How RadioRanked Scores Radios and Antennas

Every scored product here carries a score out of 100. This page is how that number is produced: the inputs, the weight of each dimension per category, the arithmetic behind each one, the missing-data rules, and a worked example. The weight tables, the curve and the example are rendered by the code that scores the catalogue, so this page cannot disagree with the site.

What the Score Is, and What It Is Not

The score is a composite of published specifications, the current Amazon price, and the Amazon owner rating with its review count. It is arithmetic on listed data, not a judgement I make product by product, and no brand can buy a better one. The daily price refresh recomputes all four scored catalogues, so a score moves when its data moves and at no other time.

What the score is not:

• I do not bench-test radios. No RadioRanked measurement of power, runtime, sensitivity or audio is in it. Every specification comes from a published data sheet, manual or product listing; none is a RadioRanked bench measurement.
• It cannot see build quality, audio quality or long-term reliability. Those take years of use to judge.
• The Amazon owner rating is my proxy for that gap, and an imperfect one. Owners rate a radio in its first month, and often rate the seller along with it.
• A published IP rating is the only durability input. A manufacturer that publishes none leaves the radio scored as though it has none.
• The reviews on this site are editorial. They are my opinion, and they do not feed the score.

Where the Inputs Come From

• Manufacturer data sheets and manuals. Power, frequency coverage, modes, memory channels, weight, battery capacity, IP rating, and for antennas length and connector.
• Amazon listings. Price, owner rating and review count, refreshed daily by the price cron. Listing copy is also parsed for specifications a data sheet omits, with marketing text trusted ahead of the seller's attribute dictionary, which has invented features on radios that lack them.
• FCC equipment authorisation filings. Shown on product pages, never scored. A grant records that a device was authorised for a service, not how well it performs.
• A hand-curated receiver class for HF. A tier from 1 to 5, anchored to published receiver test rankings. It is the one input I assign by hand.
• Length and connector for antennas. The antenna engine runs on measured length and the bands covered, not the gain printed on the packaging.

Weights, Category by Category

Each category combines its dimensions as a weighted average. The third column is the set used when a product has no current price: value is dropped and its weight redistributed. Every set sums to 100%.

Handheld radios

The reference category. A handheld is carried and is often a first radio, so portability and beginner friendliness are both scored.

Mobile radios

Portability is dropped, because a rig bolted into a truck has no meaningful weight and battery axis. Transmit power carries more of the feature dimension instead, against a 50 W ceiling, since the handheld curve saturates on every mobile.

HF transceivers

Receiver quality replaces beginner friendliness and becomes the heaviest dimension, because on HF the receiver is the headline buying criterion. Portability stays, since field and QRP operation is a real HF use case.

Antennas

The feature slot carries RF performance, meaning electrical length rather than advertised gain. Scores are normalised within a family, so an 85 on a handheld whip means excellent for a handheld whip, not better than a mobile antenna.

GMRS radios

GMRS radios are not scored, by design. Part 95E caps power and fixes the channel plan, so the specifications a score would weigh are near identical across the category. The buying guides rank by published transmit power, then by Amazon rating, and each says so rather than claiming a scoring algorithm ran.

How Each Dimension Is Computed

Every dimension lands on a 0 to 100 scale before the weights apply. The constants below are the ones in the engines.

Value. Capability per dollar through a logistic sigmoid, mapped onto 5 to 95. Capability is not the feature score. On a handheld it is a separate RF capability index: the digital mode points, the memory and power curves and the colour display from the table below, without the IP, CHIRP and multi-mode bonus points, plus up to 15 points of Amazon quality signal (rating and review count together). A mobile uses the same index with an 8-point floor, its power curve worth 15 points against 50 W, and the same 15-point signal. On HF it is the feature score plus half the receiver score, and on an antenna the RF performance score. Each category is centred on its own price floor.

Features. Additive points, capped at 100, with the handheld set below as the reference. Mobiles use the same flags with power worth up to 15 points against a 50 W ceiling, and HF scores 20 for HF coverage, 15 for adding VHF or UHF, 5 for 6 m, up to 20 for power against a 100 W ceiling, 2 per transmit mode up to 10, 12 for an antenna tuner, 6 each for D-STAR, C4FM and a colour display, 5 for an internal battery, and up to 6 for memory channels topping out at 1,000.

Portability. A handheld blends 45% weight, 45% battery and 10% of a battery-to-weight bonus. The weight curve is 100 times one minus the ratio of grams to 500 raised to the power 1.3. Battery follows a square-root curve worth up to 80 points, topping out at 4,000 mAh, and the bonus is 3 points per mAh per gram, capped at 10. HF rigs take 45 points at or under 800 g with a logarithmic taper beyond, 30 for an internal battery and 15 for a transmit power of 25 W or less, floored at 5. Antennas score 100 times one minus the ratio of length to a reference raised to the power 0.85, the reference being 900 mm for a handheld whip and 1,800 mm with a ground plane, plus 8 for a flexible whip and 4 more if it folds.

Beginner friendliness. Starts at 30. CHIRP compatibility adds 20, since free open-source programming removes the biggest obstacle a new operator hits. A colour display adds 8, Bluetooth 5, an IP rating 4. Price contributes up to 20 points on a logarithmic curve anchored so $20 earns the full 20 and $400 earns nothing. Radios with no DMR, D-STAR or C4FM take a simplicity bonus of 10. Up to 10 more come from Amazon social proof, mapping 3.5 to 5.0 stars against a review-count confidence saturating at 3,000. Clamped to 0 to 100. Mobiles use the same shape without the IP-rating bonus and with the price curve recentred on $100 to $600, and HF drops the dimension.

Owner reputation. The rating maps linearly from 3.0 to 5.0 stars onto 0 to 1, then rises to the power 0.7 so the top of the range is rewarded. Review count becomes a confidence factor: the base-10 logarithm of the count over the logarithm of 5,000, capped at 1. The two combine as the rating curve times 0.6 plus 0.4 times that confidence, and the score is 5 plus 90 times the result. No rating, or no reviews, scores a neutral 50 rather than a zero. On mobile and HF, where Amazon often gives an average without a total, a known rating with an unknown count is held at a fixed confidence of 0.5.

Receiver quality, HF only. The curated tier sets the base: tier 5, contest grade, 85; tier 4, high-performance direct sampling, 70; tier 3, solid mid-range, 55; tier 2, 40; tier 1, 25. An unassigned rig gets 15. A waterfall display adds 8, and general-coverage receive, meaning the receiver starts at or below 0.6 MHz and runs to at least 29.7 MHz, adds 7. Capped at 100.

Antenna RF performance. Gain claims on cheap antennas are frequently fiction, so the engine never reads the claim. It computes electrical length: physical length divided by the wavelength at each band covered, using 146 MHz for 2 m, 223 MHz for 1.25 m, 435 MHz for 70 cm and 465 MHz for GMRS. That length in wavelengths becomes a plausible gain in dBi through a piecewise curve, with a separate curve for antennas working against a real ground plane, because a quarter-wave monopole on a vehicle roof behaves close to theory and the same element in your hand does not. The gain is normalised inside a window of minus 2 to 7 dBi with a ground plane and minus 6 to 3.5 dBi without, averaged across the bands covered and multiplied by 92, with 6 points for covering both 2 m and 70 cm and 2 more for 1.25 m. The advertised gain is still shown, used only to flag claims the antenna's own dimensions cannot support, allowing 2 dB of tolerance. That flag is kept out of the score: an antenna radiates however it radiates regardless of what the box says.

What Happens When Data Is Missing

A missing input must not read as a bad one. Where a sensible default exists the engine uses it and says so. Where none does, it declines to score.

• No price. Value is dropped and the remaining weights rescaled, the third column of every table above. An unpriced product is not penalised, nor credited with being a bargain.
• No weight. A handheld with none published is scored as 250 g, mid-range for the category. An HF transceiver with none is treated as a desk rig and given low portability credit. Mobiles do not score portability at all.
• No rating. Owner reputation is the neutral 50 in every category. A rating with no review count is also 50 on handhelds and antennas, while mobile and HF keep the rating and assume a 0.5 review confidence, because their counts come from Amazon's star histogram rather than a total.
• No antenna length. RF performance falls back to a neutral 50 rather than a guess that would rank an unmeasured antenna against a measured one.
• Antenna families the model does not describe. Directional antennas and mounts get no score. A beam's gain comes from element count and boom length together, not the resonant length of one radiator, so the model reads a Yagi's boom as a radiator and predicts a fraction of the gain it makes. No score beats a wrong one.
• Two aviation transceivers. The handheld catalogue holds two transceivers that transmit AM on the aviation band and cannot transmit on the amateur bands at all, so they are excluded from rankings and comparisons. The engine looks for DMR, APRS, GPS and CHIRP, and was marking avionics equipment poor for lacking ham features it never advertised. Their product pages stay, because a pilot who lands on one is reading something accurate.

The Overall Curve

The weighted average of the dimensions, which I call the base, is not published directly. It goes through one logistic curve shared by all four catalogues, so every score sits on the same scale. The overall score is 5 plus 93 divided by one plus e raised to the power of minus 0.07 times the base minus 46, rounded to a whole number.

The curve exists because raw weighted averages bunch up. Most radios in a category land close together, and a 10-point difference in specifications would otherwise show as a 2-point difference in the published score. Stretching the upper range makes real differences visible, at the cost of compressing the extremes, where almost nothing sits.

A Worked Example, Computed Live

Here is the whole calculation for one radio, run when this page was last built. The inputs are the stored values for Baofeng BF-5RH PRO, and the dimension scores come from the same function the catalogue calls, so nothing here is a retelling.

Those inputs produce the 5 dimension scores in the selected set, each multiplied by its weight from the priced handheld set.

These figures are computed from today's stored data by the same function the rest of the site uses, so they can differ from a cached product page for a few hours after a price change. The number on the product page is the one written at the last nightly recalculation.

How Prices Move the Rankings

Prices refresh daily and the four scored catalogues are recalculated with them. A price change moves the value dimension in every priced category, and on handhelds and mobiles it also nudges the beginner score, so a discount shifts a radio by a point or two rather than rewriting a ranking. The larger effect is a product going out of stock: it loses its price, value is dropped, and the remaining weights are rescaled. That is usually why an unpriced radio sits somewhere surprising in a list, its score answering a slightly different question from the priced products around it.

Scoring Changelog

• 2026-03-14. Scoring engine v2. The owner reputation dimension and the shared logistic overall curve replaced the original linear weighting.
• 2026-08-09. The antenna engine began scoring electrical length instead of advertised gain, and declining to score families the model does not describe.
• 2026-08-11. Two aviation transceivers in the handheld catalogue had their scores cleared and were removed from rankings and comparisons.
• 2026-09-15. This page published. The weights and the overall curve moved into one shared module the page renders directly, and the About copy was corrected: it had described the score as pure specification data, which understated the owner reputation weighting.