The four pairs that actually matter
Most metric and inch wrench sizes are nowhere near each other and cause no trouble. A handful sit close enough to go on, and those are the ones that round fasteners. Sorted by how dangerous they are:
| Pair | Exact values | Gap | What happens |
|---|---|---|---|
| 3/4 in and 19 mm | 19.050 and 19.000 mm | 0.05 mm | Effectively interchangeable. This one is genuinely fine |
| 5/8 in and 16 mm | 15.875 and 16.000 mm | 0.13 mm | Usually fine, a fraction loose on a 5/8 fastener |
| 7/16 in and 11 mm | 11.113 and 11.000 mm | 0.11 mm | 11 mm will not go onto a 7/16 fastener; 7/16 is loose on an 11 mm one |
| 9/16 in and 14 mm | 14.288 and 14.000 mm | 0.29 mm | 9/16 on a 14 mm fastener is loose enough to matter under load |
| 1/2 in and 13 mm | 12.700 and 13.000 mm | 0.30 mm | The classic. 13 mm goes onto a 1/2 in fastener and rounds it |
| 3/8 in and 10 mm | 9.525 and 10.000 mm | 0.48 mm | Nearly 5% loose. Reaches for the corners immediately |
The pattern to take away is that a gap under about 0.15 mm is unremarkable, a gap around 0.3 mm is where damage starts on anything tight, and half a millimetre is a rounded fastener waiting to happen. The tolerance field on the tool lets you set that threshold in percentage terms, which is the more honest way to think about it, since 0.3 mm on a 6 mm nut is five percent and on a 24 mm nut it is one.
Six point beats twelve point when the fit is marginal
Whether a slightly loose wrench rounds a fastener depends heavily on how the wrench contacts it. A six point socket bears near the middle of each flat. A twelve point socket, with twice as many engagement positions, necessarily contacts closer to the corners, and corners are where a fastener yields first. On a fastener that is already worn, already slightly rounded, or being turned with a slightly oversize wrench, that difference decides the outcome. The same logic explains why flare nut wrenches exist for soft brass fittings and why an open-ended wrench is the last resort on anything seized: it grips two flats instead of six, so it splays.
Hex keys have their own trap
Hex key sizes are stated across the flats of the socket, the same convention as wrenches, and the near pairs are proportionally tighter because everything is smaller. 5 mm against 3/16 inch is 5.000 against 4.763 mm — a five percent gap in a recess where there is no corner to reach for, so the key simply twists and the socket becomes a circle. 6 mm against 1/4 inch is 6.000 against 6.350, so a 1/4 inch key will not enter a 6 mm socket at all, and a 6 mm key in a 1/4 inch socket is a quarter millimetre loose in a fastener whose whole drive depends on the flats. Metric fasteners take metric keys and inch fasteners take inch keys; where a set is mixed, the sizes worth marking are those two pairs.
Ball-end keys make this worse rather than better, because the ball contacts a much smaller patch than the flat end. They exist to let you approach at an angle, not to increase grip. Break a fastener loose with the plain end and use the ball end only for the turns that follow.
Wrench size is not derivable from thread size
People ask what wrench an M10 bolt takes, and the honest answer is: measure it. Head sizes for a given thread are set by the standard the fastener was made to, and there is more than one. Metric hex bolts to older German standards commonly take 17 mm at M10, while the ISO pattern moved several sizes down, so both 16 mm and 17 mm heads are in circulation on the same thread. Imperial fasteners split the same way between finished hex, heavy hex and structural patterns. Anything sold as flange bolts, small-pattern bolts or space-saver heads deviates on purpose. Calipers across the flats take four seconds and are right every time.
Working out what the fastener actually is, rather than what turns it, is a separate measurement — the screw thread identifier takes a diameter and a pitch and ranks the candidate designations. If the job is drilling and tapping to match, the tap drill size calculator covers the hole.
Questions people ask
Can I use a 13 mm socket on a 1/2 inch bolt?
It will go on, and that is the problem. The gap is 0.30 mm, so the socket contacts near the corners of the head and concentrates the load there. On a loose fastener nothing happens. On a tight or corroded one, the corners deform and the next attempt has less to grip. If it is the only tool in reach, break it loose in the smallest increments you can and expect to inspect the head afterwards.
Which metric and inch sizes really are interchangeable?
Only 3/4 inch and 19 mm are close enough that nobody thinks about it — 19.05 against 19.00 mm, a twentieth of a millimetre. 5/8 inch and 16 mm at 0.13 mm apart are usually uneventful too. Everything else on the near-miss list is a compromise, and the direction matters: the larger of the pair goes onto the smaller fastener loosely, while the smaller does not go on at all.
Why does my socket set skip sizes?
Because fastener head sizes cluster and the gaps between clusters are not worth stocking. Metric sets typically jump from 19 to 21 or 22 mm, since 20 mm heads are uncommon in the sizes home sets target. Inch sets often skip 11/16 or 15/16 for the same reason. A set that skips a size you need is normal rather than defective, and single sockets are sold individually.
The tool says my size is not on the list at all.
It shows the three closest entries regardless, with the gap for each, so you can see where your measurement falls. A reading that sits between two standard sizes usually means the flats are worn, the head is coated, or the fastener is not a standard hex at all — external Torx, twelve-point flange and spline drives all look roughly hexagonal at a glance and none of them take a hex wrench.
Is across flats the same as across corners?
No, and mixing them up produces a reading about 15% high. Wrench and hex key sizes are always across the flats, meaning the perpendicular distance between two opposite faces. Across corners is the diagonal, which for a regular hexagon is exactly 2 divided by the square root of 3 times the flat dimension, or about 1.155 times. If your caliper reading looks like a size and a half, you measured the diagonal.