Pinotage said:
The average party level is 2 (13/6), although the group should be able to handle CR 3 creatures as a standard encounter given that they are 6 characters instead of the normal 4.
The technique I use for the party level is to calculate it the same way that the EL would be calculated for any encounter, then subtract 4.
For example, a party of 4 characters, all of them ECL 6 would have an EL of 6: CR6 + CR6 = CR8, and because there are two of them, CR8 + CR8 = CR10. Minus 4 becomes EL 6.
Now add a single ECL 3 character. You now have "average party level" of 27/5, or 5.4.
Notice how the average party level WENT DOWN instead of up?! If the party gets stronger, the APL (average party level) should go up, shouldn't it?
Using the EL calculations used for opponents in an encounter, we get a better result. First, start with the lower ECL creatures first. So do the CR3 + CR6 first. Unfortunately, the formula in the DMG doesn't give an exact value for this one. If a monster with CR(X) is paired with a monster of CR(X+2), the result is supposed to be CR(X+3). This pair will be a little less powerful than that, so I'll call it CR6.8 (there's actually an exponential formula* that gives precise results, but I don't have it handy). So now we're left with CR6, CR6, CR6, and CR6.8. CR6 + CR6 = CR8, and CR6 + CR7 should be about CR8.5 (a guess based on CR6+CR6=8 and CR6+CR8=9), but we have CR6.8, not CR7, so I'll call it CR8.4 instead of CR8.5. That leaves CR8.4+CR8, or about CR10.2. Subtract 4 points, and the encounter level for the party is 6.2.
Notice how the EL calculation better approximates the level for the party? Using an average is great (and easy!) if you have exactly 4 PCs, but otherwise it breaks down. Yes, the math is a bit of a pain, but you only need to do it when the composition of the party changes.
--
* The formula recognizes that the CR ratings are on an exponential scale. So convert the CR's into logarithms, add them together, then convert back. When using the logarithms, the math is much easier: convert all CR's into logarithms, add them all together and divide by the number of PCs, then convert back. There's none of the guesstimating that I used in my example. If anyone's interesting, I'll try to find the formula...