[RPG] Does the Cleric’s Channel Energy ability heal living enemies too

Necromancy is not inherently evil

Necromancy spells are only characterized by their use of negative energy; negative energy is entropic in nature but is not a moral force. The inflict line of spells, for example, aren't evil.

But...

Necromancy does have some evil spells. Thankfully, they'll helpfully be labeled [Evil] in the spell descriptions!

Spells associated with particular alignments are indicated by the chaotic, evil, good, and lawful descriptors in their spell descriptions.

Per the Cleric class description, the Cleric cannot use spells with an alignment component opposed to their own; Good Clerics can't use [Evil] spells, Evil Clerics can't use [Good] ones. When in doubt, just check the spell's description!

Further, Channel Energy does not affect non-spontaneous spellcasting. While Channel Energy determines what spells a Cleric may cast without preparing, it doesn't affect spells that the Cleric memorizes ahead of time and does not otherwise restrict their casting list, though their alignment may do so (see above).

I would like to start by noting that this question has also been discussed on the paizo messageboard, without a definite answer unfortunately.

That being said, there is no official answer to this question, and I can give only a recommendation on how to handle it, based on the fairness of the math involved.

There are three (and a half) readily available possibilities to go about this. Ideally, what you want is something that will not alter the statistics of the roll too much. The mean and standard deviation of the complete roll should ideally equal half the mean and deviation of Xd6 (see here for X=5)

• 1d3 has a mean of 2, with all values equally distributed. The mean is matched rather well, However, the deviation of (X/2)d6+1d3 is quite large.
• 1d6/2 (round down as is usual in PF), has a mean of 1.5, with 0 and 3 being half as probable as 1 or 2. Again, the mean is alright, but the deviation for X=5 is even larger in this case.
• 1d6/2 (min 1) has a mean of 10/6 = 1.667, with 1 being thrice as probable as 3.
• Xd6/2 will actually yield a probability distribution similar to the distribution of Xd6.

From a probability standpoint, the last solution seems best, and is consistent with the way +50% is calculated as well. However, it directly contradicts RAW, as you would roll 4d6/2 instead of 2d6 on level 7. If you want to stay RAW, I would suggest using 1d3 instead of 1d6/2 for a half d6, because it does not break the symmetry of the bell curve.