Climate models have done a horrible job predicting recent trends, but they can't even explain the past

Predicting the future is much harder than simply running a regression that explains the past.  The whole point of a regression is to fit a line to data.  Presumably they must have used past data at some point to get their idea of what causes climate to change.  If they can't go back then and explain the changes, it probably means that their models deviated then from the results that they had previously gotten.  The whole thing is pretty troublesome.  From Watts Up with that:

One of the main points of criticism of the CO₂-dominated climate models is that they fail to reproduce the temperature fluctuations over the last 10,000 years. This surprises no one as these models assign scant climate impact to major factors, i.e. the sun. As numerous IPCC-ignored studies show, the post-Ice Age temperature curve for the most part ran synchronously with solar activity fluctuations. The obvious discrepancy between modeled theory and measured reality has been brought up time and again.
The journal Climate of the Past Discussions has published a new paper written by a team led by Gerrit Lohmann of the Alfred Wegener Institute (AWI) in Bremerhaven, Germany. The group compared geologically reconstructed ocean-temperature data over the last 6000 years to results from modeling. If the models were indeed reliable, as is often claimed, then there would be good agreement. Unfortunately in Lohmann’s case, agreement was non-existent.
Lohmann et al plotted the geologically reconstructed temperatures and compared them to modeled temperature curves from the ECHO-G Model. What did they find? The modeled trends underestimated the geologically reconstructed temperature trend by a factor of two to five. Other scientists have come up with similar results (e.g. Lorenz et al. 2006, Brewer et al. 2007, Schneider et al. 2010). . . .