Where these numbers come from

The test condition

Every measurement on this site was taken under one condition. It is not your tension, your swing or your racquet — which is exactly why the numbers are comparable to each other, and why none of them is a prediction about your setup.

Reference tension

51 lb (23.1 kg)

Swing speed

Fast

Strings in the database

788 across 44 brands

Data last fetched

2026-07-31

Source

TWU string performance database

Joined with

TWU string comparison tool

The single biggest caveat. A swing-speed and tension condition this specific means the rankings describe strings relative to each other under one protocol. Tension almost certainly matters more to how your setup plays than the gap between two strings ranked a few places apart — which is why the tension calculator exists and why the finder gives a nudge rather than a number.

How much of the database carries each measurement

A ranking can only include strings that carry every measurement it depends on. This is why a spin ranking is drawn from a smaller pool than a comfort ranking, and the result pages say so.

Every column, as the fetcher records it

Printed straight from the data file rather than written by hand, so it cannot fall out of date with what the site is actually serving. 9 of these are published on string pages; the rest are carried for provenance or filtering.

id

Stable key. TWU product code (pcode) where one exists, otherwise derived from the display name for strings only reporter2 lists.

brand_source

'comparestrings' (authoritative), 'inferred_from_name', or null

material

String construction family (Polyester, Nylon, Gut, Nylon/Zyex, ...)

gauge

Gauge label parsed from the display name. Higher number = thinner; an 'L' gauge is thinner again than its own base, so 17L != 17.

diameter_mm

TWU-measured diameter, falling back to the manufacturer figure. Both are range-checked to 0.9-1.55 mm and dropped if implausible, so this is null rather than wrong. Expect it to differ from any diameter in the product name: that one is the manufacturer's nominal figure, this is what TWU measured.

stiffness_lb_in ranked on

String bed stiffness in lb/in; higher = stiffer

tension_loss_pct ranked on

Percent of the reference tension lost under the test protocol; higher = loses tension faster. NOT pounds.

tension_loss_lb

The same loss expressed in pounds

energy_return_pct ranked on

Percent of energy returned; higher = livelier / more power

cof_static ranked on

String-on-string coefficient of friction; lower = better snapback

cof_ball ranked on

String-on-ball coefficient of friction; higher = more bite

spin_potential_ratio ranked on

TWU spin potential (cof_ball / cof_static); higher = more spin

currently_available

TWU flag for whether the string is still in production

has_data

False when TWU lists the string but publishes no measurements

stretch_pct_at_40lb charted

Elongation (%) when tensioned to 40 lbs (18 kg)

stretch_pct_at_51lb charted

Elongation (%) when tensioned to 51 lbs (23 kg)

stretch_pct_at_62lb charted

Elongation (%) when tensioned to 62 lbs (28 kg)

Every test condition, not just the headline one

Each string is measured at up to three reference tensions (40, 51 and 62 lb) crossed with three swing speeds (slow, medium, fast) — nine runs. The figures quoted across the site are one of those nine, because they have to be comparable string-to-string. All of them are published on each string page.

Strings with at least one condition

786 of 788

Strings with a full tension curve

502 — tested at every reference tension, so their tension response is measured rather than assumed

Strings with all three swing speeds

504

This is what makes the tension calculator work. Where a string has the full curve, a line is fitted through its measured points and solved directly. Where it does not, the calculator falls back to proportional scaling and says so on the result. The two answers differ by several pounds, and the difference is not noise: a real string bed carries roughly 60–65 lb/in of stiffness that does not come from tension at all, so scaling proportionally always overshoots at low tensions.

Measured versus inferred

This is the distinction the whole site rests on. Everything in the first list came off a testing machine. Everything in the second is arithmetic performed on top of it by StringSpec, and is labelled as an estimate wherever it appears.

Measured

• StiffnessString bed stiffness in lb/in. Higher values are stiffer and more control-oriented; lower values are softer and more comfortable.
• Tension lossPercentage of the reference tension lost under a fixed test protocol. Lower values hold tension longer.
• Energy returnPercentage of impact energy returned to the ball. Higher values play livelier and more powerful.
• Spin potentialRatio of string-on-ball friction to string-on-string friction. Higher values generate more spin.
• Ball biteString-on-ball coefficient of friction. Higher values grip the ball harder.
• SnapbackStatic string-on-string coefficient of friction. Lower values let the mains slide and snap back.
• Dwell timeHow long the ball stays on the string bed. Longer dwell is the physical basis of what players call hold or pocketing.
• Bed deflectionHow far the string bed moves at impact. More deflection means the bed is doing more of the work of stopping the ball.
• Peak forceThe highest force perpendicular to the string bed during impact. The spike the frame and the arm actually receive.
• Tension at impactWhat the bed is actually pulled to by the time the ball arrives, after it has settled. Always lower than the tension it was strung at.
• Loss: settlingTension lost simply from being strung and left alone, before a ball is hit.
• Loss: bedding inTension lost while the bed settles over its first impacts.
• Loss: impactTension lost to the impact itself. Frequently negative — a bed often gains tension across an impact rather than losing it.
• Loss: totalThe three losses above added together, in pounds.
• Elongation under load Measured at 40 lb, 51 lb, 62 lb — three separate measurements, plotted as a curve on every string page that carries them.

Inferred by StringSpec

• Percentiles and rankings Where a string sits in this database on a measurement. Exact arithmetic, but it describes this database — a percentile moves when strings are added or removed, and it is never a statement about strings that are not in here.
• Need rankings Each need is a signed weighting over those percentiles, published in full on the page that uses it. No playtest scores and no opinion — but the choice of weights is a judgement, and a different reasonable person would pick slightly different ones.
• Tension equivalence One of these two has not been tested at every reference tension, so this falls back to scaling bed stiffness in proportion — the way stringers work by feel. It is slightly sub-linear in reality, so treat it as a starting point and expect to adjust a pound or two.
• Playable life Scaled between 40 and 10 hours by the string's tension-loss percentile. The endpoints are the conventional figures stringers quote, not measurements. This estimates when the bed goes dead, not when the string snaps. If you break strings before you notice them going dull, you are a breaker — thickness and material matter far more to you than tension hold does.
• Hybrid beds Estimated, not measured. No hybrid combination has been tested as a bed — these figures are the two strings' own lab measurements weighted by how much of the bed each one actually is. Treat them as a guide to which way a pairing moves, never as lab data.
• "Like this, but…" alternatives Nearest neighbours in percentile space across the primary measurements, required to move at least 12 percentile points on the axis you asked for. Geometry, not judgement — but proximity in these six measurements is not the same as playing alike.
• Durability The database publishes no durability figure at all. The durability ranking is scored on proxies — thickness, string-on-string friction and material — and is the weakest thing on this site. It says so on the page.

What StringSpec deliberately does not have

• PricesNo price data of any kind. Nothing here knows what a string costs, so nothing here can tell you whether it is worth the money — a real gap, and stated rather than papered over.
• Playtest opinionsNo reviews, no star ratings, no feel scores. Everything is either a machine measurement or arithmetic on one.
• Conditions for every stringAbout a third of the database has been tested at only the reference condition, so those strings have no measured tension response and fall back to a proportional estimate. That is an upstream gap, not a choice.
• Hybrid measurementsNobody has tested hybrid beds. The builder estimates them and never claims otherwise.

Browse all 788 strings See the whole database plotted