Anyone moving from recurve to compound, or trying a clubmate's bow, notices it on the first draw: two compounds with the same draw weight can feel like two different bows. One climbs sharply and lets off early, the other holds its load for longer; one rolls gently into the valley, the other stops dead against the wall; at full draw one is held with very little, the other asks more of the shoulder. On the shooting line people say it comes down to the cams, that the limbs matter less, and that the nocking point must be set so the arrow leaves straight. A small body of engineering and physics research covers these three parts of the compound, and here I follow it to see what it has measured and how firmly.
First, none of these studies measured what the archer feels. Force is read with a load cell or a winch, and many of the results come from mathematical models calibrated on a single bow, almost all of them by two authors. That is why the registry's claims about the compound sit between transferred and reported: the direction is consistent from one study to the next, but it rests on calculations or on a handful of measured bows.
The shape of the force-draw curve
A compound's draw curve has a shape the recurve does not. On a recurve the force rises steadily over the whole draw, and at full draw the archer holds the maximum while aiming. On a compound, as one of the key papers describes it, the force climbs quickly over the first part of the draw, stays roughly constant for much of it, drops into the valley just before full draw and then rises abruptly at the wall, which the archer leans against while aiming; the position of the wall therefore sets the bow's draw length.
The most complete measurement on a production bow was made by an American physicist in 2015. The force rose to about 200 newtons and dropped by about 65% at full draw, so the archer held roughly a third of it while aiming; integrating the curve, the bow stored 73.7 joules and gave back 71. Two further single-bow measurements give 96.5 joules for a twin-cam compound, derived from a model that reproduces its measured curve, and 92.8 for a 53.5 lb target compound, measured in a study on wind alongside a recurve that stored 45.1. For this reason the claim that a compound stores between 70 and 100 joules and sheds about two thirds of the force at full draw is reported: the three bows fall within the range, but let-off was measured on only one, and draw weight and draw length change from study to study.
The comparison with a recurve at the same peak comes from a 2012 model in which the recurve exists only on paper: at equal peak force the compound stores 75% more energy, and at equal energy it is held at full draw with 2.3 times less force. How the virtual recurves' curves were built is not explained, so the claim that at the same peak force a compound stores far more energy than a recurve and is held at full draw with far less force remains transferred. From the archer's side there are only two master's theses, from 2000 and 2001, on seven and six archers shooting a recurve and a compound of 55 lb: with the compound, activity in the shoulder, upper arm and forearm fell after let-off. It is the closest the research comes to what aiming feels like, and it measures neither force nor energy.
What the cams shape and what the limbs shape
Almost everything known about what sets this shape comes from three mathematical models by the same author, published between 2015 and 2017 and calibrated on real bows. The first shows that the geometry of wheels, limbs and cables is enough on its own to rebuild the measured curve of a round-wheel compound. The second moves on to a commercial twin-cam compound and finds the calculation very sensitive to the cam profile: redesigning the cams for the same peak and draw length would take the stored energy from 96.5 to 106.5 joules, about 10% more, gained almost entirely by making the first part of the curve steeper. At the same peak and draw length, in fact, the curve that stores the most energy is a rectangle: a force that reaches its maximum at once and stays there to the end.
The third paper returns to round wheels and changes the eccentric parameters one at a time, and these turn out to be the most influential: they move the peak, change the shape of the curve and set the let-off. With the cable eccentric offset by 16.7 millimetres the let-off reaches about 80%, but the curve empties, and its quality coefficient, a measure of how close it comes to the rectangle, falls from 0.619 to 0.589. Changing the starting angle of the eccentrics has the opposite effect, filling the curve and raising the coefficient to 0.685 or 0.737, but the let-off falls to about 50% or below.
The same paper then builds three virtual bows with the same peak, 223.9 newtons, and the same draw length, changing wheels and limbs together. The third stores 18.5% more energy than the first, but at full draw it is held with 88.3 newtons rather than 73.1, which means a let-off of about 61% rather than 67. The second, though, stores more energy than the first with almost the same let-off, so the trade-off between the two is not a fixed rule.
All of this is calculation: nobody has fitted different cams to the same bow and measured what changes. So the claim that on a compound the shape of the force-draw curve is governed by the cam profile is transferred: the three papers agree on the direction, but the energy gains are simulation figures.
In the same models the limbs do something else. In a 2016 paper on limb deformation, changing the material, thickness, width or profile in the calculation mainly shifts the height of the curve and the full-draw position, while the shape stays much the same; thickness matters far more than width, because stiffness grows with its cube. Even the limb profile, which changes a recurve's curve a great deal, makes little difference on a compound, because the force always acts along the axis of the cam. The claim that on a compound the limb mainly changes how much force is needed, and changes the shape of the force-draw curve little is therefore transferred, like the one on cams. Together they give a simple picture: the limb sets the height of the curve, the cams its outline. It is a picture built from models, though, and from a single author.
Between draw and let-down
The curve is also not the same on the way out and on the way back. Drawing a 55 lb round-wheel compound slowly with a winch and then easing it back, the force on the draw was systematically higher than on the return, and a year later the same happened on a twin-cam compound. The author puts it down to friction in the wheels and hysteresis in the limbs, string and cables, but gives no figure for the difference and does not separate the two causes. The claim that on a compound bow the draw force is a little higher than the let-down force is reported: two measurements by the same author, one bow at a time, observed in passing while calibrating a model.
The two limbs
Almost every compound tune starts from one premise: the upper and lower limbs are identical. The only measurement dates from 2015, on the compound used to calibrate the model, where limb length, wheels and resting angles, measured separately at the top and the bottom, showed no significant differences according to the author. It is a single bow, the separate values are not published and the stiffness of the two limbs is not compared, so the claim that on a production compound the upper and lower limbs are identical is reported in the narrowest sense: on that bow no differences emerged.
Identical parts do not mean identical bending, though. The nocking point almost always sits above the centre of the string, so the two halves of the string meet the limbs at different angles and load them differently. In 2010, on several compounds with different cam configurations, the deflections of the two limbs matched to within half a degree over the whole draw, and in the worst case, a short axle-to-axle bow, the calculated difference of 0.22 degrees shifted the nock path by half a millimetre. In 2012 a model that does not impose symmetry, checked against a photograph of a compound at full draw, found limb angles of 0.124 and 0.110 radians, measured at 0.125 and 0.106: a difference of 12% of the mean angle, which the author judges to be many times larger than the measurement errors.
The two papers agree on the direction but not on the size, so the claim that on a compound the two limbs do not bend by the same amount at full draw is reported. The 2012 paper also scales down the consequence: treating the bow as symmetrical changes the limb angles, but the forces in the string and cables change by between 0.1 and 0.2%, about as much as the measurement errors. How much the difference matters for tuning, or on the target, has not been measured.
The path of the nocking point
On the nocking point the mechanics have been worked out most carefully, almost entirely by one author. A single formula that applies to compounds, recurves and longbows shows that with the nock above the geometric centre of the string, as it nearly always is, the nocking point rises throughout the draw, and rises more the further it sits from the centre. On a production single-cam bow with 0.865 metres axle to axle, the path without correction would stray 21.8 millimetres from a straight line; with the cam grooves corrected by the manufacturer, the measured path stays within 1.5. In 2009 the same author had measured another single-cam bow whose path stayed within one millimetre of a straight line.
Geometry decides it. A longer bow axle to axle strays less, while the resting angle of the limbs matters little. Counting the cams' degrees of freedom, only single-cam and hybrid-cam bows can straighten the path by design; on twin-cam and binary-cam bows all that is left is to bring the nock closer to the centre of the string. Putting it exactly at the centre would remove the problem, but would place the hand well below the middle of the bow, which then feels top-heavy, something archers generally avoid.
Why the path matters is explained by a second paper from the same author, a model of the arrow launched from a compound: in its sensitivity analysis the arrow's behaviour changes little with the string force or with the rest spring, and markedly with the path of the nocking point. An independent 2012 model then finds that at full draw the bow's force does not point along the arrow, because 104.5 newtons along it come with 9.4 across it, about 9%. The registry counts these papers as models, since the measurements on bows serve to check the calculation, so the claim that on a compound the nocking point does not travel in a straight line during the draw, and how far it strays depends on the bow's geometry is transferred. A group of arrows or a score comparing a corrected bow with an uncorrected one has not yet been measured.
The rest, the arrow and the release aid
The same arrow model also addresses where to set the rest. Moving it 2 millimetres sideways mainly changes the angle at which the arrow enters the air, from 0.1 to -0.2 degrees, while the path of its centre of mass stays very similar; the author concludes that the position is not critical, provided the back of the arrow, above all the leading edge of the fletchings, clears it. The margin does depend on the cams, though, because with the nock path of a binary-cam compound the rest has to sit 4 millimetres below the nock rather than 2. In the same calculation torque on the bow carries more weight, and a single degree is enough to take the horizontal angle to -0.4 degrees. With no shooting tests behind it, the claim that on a compound the rest position matters little, as long as the arrow passes freely is transferred.
With a single-cam compound mounted on a shooting machine and a release aid, photographs and the model gave a maximum vertical flex of a few millimetres in the arrow, against the 10 to 15 the author expects, on the basis of other work, from a recurve. The two are not compared in the same test, and the claim that with a compound the arrow flexes less than with a recurve and leaves almost aligned is transferred. A direct comparison between release aid and fingers has been made only once, on a recurve mounted on a robot: with the release aid the arrow oscillated by 1.22 millimetres just after leaving the bow, by 6.86 in shots from two Japanese archers of national standard and by 14.8 with a mechanism that imitates the fingers. The claim that with a mechanical release the arrow flexes far less than it does off the fingers is reported: the test does not involve a compound, it counts 72 shots in all, and it did not look at where the arrows landed.
Where the studies stop
Taken together, these papers explain the mechanics of the compound well and say almost nothing about what the archer feels: curves are measured one bow at a time, cams and limbs change only in calculations, and none of the work reaches a group of arrows or a score shot by an archer. The starting question, what makes one draw easier to manage than another, has not been tested: the two theses on muscle activity are the only place an archer appears, and they compare a compound with a recurve, not two compounds with each other. A broader comparison between the disciplines is in the article on barebow, Olympic recurve and compound.
What I take from it for the shooting line
None of this points to one cam being better than the others, nor to a correct let-off; what it gives is a few ways of reading the bow in your hands, with the limitation that they come from models. The first is that the shape of the curve and the let-off are design choices made in the cam, not fixed properties of the compound, and in the 2017 model they often pull in opposite directions: a cam that lets off more does not store more energy for that reason. If two bows with the same peak have curves of a different shape, the calculations put the difference in the cams rather than the limbs, which mainly shift the draw weight; choosing between two profiles still means trying them.
The second concerns the nocking point. On a twin-cam or binary-cam bow, where the cam cannot correct the path, the only lever the geometry leaves is to keep the nock as close as possible to the centre of the string, within a hand position that does not make the bow top-heavy. On a single-cam or hybrid bow corrected by the manufacturer, the position the calculation predicts is level with the rest mounting hole. Neither point has been tested on the target, and I would use them as a starting point for tuning the bow, not as a recipe.
The third concerns the rest. In the model, clean clearance for the arrow, fletchings included, matters more than its exact position, and a bow hand that does not twist the bow matters more still. Before moving the rest by a millimetre I would check whether the fletchings touch it, with slow-motion footage or a dusting of powder, and whether the hand turns on the grip from one arrow to the next. For anyone coming from recurve there is one last difference to bear in mind: with a compound and a release aid the arrow flexes much less, and what is known about spine on the recurve, where the flex starts with the string sliding off the fingers, needs reading with that in mind.