2nd last sounds good so long as there is precedent for it. Perhaps the real pattern here is do something different every year with just a months or so notice so a wait and see approach is most wise. There is an election promised for 2014 which might result in a change.<br>
<br><br>
<br><br><br><br><br><div class="gmail_quote">On 30 August 2012 04:01, Robert Elz <span dir="ltr"><<a href="mailto:kre@munnari.oz.au" target="_blank">kre@munnari.oz.au</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
Date: Wed, 29 Aug 2012 15:37:11 +0100<br>
From: Ian Abbott <<a href="mailto:abbotti@mev.co.uk" target="_blank">abbotti@mev.co.uk</a>><br>
Message-ID: <<a href="mailto:503E2917.2090101@mev.co.uk" target="_blank">503E2917.2090101@mev.co.uk</a>><br>
<div><br>
| "2nd last Sunday" seems a little obscure though. Probably just a<br>
| coincidence it fits two years running.<br>
<br>
</div>For October there have been (or will have been, after this next one)<br>
three of them, which is more like a pattern, particularly if it happens<br>
again next year. For summer time ending this will be only the<br>
second time, so basing forward estimates on that is more problematic.<br>
<br>
The "2nd last" thing isn't as weird as it sounds, though we don't yet<br>
have enough data to know if that's how they're picking dates (deliberately<br>
os just because some other factor leads to the same result).<br>
<br>
Another plausible possibility is "Sun>=19" (the last Sunday that results<br>
in a full week after the shift) (in a 31 day month of course).<br>
<br>
Assuming the basic pattern continues, we'll know which of those it is<br>
(assuming it is either) when the dates for 2015 are announced (the<br>
2014-2015 dates for the Jan date, and the 2015-2016 dates for Oct).<br>
<br>
kre<br>
</blockquote></div><br>