http://www.dfrobot.com/wiki/index.php?title=Capacitive_Soil_Moisture_Sensor_SKU:SEN0193
roryaronson personally I believe you have gone completely with the wrong sensor.
This is a person building own.
Capactive sensors give broader reading than two probe in ground.
http://www.mait.com.au/products/aquaflex-soil-moisture-sensors-soil-moisture-probes/
Note here capacitive does not have to come in contact with soil at all to get a reading.
http://www.coppersulfatecrystals.com/killroots.html
Putting copper in and out of the ground is not something you want to be doing. Higher grade sensors resistive sensors like I guess you are using are gold plated but that plate does wear off then you come back to copper contamination reacts with sulphates in soil to make copper sulphate that plants don’t particularly like. The issue you have is soil dry is like very light sand paper. So probe only been in the ground for a few min day is only fix one problem the fact its being progressively sanded off into the soil is another. The gold plated resistive base sensors are designed to be inserted and left not put in and pulled out over and over again.
Commerical farm usage is capacitive sensors were possible. 1 they last 2 you don’t end up with contamination. Putting fixed pipes in and running capacitive sensors inside that is done in fact to reduce wear and read to a lot deeper depths.
Something as simple a lettuce is 6 to 21 inches of root depth depending on soil compaction. Good garden you expecting at least 17 inchs. Current design you only measuring the top bit as well.
10% difference in readings between night/day or summer/winter<< There is something that can cause greater than this.
Above is a kids experiment but shows key properly of Capillary Action. Hotter the temperature of the soil the more the water will rise up though the soil and the slower water with sink into soil and as the soil cools the slower the water will rise though the soil more the water will sink into the soil until you get to freezing. So yes water in soil does move up and down between day and night and the rate water sinks in changes between summer and winter
Taking surface readings can lead you up the garden path thinking you need to water when the plant has tones of water. Why the water has sunk in deeper than where you sensor is reading but not past were the plants roots are getting at it. Also it can trick you the other way were the surface is wet and the subsoil is dry so the water disappears very quickly even that you appeared to have had quite a good reading.
Here is what is really going to ruin you day is taking a soil surface temperature will not define how fast water is sinking into the lower layers. You will need subsurface readings. Reason why I use wicking garden beds is it make the water layer come up from under the plants. This creates a supported stack of water. Watering from above it can basically can have no moisture under so it sinks into the ground away from the plants. A supported stack of water does not move up or down as much when the temperature of soil changes. So the overall design of the garden bed has huge effects on how water will behave in it.
Commercial farms don’t go to all the effort of taking reading down through the complete soil profile for no good reason if you don’t you are only getting information that will make you make mistakes. Yes using capacitive sensors to map the complete profile on farms you can see in the numbers the water moving up and down through the day and the amount of movement linked to temperature not surface temperature on it own either…
There is a lot of science to taking care of plants that if you are not aware of can make you make huge mistakes. Yes something that sound simple as a soil moisture reading its possible to mess method up completely because its not as simple as it sounds. Taking the surface temp of bed is not going to tell you where the sub surface temp is 3 inchs in. Temperature probe will not fix this error if it would commercial farms would be using it.