The tube arrives with twelve shafts of the same model, all with the same spine number printed on the label. At that point archers almost always ask me two things: whether that spine suits their bow, and whether the twelve arrows that come out of it will be the same as one another. They are practical questions, and research has taken them more seriously than most people think, although nearly always with a machine in place of the archer.
Here I set out what has been measured, from choosing the spine to the stiff side of the shaft and the nock, and each time I say how solid it is. Why the arrow bends as it leaves the bow is covered in the article on the archer's paradox, and adjusting the bow in the one on bow tuning; this article starts one step earlier, with the box of shafts.
Matching spine to draw weight and length
Spine measures how stiff a shaft is. The shaft rests on two supports 711 millimetres (28 inches) apart, a weight of 880 grams hangs from the middle, and you read how far it bends in thousandths of an inch: a 380 deflects 9.65 millimetres, or 380 thousandths, and the lower the number, the stiffer the shaft. The rule everyone knows says that spine has to be chosen according to draw weight and arrow length, because a more powerful bow needs a stiffer shaft and a longer arrow behaves like a softer one.
The registry holds this rule as reported. The mathematical models that have studied it since 1990 all arrive at it, each by its own route, and one of them also calculates the window: on a competition recurve, moving to the next spine size takes about 11 newtons of draw weight to stay tuned, while a bow's draw weight can only be adjusted over a range of about 10% of its nominal value, so for a given draw length and arrow length only one or two spines can be tuned well. The part that concerns archers is thinner. In 1990 the equipment of eighteen of the twenty-two best Polish archers fell within the band the model indicated, and nobody has yet had archers shoot well-matched shafts alongside deliberately mismatched ones to see how much the group changes.
Manufacturers' charts
In practice spine is chosen from the manufacturer's chart, by crossing your actual draw weight, the weight you hold on your fingers at your own draw length, with the length of the arrow. That the best archers do this is reported from a single observation, but a clear one: in high-speed footage of sixteen recurve archers from national teams, filmed at the London Olympic test event, arrow stiffness was in every case what the charts predicted for their draw weight and arrow length, and fifteen of the sixteen shot the same shaft model. Draw weights and lengths were as declared by the archers, and the shafts had very probably been chosen from those same charts, so the finding says that the best archers follow the charts, not that the charts point to the best shaft.
On that second point there is only a numerical model from 2025, not yet checked against real shots. Overall it reproduces the commercial charts, but at the extremes it finds them too stiff, by up to four sizes, for long arrows and low draw weights. It also notes that the charts move in steps of four to six pounds and one inch, while in many rows half an inch of arrow is already enough to change the recommended shaft. For anyone shooting a low draw weight or a long arrow, then, the chart remains the starting point, and confirmation has to come from tuning.
The number on the box and the arrow in flight
The spine value comes from a static test, whereas in the shot the string pushes the arrow from behind, and the shaft compresses and flexes as it accelerates. The claim that a measurement at rest is not enough to predict how the arrow behaves when shot is transferred: three models over twenty-five years support it, but no test has yet shot the same shafts to compare the two rankings. In one of those models an X10 550, 4% stiffer than a ProTour 570, needs less draw weight to leave the bow cleanly, 175.5 newtons against 180, because the two shafts distribute stiffness and mass differently. This is why the author of the model warns that shafts of different types with the same spine number should not be treated as equivalent, and why a change of model is best started again from that model's own chart.
Some archers also look at natural frequency, the rate at which the shaft vibrates. That stiffer shafts vibrate faster is reported: three arrows from the same series, from the softest to the stiffest, vibrate at 74, 77 and 80 hertz. Between one spine and the next, though, there are only three hertz, and nobody has shown that measuring frequency helps in choosing a shaft.
What a competition arrow weighs
Of the figures people quote from memory about the recurve arrow, speed and length hold up: arrows leave the bow at roughly 57 to 65 metres per second, and the 39 elite archers from four countries followed in one study reported an average arrow length of 64.4 centimetres. Weight, on the other hand, is contested, because it depends on the shaft. The figure of around fifteen grams belongs to the A/C/E, and 14.3 grams is the weight of the arrow Japanese laboratories use as their reference; the two X10s weighed on the bench in a Japanese study came in between 19.7 and 21.2 grams, while those in the 2025 French test, which I come to below, cut to about 81 centimetres for a 52 lb bow, reached 27.25.
The material matters too. That at a similar spine an aluminium shaft weighs more than a carbon one is reported from a single measurement, on a rig that estimates mass by making the shafts vibrate: between one and a half and 1.7 times as much, with the 2219 aluminium at 22.7 grams against 13.5 for the carbon 300, although this is a mass estimated by the rig and not the weight of a finished shaft. As for length, the only study that set it against score, in the same 39 elite archers, found no link. Length was self-reported, though, it varied little and it depends on each archer's draw length, so the result does not say that shortening or lengthening your own arrow makes no difference.
How alike the arrows in a dozen are
This brings us to the second question. The manufacturer states that within a dozen the shafts differ by no more than half a grain, about 0.032 grams. A French group weighed twenty-six dozen X10 shafts belonging to their federation, in seven different spines, one by one on a scale reading to the milligram, and not a single dozen stayed within that tolerance; mixing dozens of the same spine, the range ran from 1.96 to 3.69 grains, and up to 11.64 in the 380s. According to the ballistic calculation in the same paper, at 70 m half a grain more lowers the impact by almost a centimetre, but the arrows of each dozen would all stay in the 10, and points would only be lost by mixing different dozens. It is a conference abstract with a simple model that leaves out shaft flex, but the practical message is clear: the stated weight needs checking, and dozens should not be mixed.
Six years later the same group carried out the most detailed test that exists on this question. Fifty-nine X10 shafts in spine 325 were measured one by one on an automated bench, assembled with point, vanes and nock by an archer of national standard, and shot by a machine at 70 m indoors, five times each and in random order. Almost all the weight variation came from the shaft and the point, while nocks and vanes weighed practically the same; between the lightest and heaviest assembled arrow there were 0.282 grams, and the authors calculate that pairing the heaviest points with the lightest shafts would bring the spread down to 0.027.
That weight and nominal spine are not enough to make a set consistent is therefore reported, from two independent bodies of evidence, neither of them with archers shooting. Out of two hundred and fifty-five combinations of characteristics, the one that best explained how close the arrows landed to one another was mass, mean spine and the frequency of the stiffness variation around the shaft. For the position of each arrow on the target the authors built two separate models: mass and mean spine accounted for just under half of the height differences between arrows, while for the sideways ones mean spine and the amplitude of the stiffness variation explained less than a fifth.
Those same fifty-nine arrows, however, were not matched by weight, and they still kept their mean positions within 8.8 centimetres vertically, inside a 10 ring that measures 12.2. Matching matters, but with arrows of this quality it is not what separates a 10 from a 9.
The stiff side of the shaft and the vanes
The third characteristic in that combination, the variation in stiffness around the shaft, is the newest. Spine is measured at a single point, whereas the bench in the 2025 study flexed each shaft at every sixteenth of a turn for at least four full turns, sixty-four measurements per shaft. Stiffness almost always followed the same wave, with two peaks and two troughs per turn, in other words a strong axis and a weak axis. That shafts have a stiffer side is therefore reported, but the difference is small: on average 0.014 millimetres of deflection out of about 8.3, ranging from a thousandth of a millimetre, practically nothing, up to 0.038. When shot by the machine, the shafts with the widest wave landed further right, and a few showed a slower, different wave that the authors read as a possible manufacturing or batch defect; that these are the shafts that leave the group, however, remains their interpretation.
This is the origin of a widespread habit, fletching every arrow with the same orientation relative to the stiff side, usually taking the label as the reference. The same study recorded the angle between the reference vane and the stiffest point of each shaft, and that angle did not enter any of the models predicting the point of impact: that vane position relative to the stiff side does not change the point of impact is reported, from this one test. The detail that struck me most is a different one: the label is printed at random relative to the stiff axis, so anyone who fletches off the label believes they are orienting the vanes and in fact places them at random. The question that matters most on the range remains open, namely whether turning the nock by 120 degrees, as archers do with an arrow that leaves the group, moves the shot: the authors leave it open, and nobody has yet checked that it brings the arrow back into the group.
Even less is known about an idea that circulates among people who fletch their own arrows, that some shafts leave the bow already spinning in one direction and that the vanes should be fletched with a helical twist in the same direction. The only source is a 2019 review reporting unpublished work, with neither data nor method, and for the registry the question is not studied.
Straightness and nocks
The straightness tolerance is printed on the box too, and studies of its effect point in opposite directions, so the registry holds it as contested. With a compound in a machine at 25 m, aluminium shafts straight to within a thousandth of an inch grouped in 27 millimetres, bent to three thousandths in 30 and to six in 48, and between two carbon series of the same model, the one straight to a thousandth grouped in 30 millimetres against 42 for the one straight to three. In the 2025 study, by contrast, straightness did not enter any model, and the data show why: those X10s were far straighter than the tolerance required, 0.007 millimetres on average with a maximum of 0.015 against the stated 0.025. Read this way the contradiction may only be apparent, because within a modern competition set straightness varies too little to count, while between shafts of different grades it can matter.
The 2017 test also holds a finding that concerns the nock more than it might seem. Before the main trials began, on the same straight shafts, the group went from 44 to 27 millimetres just by changing the nock (the worst one also needed an extra adapter), and on another shaft size from 32 to 15. That the type of nock changes group size is reported from this single test, run for another purpose and with a compound shot by a machine, so it says nothing directly about a recurve shot off the fingers; it does show, though, that on the same shaft the nock can weigh as much as straightness.
Arrows and wear
People say arrows do not tire, and that as long as the shaft is intact it stays as it was on the first day. The only test is a 2019 preprint in which one arrow per material, carbon, aluminium and cedar, was taken through about 600,000 vibration cycles and then measured again. The authors write that nothing significant changes, but in their tables the estimated stiffness after the cycles is higher in every measurement, by 14 to 52%; the estimated mass of the carbon arrow also falls, though, which an intact shaft should not do, so part of those differences is noise from the method. The claim that arrows do not change is therefore contradicted: the evidence there is goes against it, but carries little weight, and it says nothing about impacts on the boss or arrows striking each other, which on the range add to the vibration.
What I take from it on the range
Put together, these tests suggest an order of work more than a recipe. The first choice, spine, is made from the chart for that shaft model, with draw weight measured at your own draw length and the real length of the arrow, bearing in mind that at the extremes, with low draw weights or long arrows, one model finds the chart too stiff and that confirmation comes from tuning. The second is not to mix dozens, and to weigh points and shafts separately before assembling them: pairing the heaviest points with the lightest shafts is what the 2025 study proposes, and its effect on the target is, for now, a calculation.
Then you number the arrows and watch the groups over many ends, arrow by arrow. In a small test a shooting machine picked out defective arrows at 70 m, but in the only comparison made on the same range, with a single archer and different arrows, the archer scattered more than the robot, over three times as much vertically, and a difference of a few hundredths of a gram or a few thousandths of a millimetre can easily stay below the noise of your own shooting. That is why an arrow that leaves the group should be watched for a long time before it takes the blame; turning its nock is a common attempt, although an untested one, and if it keeps straying it is set aside. Many archers check straightness by spinning the shaft over their fingernails, and the authors of the 2017 test write that by feel a shaft straight to a thousandth of an inch can be told apart from one at three or six thousandths. The check is mostly useful for spotting a shaft bent by an impact, because the X10s measured in 2025 left the factory far straighter than their tolerance.
Then there is proportion, which is what I repeat most often: the mean positions of the unmatched arrows in the French test spanned 8.8 centimetres vertically, less than the diameter of the 10, and they were shot by a machine, so without the archer's error, which in competition adds to theirs. Selection removes a small part of the error, useful once the rest of the shot is already stable, and it does not take the place of work on technique.
Where the research stops
Almost everything known about arrow selection comes from a machine, which removes the variability of the release and of aiming. The most complete test used one bow, one setup and one of the stiffest spines on the market, and the authors themselves warn that things could change with softer shafts. Three comparisons are missing: a set selected by weight and spine against one also selected for the stiffness variation around the shaft, shot by a machine and then by archers; well-matched shafts against mismatched ones, shot by archers with many arrows per condition; and a check on whether turning the nock brings back into the group an arrow that has left it. Until they exist, arrow selection rests on careful laboratory measurements and machine tests, while the step to an archer actually shooting has yet to be shown.