Formula to decide reliability of data

[Twinbee]

If say I was looking for say... a good webhost, and I wanted to search Google for how many positive to negative comments there were, I would not only look for the good/bad ratio, but also how many comments there were in total to express reliability.

Which formula best expresses this extra factor out of the following three:

1: (comments - comments^0.5) / comments
2: comments / (comments^0.5 + comments)
3: (comments/30) / (1 + comments/30)

(30 is chosen arbitrarily in the 3rd example). For each formula, a result of 1 is total reliability, and 0 is total unreliability, but other than that, they differ with the values in the middling range 0 to 1.

The end formula for deciding the best webhost would be as follows:
(good/bad) ^ ((comments - comments^0.5) / comments)
...or...
(good/bad) ^ (comments / (comments^0.5 + comments))
...or...
(good/bad) ^ ((comments/30) / (1 + comments/30))

Which is the more 'genuine' formula to use, or maybe another is preferred?

 

[Valkarie]

It is difficult to definitively answer this question without knowing more about how you are defining the "good/bad" ratio, and what kind of information is being collected from the comments. Each formula might be more appropriate for different types of data. If possible, it is best to experiment with all three formulas and compare the results to determine which one will provide the most accurate and reliable results.