That test above is very far from conclusive, but it does appear to indicate the possibility FFD has an effect.
This I don't agree with. These tests don't pass the bar for sufficient statistical rigor to demonstrate any significant effect. Ground combat casualties can vary quite a lot from round to round, I have seen variance on the scale of +100% from one round to the next on occasions, so a net +8% advantage after ten rounds is minor by comparison to the variance.
We can make a rough estimate of the variance here by assuming that (1) the kill counts are uncorrelated, allowing us to use a simple Poisson distribution, and (2) the number of kills in each round is the same, which almost certainly
underestimates the uncertainty but it is the best we can do with this data. For a total kill count K, the uncertainty is then Utot = sqrt(K), or a relative uncertainty of 100% / sqrt(K).
Results with uncertainty:
| 10 Rounds: | Blue | Red |
| FFD Yes: | 910 ± 9.3 | 833 ± 13 |
| FFD No: | 896 ± 10 | 899 ± 10 |
Blue kills are 167±13 with FFD and 101±10 without FFD. Running these values through a
Student's T-Test gives a t value of 4.02, which since we only have one data point for each test places us at less than 85% confidence of a meaningful result according to a
t-test table. This means there is around a 1-in-6 chance that the observation is spurious and does not indicate a real effect by any commonly accepted standards.
I want to emphasize that the second assumption we made above
lowers our estimated uncertainty. In reality, the effect of one statistical outlier combat round could drastically shape these results, so even claiming 80-85% confidence here is probably a big overstatement.
My recommendations here if we want to show a real effect (that goes against current understanding, so we need a convincing proof here) would be:
- Track the kills per round to obtain a correct total uncertainty and account for large outlier values.
- Run multiple tests of this sort and obtain correct statistical averages and uncertainties for the aggregated data.
Also, there's this post of Steve's that appears to have been conclusively drubbed over in thrashing out what FFD actually does: http://aurora2.pentarch.org/index.php?topic=8495.msg105824#msg105824
which is to say that was written, but what we got was different from the implications there regarding FFD.
That is correct. A lot of the changes listed in the "Change Log" were development decisions that Steve may have changed during the C# development process and forgotten to go back and edit the posts. This is one of those cases but far from the only example.
See, for example,
this dev post on ground combat which describes the supporting bombardment mechanics. FFD is not mentioned, and if you check this screenshot there are no FFD elements in the formation heirarchy. Note that this post is about 8-9 months after the one you linked, since it is more recent it is more likely to represent the final state of C# mechanics.
Given that, how do people like to use FFD, if at all?
FFD is more or less only useful for front-line formations, to the best of my knowledge, due to
the targeting rules for FFD being the same as for MBL/HB components. Bombardment components
fire with the front-line unit they are supporting - i.e., they will not fire in support of a unit which is already firing in support of a unit (there is only one level of supporting fire). Therefore, I only put FFD in my front-line formations, usually 2 units for smaller (~5,000 ton) formations and 4-6 units for larger (15,000-20,000 tons) formations. I like to have at least 2 even in smaller formations for redundancy, so the orbital supporting fire is not lost if one of the FFD units is destroyed.
