If the W/NW ratio is at least 1.5 x the median, the year is considered excess.
However, if the W/NW ratio is 1.25 x the median, the year is not considered
excess, so that extra 25% above the median does not figure into the excess
ratio.
All non-excess years are treated alike, whether or not their W/NW ratios exceed
the median. All of the losses from such years are non-excess, and they are
included fully in the Non-Wind/Non-excess ratio. Further, the load is applied
to all non-excess losses.
For an example, look at ACTEX B13. If you change the non-wind losses to
1,083,333 in the first year, you preserve the median of .20, and this year now
becomes a non-excess year with a W/NW ratio 1.25 x the median. The first year
is no longer considered an excess year, so now no losses contribute to the
excess ratio; the overall average drops from .245 to .128. (If you are
putting that 25% above the median into the excess bucket, then I believe you
are doing the problem wrong. The load is for "losses greater than the median"
but only for excess years.) The non-excess losses increase to 1,300,000 in the
first year, and the NW/NE ratio remains .833. The load is now 1 + (.128)
(.833) = 1.1067, which is applied to the non-excess losses in all years. Note
that the first year contributes nothing to the load, so there is no
redundancy. (Does this example make sense?)
There are still overlaps which occur in pricing. Tiller gives a good example
on page 134 of overlaps in the NCCI Retrospective Rating Plan.