Every project is different and there are many factors that determine what kind of device would be best fit for your purposes. These factors include for example, who are your participants? Are they small kids whose wrist might be too small for big bulky devices? Are they elderly who might not be accustomed to using smart phone or even email?
Do you need to provide participants feedback only after the intervention or already during the intervention? So, if you are doing projects which aims to behaviour change you possibly need to be able to give feedback right after the measurements. There are actually big differences related to this between different devices and this is something you might not remember to consider when focusing too much attention to specifications of the devices.
One important factor is also, whether physical activity is main outcome variable in your study, or whether you just, for example, need to exclude that changes in physical activity did not influence your results. In the latter case you might just want to have a device that just makes data collection and analysis as easy as possible.
Do you need also accurate results related to sitting and sedentary behaviour? In that case you want to consider thigh-worn device as it is basically the only wearing position that can do posture allocation accurately.
In general, wearing location is of crucial importance for any project. In my opinion it comes down to compliance and accuracy.
Many studies are using wrist-worn devices as people have been used to wearing watches, and watch-like activity tracker is really easy to wrap around the wrist. So, its easy and compliance generally good.
Then there are thigh-worn devices. Influential ProPass consortium is recommending thigh as a preferred location for accelerometers as it provides more accurate results than other wearing positions. This is straightforward as we are moving with our legs and not with our arms and in sitting vs standing thigh is changing its angle, which we can detect with thigh-worn device.