Oversampling doesn't solve your real problems.

Oversampling can be used for a number of reasons. To move the break frequency for a low-pass filter way outside the frequency range of the measured signal, so a low-order filter with good phase behaviour can be used.

It can also be used to reduce problems with individual steps in the ADC not being at their theoretical positions, i.e. you sample many times, while adding a known random offset to the signal just to get the sampling to happen at different locations on the ADC curve. It can both increase the resolution of the sampling process and solve some of the linearity problems with the ADC.

But there are even 16-bit or 20-bit or 24-bit ADC. It's far harder to get 0.1 degree [u]precision[/u] from the sensor than it is to get 0.1 degree resolution from the electronics that samples the sensor.

Matching a Fluke? What Fluke. A "cheap" Fluke isn't really cheap. And it will not manage any 0.1C and be correct about it. And a tremendously expensive Fluke can't be matched with a cheap hobby design.

What do the customer [u]really[/u] need? Something that really gives 0.1C correct readings? Something that fools him into thinking he has that precision? Something else?