What is the point of a wheatstone bridge? Everywhere, like here says that it's used for measuring an unknown resistance, but I don't understand why it's so complicated.
Can't you just use a voltage divider to measure resistance? The equation for that is \$V_{out} = V_{in}R_2/(R_1+R_2)\$ , rearranged to make \$R_2 = V_{out}R_1/(V_{in}-V_{out})\$.
Yes, you can use a half-bridge and measure the excitation voltage and output voltage. It's simpler and may be acceptable in many cases, especially with (sometimes very expensive) modern ADCs that effectively form the other half of the bridge.
Wheatstone bridges are from a time when it was easier to get precise resistors than precise well-regulated voltages. A poorly regulated voltage (for example, from a battery) could be used to excite the bridge and then the output voltage could be nulled out by adjusting another resistor value, perhaps observed on a galvanometer or a sensitive mirror galvo. Since you're only looking for a null, only the resistor needs to be precisely known.
In more modern use, it's useful in situations where we are after a change in resistance from a certain value. For example, say a certain Pt100 platinum RTD over the range 10°C to 30°C will change from 103.903Ω to 111.673Ω. By offsetting the lower value with a precise and stable resistor we can amplify and measure the difference and reduce the dependency on our reference and measurement circuit by a factor more than 13:1. That kind of improvement in a measurement circuit is not trivial at all.
On the precision side, you can easily buy a resistor off the shelf with a stability of +/-0.2ppm/K for the price of a couple cups of coffee. Without resorting to temperature control, the best monolithic voltage references are 4-5x worse.
The Wheatstone bridge has many advantages at little complexity/cost, as mentioned in the others answers and comments.
To add some grey in all the black & white, let me illustrate an example where the Wheatstone bridge actually offers little advantage over a properly designed half bridge measurement:
simulate this circuit – Schematic created using CircuitLab
This is a multiplexed half-bridge measurement, where the DUT (R2) is interleaved with a single reference Rref. That way, the ratio between the measured resistances R2 and Rref directly corresponds to the exact ratio of R2 and the known Rref. Importantly, this result will be independent from any drift from V1 or even R1 or VM1. It also doesn't require good matching of Rref and R2.
These advantages are greater than those of the Wheatstone bridge, but the drawback is that the VM1 sees a much larger DC voltage compared to the compensated bridge voltmeter and it is much harder to achieve low noise here (see Spehro's answer).
Still this situation is acceptable, if the Voltmeter is very good, which isn't so hard today. Indeed it can be easier than to maintain accurate bridge matching.
We use this scheme for measurements of resistor noise index. It is limited only by the noise index of Rref, which can be made arbitrarily low by increasing its volume. In contrast, a Wheatstone bridge will diminish the source's 1/f noise only by its own matching ratio (e.g. -60 dB) but not entirely eliminate it. And to complicate matters more, the matching ratio of a Wheatstone bridge is itself subject to drift.
