How we score paddles

The four provenance labels

Every stat on this site carries one of these tags. They mean different things and should be trusted differently.

Swing weight index

Swing weight is how heavy a paddle feels through the swing — a moment of inertia about an axis 10 cm from the butt cap. It depends on where mass sits, not just how much there is, which is why two 8.0 oz paddles can feel completely different.

We model the paddle as two uniform rod segments — the handle and the face — with mass split between them, and integrate:

I = Σ (m / (b − a)) × [ (b − x₀)³ − (a − x₀)³ ] / 3

  a, b  segment start and end from the butt cap (cm)
  m     segment mass (g)
  x₀    swing axis, 10 cm from the butt cap

The head carries about 70% of total mass, adjusted up for thermoformed and foam-injected builds (which add perimeter mass) and down for edgeless designs.

We publish the result as a 0–100 index, not an absolute figure. The model ranks paddles against each other reliably, but converting it to a real kg·cm² number would imply a calibration against measured paddles that we don't have. Publishing a precise-looking number we can't stand behind would be worse than publishing an honest index.

One exception: the swing weight calculator reports the change from adding lead tape as an exact figure. That's pure physics — mass m at radius r adds m·r² — and needs no calibration, so it is exactly right.

Power

Weighted from core thickness (inverted — thinner cores rebound harder), swing weight, static weight and face material, then adjusted for construction:

• Core thickness, 34%. Mapped from 10.5 mm (maximum pop) to 19 mm (minimum), inverted.
• Swing weight, 26%. More mass moving through contact means more energy transferred.
• Static weight, 16%. Mapped 7.0–8.8 oz.
• Face material, 14%. Fibreglass rebounds hardest; raw carbon is comparatively muted.
• Adjustments. Thermoformed +8, foam-injected core +14, unibody +3, elongated +10, widebody −6.

Control

Deliberately not the inverse of power — if it were, half the comparison chart would be redundant. Control weights core thickness at 40%, swing weight moderation at 24% (peaking at a middling swing weight rather than the lowest), face softness at 16% and forgiveness at 20%, with a widebody bonus and an elongated penalty.

Spin

Dominated by face material, because that is what actually grips the ball: raw carbon 88, Toray carbon 82, Kevlar 78, composite 62, fibreglass 55, graphite 48. Thermoforming adds 4 and an elongated shape adds 5 for extra head speed.

An important caveat: face grit is a surface property and it wears. Any spin figure — ours or a lab's — describes a new paddle. Expect meaningful falloff after months of regular play.

Forgiveness

A stand-in for twist weight: resistance to the head twisting on an off-centre hit. Because twist resistance scales with the square of distance from the paddle's long axis, face width dominates at 62%, static weight contributes 28%, and shape adjusts the rest.

Why scores use fixed bounds

Every index is mapped onto 0–100 using fixed domain bounds rather than being ranked against the current catalog. This matters: with catalog-relative scoring, adding one new paddle silently changes every other paddle's number, which would quietly falsify every blog post and cached page that ever quoted one. A paddle's score here changes only when its specs change.

How we combine other reviewers' numbers

Independent reviewers publish their own measurements and ratings. We collect those, keep each one credited to the site it came from, and combine them only where combining them is honest. Two rules govern this, and they produce different answers today.

Measurements are combined. Swing weight and twist weight are physical quantities on the same scale no matter who measures them, so where more than one reviewer has measured a paddle we show the average, the number of sources behind it, and every individual reading. 202 paddles currently have a swing weight from two or more sources, 20 of them from three.

Ratings are combined by percentile, never raw. Five of our sources publish an editorial rating: Matt's Pickleball, STS Pickleball, PaddleReviewHub, BestPicklePad and RacketRise. Averaging those numbers as printed would be arithmetic on incompatible scales. Matt's rates in whole stars out of five, so his ratings can only ever land on five values. PaddleReviewHub publishes out of ten with almost everything above 8.5. STS publishes no overall at all, only four category scores out of ten. A raw average across them would move mostly with which sites happened to review a paddle — adding one source's score would lift nearly every paddle and adding another's would drop nearly every paddle, whatever the paddle is.

So each rating is first converted into where it sits among everything that same reviewer has rated. We hold every rating each source published — 475 from Matt's, 102 from STS, 70 from PaddleReviewHub, 64 from BestPicklePad, 7 from RacketRise — sort each source's own list, and find the rating's position in it, splitting ties down the middle so the dozens of paddles sharing a five-star bin all sit at the centre of that block rather than at its top or bottom. A 9.2 from a source whose median is 8.8 becomes “above about half the paddles they rated”. Those positions are on one shared scale by construction, so we average them, and that average out of 100 is the aggregated reviewer score.

Four rules keep it honest. Two sources minimum — with one reviewer there is nothing to aggregate, so the score simply does not appear rather than relabelling that reviewer's rating as ours. The source count is always on screen, because a score from two reviewers and a score from four are different claims and must not look identical. Disagreement is marked, not smoothed — when the highest and lowest positions are more than 35 points apart we say the sources disagree, which is a fixed, stated threshold rather than one fitted to make a number look good. Crowd ratings stay out. MyPickleList's owner ratings are shown on their own card but never enter the aggregate; it combines reviewers, not audiences.

What that score is not: it is not objective, not unbiased and not a measurement, and we will not describe it in those words. It is exactly what it says — the average of where each named reviewer ranked this paddle among the paddles they rated. With two or three sources any one of them still moves it noticeably, which is why every reviewer's own figure and their own position is printed underneath it.

Disagreement is shown, not smoothed. Independent rigs and protocols differ. Across the paddles both our sources have measured, their swing weight figures typically differ by about 1.5 and occasionally by as much as 11; on twist weight one source reads consistently around 0.18 higher than the other, which is a difference in method rather than noise. When two sources are further apart on a paddle than they are on roughly 90% of the paddles they have both measured, we mark it sources disagree rather than quietly averaging it away.

A figure drawn from two sources is not the same claim as a figure drawn from twelve, so the source count sits next to every combined number. Fewer sources means a weaker signal, and the interface should never let those look identical.

Measurement chips are context, not judgement. Next to each reviewer's reading we mark whether it falls in the bottom quarter (low, red), the middle half (average, amber) or the top quarter (high, green) of every reading of that measurement we hold across the site. The cut points are the 25th and 75th percentiles of the pooled readings and are recalculated as the catalog changes, so a chip describes today's market rather than a fixed standard; a measurement with too few readings site-wide to split honestly gets no chip at all. The colour marks where a reading sits on that scale, never whether sitting there is good: a high swing weight is harder to swing, a high twist weight is more stable, and which one you want is your call, not ours.

What we take from other sites, and what we don't

• Numbers and factual specs only. We do not reproduce review prose, verdicts, pros-and-cons lists, headlines or photography. Our contribution is the cross-source comparison, which is not something any single source publishes.
• Public pages only, at a polite rate. Nothing behind a login or paywall, nothing a normal visitor couldn't load, and never fast enough to burden anyone's servers. We honour each source's robots.txt.
• Always attributed. Every third-party figure names its source and links to the page it came from, with the date we collected it. Naming a reviewer is how you can check our work against theirs.

How the recommender works

Find My Paddle turns your answers into a target profile across the five indices, then scores every paddle by weighted distance from it. Budget and handle length are applied as hard constraints rather than preferences — a paddle that can't fit two hands is not a compromise for a two-handed backhand, it's wrong. Each recommendation shows the specific reasons it scored well.

What we don't do

• We don't invent ratings. Paddles with no owner submissions display no rating, and we never emit aggregateRating structured data without real ratings behind it.
• We don't take payment for placement. No brand can buy a position, and none has been offered one.
• We don't quote live prices we can't verify. We show MSRP and link out, rather than displaying a scraped price that may be hours or months stale.
• We don't pretend estimates are measurements. Ever. That's the entire point of the provenance labels.

Corrections

If a spec here is wrong, tell us and we'll fix it and note the change. Email hello@dinkdex.com.