Kellets add what is effectively a point weight to the rode, and are intended to perform one of a number of possible tasks. They are frequently hailed to increase the ultimate performance potential of the anchor, although with an adequately sized modern anchor, this claim is less than completely correct.
Buoys employed in a similar position to kellets are intended to provide shock absorption, as the rode must pull the float down against its buoyancy before it can straighten.
The kellet, also known as a sentinel, angel, chum, or buddy, can consist of any dead weight which is then attached to a point on the rode. A common method of making one is to simply loop a section of the chain, or using a bundled length of separate heavy chain. A secondary anchor could be used, or dedicated kellets are available which make use of specialist deployment and retrieval mechanisms.
Kellets are thought by many to improve the possible performance of the anchor (i.e. the point at which it will drag), by reducing the rode's angle of pull on it. Minimization of this angle is important and does improve performance. However, how much effect weight can have is limited.
The article on Peter Smith's website linked to above includes some simple computer simulations of a rode with a kellet subject to certain forces. Forces and the rode and kellet are all sized realistically for a boat in the order of 12m / 40'. The mathematics are not difficult, and it is quite simple to calculate how much effect the kellet makes. In essence, the answer is: not very much.
Kellets may be used for the various purposes outlined below. However, their position along the rode varies depending on the desired effect, which means they cannot perform all at once.
As mentioned above, kellets make little practical impact on the rode when conditions are such that the anchor is likely to be troubled. Once the rode is straight, the presence of the kellet makes no difference whatsoever. Nevertheless, to maximize what effect there is, the kellet should be placed as closely to the anchor as possible. It should not however be placed directly next to the anchor, or it will just sit on the seabed and be utterly useless.
To give the largest possible 'spring' to the rode, the kellet should be placed about halfway between the anchor and the boat's bow. This halfway point results from a balance between potential distance for the kellet to travel (from the seabed up to the point where the rode is straight), and leverage gained by placing it closer to the anchor. Like the lowered pull angle however, this use is not recommended, for the same reason: the effect disappears in bad conditions when it is most wanted.
This is really the sole use for a kellet which makes sense, although it only applies in light to moderate conditions. It will prevent the boat from aimlessly drifting around the anchor in calm conditions, and will minimize the swinging circle in light to moderate winds. The kellet should be positioned down the rode from the boat a distance equal to that of the vertical distance to the seabed (depth plus height to bow-roller). This allows the gravitational potential of the kellet to work to its full potential.
While this technique is inarguably effective, its necessity could be questionable. Swinging circles are unlikely to be of much importance in an empty anchorage where there is room (if it is small, the amount of rode deployed should be according to this). On the other hand, if the anchorage is crowded, your swinging circle should be similar to that of other anchored boats, which generally will not be using kellets. That said, where there is a bunch of mixed boat types, which may each react differently to wind and currents, a kellet can be effective at minimizing the odds of a "chance encounter" during light airs.
A buoy can be tied to the rode in order to provide the same shock absorbing properties of a kellet, only the orientation of the 'spring' is reversed. Rather than lifting a kellet in order to straighten itself, the rode must instead sink the buoy. An advantage over a kellet is that rather than requiring dead weight, its force is generated by its buoyancy (dictated by its displacement), which means it is lighter to stow onboard.
The ideal position to maximize the buffer effect is, as with a kellet, halfway between the anchor and vessel. However, this effect disappears as the force in the rode overcomes the buoyancy just like it does the catenary from the chain and weight from a kellet. The straighter the rode, the less distance the buoy has over which to provide its effect.
Moreover, the buoy has the drawback of working to disadvantage the anchor in terms of angle of pull: unlike a kellet, the rode is initially kept at a higher angle. However, as with a kellet, by the time the anchor is likely to be troubled, the rode will be close to tight regardless, so the impact on angle of pull and consequent ultimate performance will probably be negligible.
An additional use for such a buoy is an attempt to keep the rode clear of tall obstacles on the sea-bed, such as coral or dead trees. In this case the buoy can be placed at a point on the rode a distance away from the anchor a little greater than the depth of water. In light weather the buoy will then keep this final section of the rode raised, but will sink to allow the rode to provide scope when necessary.
To improve the ultimate performance of an anchor system, several other alternatives to kellets are recommended. The first and most obvious one is an upgrade of the anchor. If it is an old generation type, a new generation design will increase its efficiency on a weight-for-weight basis considerably. Also, performance is almost directly related to size. If you have a 10 kg anchor and want 50% more performance, a 15 kg anchor of the same type will provide just that. If you consider the cost and weight of a kellet compared to the same cost and/or weight instead added to the anchor, an upgraded anchor is the obvious choice.
If you need to minimize your swinging circle, again, use a larger anchor – and make use of the reduced scope this permits.
Lastly, shock absorption should be provided by rope or a dedicated snubber. This allows far more control and reliability compared to the quasi spring of a kellet.