inputs are curated relative estimates for most shafts. Only models with a ✓ Verified specs badge use published manufacturer numbers. And it matches shaft-to-shaft, not player-to-shaft: it knows nothing about your swing speed, tempo or attack angle. Use it to build a shortlist, then get fit by a pro.