Every time an "update" to the dataset occurs (which is frequently) recent years always get warmer and earlier years in the record always get cooler. What are the odds?
Year Old Temp New Temp Missing Heat Found
1998 .52 .52 +.00
2001 .40 .43 +.03
2002 .46 .49 +.03
2003 .46 .49 +.03
2004 .43 .44 +.01
2005 .47 .53 +.06
2006 .43 .49 +.06
2007 .40 .48 +.08
2009 .44 .49 +.05
2010 .50 .53 +.03
And of course, the uncertainty is usually around twice the actual temperature.
Source: HadCrut
