"Why do my arrows go left?", or right, or low, or high: sooner or later every archer asks it on the range. A single arrow cannot answer, whereas a group of a dozen says a great deal, provided you read it in the right order: first how tight it is, then where it sits relative to the centre, and finally what shape it has. Almost nobody looks at the target this way, because we look at it to find out how we did, not to understand why.
Some of these readings have measurements behind them, and I give each one the weight it has. That twisting the bow with the hand moves the arrow is shown by models and machine tests; that the way the hand spreads pressure over the grip matters, on the other hand, is not borne out by the only study that has measured it. At one World Cup stage off-centre groups cost between 1.35 and 2.92 points per 36 arrows at 50 m on average, and elite archers' groups are usually a little wider than they are tall. The rest is the map I use with the archers I coach: it does not hand you a ready-made diagnosis, but it narrows things down quickly and saves you months spent fixing something that was never the problem.
The target is a report, not a verdict
The first rule is also the hardest to follow: never read an arrow on its own. A single arrow means almost nothing, because you can land a perfect one by luck and a terrible one from a flicker of distraction, and neither tells you how you actually shoot. What matters is where the arrows end up when you look at ten, twelve, a whole series.
The reason is statistical: every shot is the sum of many small variables, such as the pressure of the hand on the grip, the exact point where the string touches the face, the milliseconds of the release, one extra heartbeat. Some of these variables are random and cancel out across the group, while others are systematic: they repeat identically on every arrow and push the whole group in the same direction. Only by looking at the group can you tell the two kinds apart, so from here on forget the single bad arrow that made you angry and ask yourself two questions in the right order.
Two questions before all the others
Before you even ask "why to the left", ask yourself two other questions, because they point to different problems, each solved in a different way: the first is whether the group is tight or wide, the second whether it is centred or off. The two need to be kept apart: you can have a tight group that sits entirely off centre, or a wide group that is roughly centred, and they are opposite situations. And when you change distance there is a third question worth as much as the first two: how your group grows with distance.
Written down like this, the distinction looks obvious, yet ignoring it is the diagnostic mistake I see most often: an archer has a wide group, moves the sight because "the arrows go left", the group stays just as wide a little further over, and the archer keeps chasing it. Moving the sight shifts the centre of the group without tightening it, so if the problem is dispersion, the sight has nothing to do with it.
A tight but off-centre group is almost good news, because it means your shot is repeatable: you do the same thing on every arrow, and that same thing simply points a little off centre, so a sight or tuning correction is enough. A wide group is a different and longer job, because it is about consistency, which is built on the movement rather than on the sight screw.
The score on its own cannot tell the two cases apart. On paper a tight off-centre group and a wide centred one can produce the same total, and the only study to have measured score and arrow position on the same shots sees it happen: in twelve young Indonesian archers, shooting an electronic bow at a target projected on a screen, two of them a point apart, at around 125 out of 150, had a spread of about 10 and about 19 cm. In the registry the claim that two archers with the same score can group very differently is reported, because it shows that this can happen, not how often it happens at competition distances. With level, on the other hand, the two improve together: at eighteen metres beginners' spread is five to seven times that of high-level archers, and seven independent samples all point the same way, that the higher the archer's level, the tighter and better centred the group. That claim is reported too, because these are comparisons between different archers rather than the same archers followed as they improve.
What an off-centre group costs
A tight but off-centre group can be fixed in minutes, yet at major competitions it costs points. One study photographed the target faces of elite archers at the 2013 Shanghai World Cup and World Championships and recalculated each archer's score after moving the centre of the group onto the centre of the target: at the World Championships they would have gained about three points each on average, with cases of around twenty points in men's recurve. A rereading of the same targets with a model gives the loss division by division: in Shanghai, at 50 m and over 36 arrows, 2.34 points in men's recurve, 2.92 in women's recurve, 1.35 in men's compound and 1.48 in women's compound, while the top twenty men's compound archers lost 0.30 on average.
Two details of that work are useful to anyone who shoots. The first is that the loss grows with the square of the offset, so a group twice as far off costs about four times as much, and for the same offset it weighs more heavily on better archers. The second is a concrete example: a recurve archer whose group sat 25.9 mm low and left scored 335 at 50 m, where the same group centred would have scored 341, which over 72 arrows comes to about 12 points. To get a sense of scale on the 122 cm face, a model by the same author calculates that at 70 m three points lost per 36 arrows correspond to a group whose centre is 32 mm from the centre of the target, still inside the ten.
In the registry the claim that at big competitions many archers lose points to a badly adjusted sight is reported, and its limit is always the same: nobody recorded the sight settings, so an off-centre group can be explained by the sight or by the archer. The author also warns that the losses are probably underestimated, because archers were correcting their sights during the distance. That the margin is real is suggested by a follow-up with no control group, in which six Australian recurve archers, after being shown where their groups had landed, went from seventeen potential points lost to three at the next World Championships. As for how to correct, the model suggests moving the sight after each end by only part of the offset of the group's centre, between a fifth and three fifths depending on the archer's level and the starting offset, because the centre of three or six arrows wanders a long way around the real one.
A smaller measurement looks at where the group falls without working out what it costs. Eighteen archers of mixed ability shot twelve arrows each at 18 m indoors, and of the two hundred left on the face 127 landed in the upper half and 73 in the lower half, while between right and left, 104 against 96, there was no difference. The claim that the upper half of the target collects more arrows than the lower half is reported, but it should be read as a snapshot of a single session: the asymmetry concerns the golds and not the red or the blue, in the text the difference between the two halves carries a p greater than 0.05, and the two hundred arrows are treated as independent although they come from eighteen people, so two or three archers with a slightly misadjusted sight would be enough to produce the whole result.
The shape of the group is a signature
When the problem is dispersion, look at the shape it draws. Not all wide groups are the same: a group stretched vertically has a different origin from one stretched horizontally, and one scattered at random tells yet another story. The shape does not hand you the exact diagnosis, but it narrows the field before you even touch the equipment.
If the group stretches up and down, something that raises or lowers the shot is changing from one arrow to the next: a draw length that is not always the same, an anchor that rides up and down a touch, a release that sometimes goes a moment early and sometimes late. If it stretches left and right, look at the release, which is clean on some shots and plucks sideways on others; at the bow hand, which turns a little; and at the head and the string, which do not line up the same way every time.
Before you read a fault into a group that is slightly wider than it is tall, though, it helps to know that this is the commonest shape even at the top. At the 2013 Shanghai World Cup, with almost no wind, most elite archers' groups at 50 m were wider than they were tall, with an average ratio of 1.15 between horizontal and vertical spread in both recurve and compound, and at the windy World Championships of the same year the ratio rose to 1.25. Indoors at 18 m a conference abstract finds the same pattern in high-level and middle-level archers and the opposite in beginners, whose spread is mostly vertical. That is why arrow groups are wider horizontally than vertically is a reported claim, with a plausible mechanism, because while aiming the bow sways more sideways than up and down. What should make you suspect a fault, then, is a clear stretch, well beyond that 15%.
The round group, scattered more or less everywhere, is the hardest to read, because it almost never has a single cause. Usually it is many small things adding up, and many archers see it appear when fatigue sets in at the end of a session, or when their head is not in it and the routine breaks down. On fatigue the measurements are few. In twelve Italian archers, after sixty 20-second holds at full draw, sideways bow sway increased in every one of them, by 20% on average, without any change in score, and in the registry the claim that muscle fatigue increases bow sway while aiming is reported; that at the end of a training session archers can still shoot, but not with good form rests mainly on a questionnaire of 59 kyudo practitioners, and it is reported too. What fatigue does to the group in a real session has not been measured. The two separate groups, on the other hand, tell you something precise: you are shooting in two different ways, perhaps a clean release and a "grabbed" one, or you changed something mid-session without noticing.
Horizontal or vertical: which way to look
That brings us back to the original question: if the group drifts mostly sideways, or mostly up and down, the likely causes are different, because they come from different parts of the movement. What follows are tendencies, not laws, since they depend on your dominant hand and your bow type and always need checking on the individual archer: use them as a starting map, to know where to begin looking.
-
L / R
The horizontal axis
Sideways, the release matters most: a messy release, a sideways pluck, a bow hand that rotates even a little. After that come the alignment between head and string and, on the equipment side, arrow spine and centre shot. If the group is compact but shifted sideways, look at tuning first; if it is scattered sideways, look at the release.
-
High / Low
The vertical axis
Vertically, what counts is the consistency of draw length and anchor, because a few millimetres of difference change the speed at which the arrow leaves the bow. Next come release timing, back tension through the finish and the nocking point. In 3D and field archery, where the distance is not marked, the prime suspect joins the list: a misjudged distance.
-
Careful
Not a lookup table
"Left equals pluck" is the kind of shortcut that wastes time, because the same direction comes from different causes for a right-hander and a left-hander, or for a recurve and a compound. The axes tell you where to look, but they will not give you the answer: use them to decide where to start, then verify with a test.
There is also a measured reason not to turn the map into a lookup table. Every study that has tried to work out how much of a single arrow's score depends on movement finds a minority share, or one that varies widely from archer to archer: in the strongest sample, thirty-nine elite recurve archers at 70 m, clicker reaction time, draw weight and sway speed after release together account for 42%, and in the world number two archer five shoulder and trunk parameters, over 60 arrows, account for 17.9%. The claim that the movement parameters measured during the shot account for only part of an arrow's score is rated as reported in the registry, because these are associations from a single session on small samples. For anyone reading a target it means that a stray arrow almost always has more than one cause in the movement, which is why you read the group.
Sideways: the bow hand and bow rotation
Among the causes of sideways error, the one where the mechanics are clearest is torque on the grip. In a model of a 166 N recurve, thirteen hundredths of a degree of difference in torque from one shot to the next moves the arrow by more than one ring at 70 m, and corresponds to about 1.7 mm of misalignment between the arrow point and the front stabiliser as seen at full draw. As early as 1990, on a shooting machine at 30 m, each degree of bow rotation moved the impact 15.8 cm in one direction and 21.9 cm in the other, and on a mechanical launcher with a Japanese bow, varying torque alone moves the point of impact at 28 m from one side of the target to the other. In the registry the claim that twisting the bow with the hand moves the arrow on the target is reported and no more than that, because the only measurement on people comes from kyudo, and torque and point of impact have never been measured on the same shot of a recurve archer.
Torque is not a beginner's fault. In high-speed footage of sixteen national-team recurve archers at the test event for the London Games, seven showed signs of twisting the bow: the tail of the arrow cleared the button by 2 to 5 mm, or by 12 or 25, instead of the usual 8 to 10, and the same images often showed where it came from, with the wrist out of the plane of the bow or the tip of the thumb pressing on the side of the riser. The claim that even among elite archers, many show signs of twisting the bow is reported, with one limit: the torque is inferred from the behaviour of the arrow, and the authors note that if the movement is repeated identically on every arrow the score can stay high all the same. What opens the group is not a constant twist but how much it changes from one arrow to the next, which is also what the models translate into rings.
The same footage answers a question archers often ask, namely whether the bow rotates while the arrow is still on the string: in none of the sixteen did the bow move sideways or rotate measurably about the vertical axis before the arrow was in free flight. The torque that matters is therefore the one set by the hand while aiming, not a movement of the bow during the release. On the compound the question moves to the vertical plane: a dynamic model calculates that during the power stroke the bow rotates one way with a single cam and the other way with hybrid cams, and that no position of the centre of mass can cancel that rotation, but the resulting movement of the nocking point is small compared with the one produced by the geometry of the cams. The claim that while it drives the arrow a compound bow rotates in the vertical plane is therefore transferred: it rests on a calculation and on an accelerometer check whose data the authors do not show.
If torque moves the arrow, it seems natural to think that where the hand presses on the grip also matters, higher or lower, more on the thumb or on the heel of the hand. The only study to have measured this directly does not find it: fifteen experienced recurve archers each shot eighteen arrows at 30 m with force sensors built into the grip, and taken one archer at a time the spread of pressure was related to where the arrow landed in one archer out of fifteen for up and down and three out of fifteen for left and right. In the registry the claim that how hand pressure is spread over the grip does not predict where the arrow goes is reported in the direction of that result, because with few arrows each a small effect could have slipped through. What can be glimpsed elsewhere concerns the consistency of the grip rather than its form: a shooting machine loses accuracy when the grip on the bow changes from arrow to arrow, and in four traditional archers each one shot best with a different grip from the others. For the bow hand, then, the useful work is a grip that lands in the same place every time and does not twist the bow differently from one arrow to the next.
The last variable of the bow hand is cant, the sideways tilt of the bow. In a scoring model, for the archer who finished eighth in men's recurve at the 2013 World Championships, of the 30 points lost at 50 m about 5 would be down to the wind, 5 to variation in bow cant, 3 to the sight and 17 to the other errors, but this is a calculated split, not a measured one. Only one piece of work has tried to act on cant: a device with an accelerometer and a compass, fitted to a beginner's bow, signals with a light or a sound when the tilt and direction of the bow come back within one degree of those recorded on shots that landed in the ten, and with the signal the average distance between the three arrows of an end fell from 17.3 to 10.7 cm at 30 m. The claim that a signal telling beginners when the bow is at the right angle tightens their groups is reported, but on the thinnest evidence there is: probably a single archer, no statistics and a device that was new to the person using it.
High and low: launch speed, the nock and the air
Vertically, the most measured mechanism runs through the speed at which the arrow leaves the bow. An arrow that leaves slightly faster drops less in flight and lands higher: at 70 m, with a compound launching at 74.5 metres per second, one metre per second more raises the point of impact by about thirteen centimetres, almost twice what a one-metre error in distance or 20% more aerodynamic drag would do. It is also the thread that links the causes on the map, because a draw a few millimetres shorter or longer changes the energy the bow gives the arrow, and a shooting machine with a recurve at 70 m spreads its arrows less vertically when the release is triggered by the clicker, that is, with the same draw length every time.
The measurements, however, are almost all on machines. In fifty-nine Easton X10 arrows of the same spine shot by machine at 70 m, mass and mean spine account for 46.7% of the vertical spread, with the heavier arrows landing lower, and the same study cuts the size of the effect down: the ballistic calculation predicted over fourteen centimetres of vertical range, while the measurement gave 8.8. In archers, moreover, speed varies very little, one or two kilometres per hour from one arrow to the next in three national-team archers, and across seven archers from the German national squads neither mean speed nor its consistency entered the model of the score. That is why the claim that small differences in launch speed account for much of the vertical spread is reported: the direction is solid, the size needs revising and the link has never been measured in a group of archers.
Then there is the nock. On the range you hear that a nock too loose on the nocking point makes the arrow change height, and the idea makes sense, because the arrow would sit in a slightly different place on the string each time and leave at a slightly different angle. How much the arrow's height on the string matters is shown by a machine test: raised by 10 mm, the tear in paper at 2 m goes from under a millimetre to 40. A nocking point moved on purpose is not a nock that wobbles, though, and the only work that mentions a loose nock is the commissioning of a shooting robot, where the authors put a vertical spread down to it without ever changing the nocking point, and also point to a second cause, a draw that was not always the same. That is why in the registry the claim that if the nock is loose on the nocking point, the arrow changes its impact height is marked as not studied. Checking that every nock clips onto the string with the same light snap is still a few seconds' work, and I cover it in the article on bow tuning.
The last explanation comes from the air. A Japanese research group has measured several times that the same arrow, at the same speed, can fly with two very different levels of drag: within a certain speed band the drag coefficient of an A/C/E arrow splits into two families, about 1.5 to 1.6 with a laminar air layer and about 2.6 with a turbulent one, and what decides it is the arrow's attitude, because four to six tenths of a degree of incidence are enough to turn the flow turbulent. The authors calculate that the difference is worth about 15 cm in height at 60 m. The claim that within a certain speed band the arrow can fly in either of two regimes, and that shifts the point of impact in height is reported, but for archers its weight is limited. The measurements come almost entirely from arrows fired by compressed air, which do not flex. The arrow shot by an elite archer at 58 metres per second gave only the turbulent value, and with a bow shot by a robot the flexing of the shaft makes the flow turbulent across the whole band when the release is made with mechanical fingers, while with a mechanical release, which barely flexes the shaft, the two values return. The shift in height, moreover, is calculated rather than counted on the target.
Away from target archery, finally, the prime suspect is a different one. In 3D and field archery the distance is not marked, and in twenty-eight archers of three styles the error band the study classes as average goes from 2.8 to 5.1 yards at 17 yards to 19.3 to 31.5 yards at 55, which is about 50 m. For the registry, when the distance is unknown, judging error grows with distance is a reported claim: it comes from a single descriptive study, which does not look at the score. For an order of magnitude, in the calculation for a compound at 70 m a one-metre error in distance moves the impact by about seven centimetres.
The method: one variable at a time
Reading the pattern is half the work, while the other half is resisting the temptation to change three things at once. Serious diagnosis works like an experiment: you isolate one variable, change it, measure again. If you change grip, nocking point and sight in the same afternoon and the group improves, you will never know which of the three helped, and you will not be able to repeat it.
The first step is to have enough arrows before you conclude anything, at least a dozen at the same distance. The reason can even be calculated: the centre of a three-arrow end typically lies about 58% of the group's spread away from the group's real centre, and that of a six-arrow end about 41%, so a single end is not enough to tell where the group is, and one flyer is noise rather than a pattern. The second step is to call your shot: before you look at where the arrow went, tell yourself where you think it landed. If your calls match the target, you are feeling your errors as you make them, so they come from execution; if they do not match, you have a tuning or perception problem, and that calls for a different approach.
The third step is to keep execution and equipment apart: a shot you felt go badly is an execution error, not data about the tuning of the bow, and it should be dropped from the equipment analysis. To take aiming out of the picture and isolate the movement, the best tool is still blank bale work, shooting at close range without a target face. To check tuning objectively, comparing a fletched arrow with a bare shaft tells you a lot about spine, nocking point and clearance without your having to interpret feelings. Then you change one thing and go back to measuring.
When it is not (entirely) your fault
Not every pattern comes from the movement: some arrive from outside, and correcting technique when the cause is external is the fastest way to get worse. Wind and light shift the group with the conditions: if the group shifts when the flag moves, it is the wind moving it, not a mistake of yours, and a group that walks with the weather should not be read as a form fault.
Then there is the equipment, which draws patterns that can look like your own errors: an old string that has stretched, a nocking point that has crept over time, a rest that has worked loose, a clicker that has slipped, an arrow that is bent or has a point different from the others. If a pattern appears suddenly without your having changed anything in the movement, suspect the equipment first, because the body changes slowly and a screw does not; and if it is always the same arrow that leaves the group, numbering your arrows makes it obvious within a session.
On matching the arrows in a set, by contrast, the measurements counsel calm. Fifty-nine Easton X10 arrows of the same spine, not matched by weight and shot by machine at 70 m, held a vertical range of 8.8 cm, less than the 12.2 cm diameter of the ten. In the registry the claim that matching the arrows in a set by weight and nominal spine is not enough is reported: how close the arrows land is better explained by mass, mean spine and the way stiffness varies round the shaft, but the tests are all done on the bench or by machine, with no archer, and matching is not what separates a ten from a nine. In the same experiment the angle between the vanes and the stiff side of the shaft did not enter any model of the point of impact, and the manufacturer's label turned out to be printed at random with respect to that side: the claim that where the vanes sit relative to the shaft's stiff side does not change the point of impact is reported, on a single study with one bow and one spine.