None of the countless species of animal in existence has three
legs. Creatures such as the kangaroo and the meerkat use their tails for
balance, but a tail is plainly not the same as a leg. This pattern does
not apply only to mammals - other kinds of animal have an even number
of legs, too. Why wouldn't having three legs work?
The answer probably lies in how legs are form, that is from segments. Most animals come from a basic design that consists of segments with one pair of limbs per segment (think of a centipede), this basic design goes right back and is fundamental to vertebrates and so you only really get pairs of legs hence you don't get three.
To look for odd numbers you have to go to the inverts and here it comes down to symmetry and segmentation. For example many starfish have five arms but they are radial symmetrical rather than bilaterally symmetrical and so they body segments are arranged around a central axis not a longitudinal axis. Five is a neat number as it ensures you don't have any obvious points of division - that is why older cars have five wheel nuts on each wheel - to ensure that opposite each nut is solid wheel and not another nut so there is less chance of a fracture line going through two or more nuts. So you can get odd numbers of legs but only in radial symmetrical organisms and most of those are invertebrates. Maybe three is too few legs to be useful in a radial design.
An animal, if ever had three legs, would find difficulty to balance itself while changing its position (moving), that is why God created all the living oranisms with either no legs (snake), or two legs(human) or four legs (horse, dog etc).
Ah, God - on this page he doesn't always get the credit due to him. I don't know why a 3-legged beast would be less stable than a 2-legged one. Even 4-legged creatures usually move 2 legs at a time. Surely a 3-legged animal could move one at a time?
In evolutionary terms arms are just modified legs, of course, so we have a pair of each as ScottB explains above. An arm is a leg with modifications so let's call it a limb. An elephant's trunk is a nose with modifications and it's certainly versatile enough to be classed as a limb. A spider monkey's tail is pretty impressively limb-like too.
Small oversight in leg counts: try 6 (insects), 8 (octopi, spiders and scorpions), 10 (squid), around 100 (centipede), around 300 (millipede).
Humans derive from 4-legged animals but have specialised for the hunter/gatherer role: partly by having big flat feet to put your CoG over; partly by standing in a vulnerable upright position to increase your angular momentum around your feet to make it easier to balance; partly by devoting about half the body height to huge legs and muscles to hold it all up; partly by devoting a fair amount of head space to sensors (in the ears) and brain power (to control balance). None of that comes for free.
Four-legged animals tend to balance on two diagonal feet while moving the other two in a trot (except for specialists like horses who have learned the gallop). That is more stable and quicker than moving one of an odd number of legs. Six-legged insects tend to move on three triangles alternately (e.g. move right-front, left-middle, and right-back legs at once).
Five feet (starfish) are still balanced with one leg up, whereas three feet in a symmetrical mode just flop on their belly with one leg up. That's not good.
Three feet in the Rolf Harris (three-in-a-row) leg layout requires that the central foot has to support the whole body weight at some point in the gait, which is even worse than two feet.
Symmetrical bodies are well favoured for efficiency in DNA, with a program to make half a body, and a mechanism to make a mirror image. This is similar to the 'segmented evolution' point made earlier. So even numbers of limbs are favoured. Hands and feet have identical bone structure because they start the same topologically in the embryo, and morph into the right shapes later.
The efficiency of DNA is further improved by the junk DNA segments: like all great programmers, God uses these for comments, so that future enhancements can be carried out quicker and more reliably.
It isn't true that there are no animals with an odd number of legs - starfish have five. At least, they really walk on lots of little tube feet, but the tube feet are arranged in groups underneath the five visible limbs.
Three limbs might actually work for something like a starfish which is basically lying down, but three is an unstable number for any animal which stands up off the floor, because as soon as it lifts one foot to go forwards it is standing on two limbs, and may topple.
Advanced quadrupeds such as dogs and cats can manage very well on three legs, because they have a good sense of balance and because their feet naturally fall close to the midline of their centre of gravity. Even though as soon as they raise a paw they are left teetering on two limbs, they only have to sway very slightly to get their weight balanced over the remaining two feet - and in fact at a fast gait quadrupeds often pick two feet off the floor anyway and balance on the other two.
Then there are bipeds such as humans or birds, who go on two feet all the time, and have to balance on one foot as they step forwards with the other. But again, such bipeds have a good sense of balance and their feet positioned well under their centre of gravity; and they evolved from quadrupeds.
However, animals which were first developing the ability to walk on land - amphibians, early arthropods etc. - tended to be low-slung with their feet placed well out to the side, and they were only just learning to balance on legs. A three-legged crab or newt would either fall over as soon as it lifted one foot off the floor, or have to sway violently to get its weight over the remaining two feet, thus wasting energy.
So emergent walkers needed four or more limbs. All land-walkers seem to have gone for an even number of limbs arranged in pairs, probably because developmentally-speaking it's nice and easy to code for one pair of legs on one body-segment, and then just repeat the same bit of code as many times as you need it. And if you've got a string of legs down either side, then your other body parts get arranged to fit, with a mouth at one end and an anus or cloaca at the other.
Re-shaping themselves to have only three limbs would thereafter be very difficult, since their body had adapted to hang its organs in the gulley between two rows of limbs, and there would be no clear evolutionary benefit to evolving from a multi-paired to a three-limbed form. There's probably nothing that three legs could do that couldn't be done better by either four or two, since a triped would have neither the fast simple gait of a biped, nor the stability of a quadruped.
The only way I could see that you would get a three-legged walker would be if you started with a thick-tailed biped such as a kangaroo, and the tail then gradually evolved into a sort of extra limb. Aquatic animals, for whom falling over isn't an issue, do sometimes end up effectively with three limbs. Whales and dolphins, for example, use their flippers and tail as their three swimming limbs and so do many fish (although some also use their pelvic or dorsal fins as well). There are some fish which actually walk on the sea-bed using their pectoral fins and tail as a tripod - although those who emerge and do so on land are unable to walk with three limbs without falling over, and either progress in awkward hops or lie on their bellies and row, so that they have nowhere further to fall.
If you are going to count starfish "arms" in the leg debate...
What about tripod fish (search for them for pics/vids)? Okay so they are modified fins and they mostly just stand on them but there are definitely 3. Give then a few million years more and we may even get there.
Just as a matter of interest, flies which have six legs walk by moving three legs at at time. This means that have always a stable footing, a tripod being extremely stable even on an uneven surface, but produces the characteristic zig-zag movement observed.