One of the most common, and least credible, complaints I hear from deniers is that climate change is not real because the models all fail. This is both false and a false argument.
Let me address the false argument first. In fact, I have addressed this in some detail before, but it's worth repeating. A model is a mathematical representation of a real phenomenon. The real thing remains unchanged whether the model is accurate or not. If we get the model correct, the real world is the same. If we get it wrong, the real is still the same. The accuracy of models has no bearing at all on the reality of climate change.
And, climate models and climate science are not the same thing. Models are merely one tool among many we use. Besides models, there are satellites, thermometers, aircraft, ice cores, mud cores, coral cores, tree rings, buoys, radar and much more.
Besides that, the fact is, models are quite accurate and getting better every year. Here is just one discussion on the topic.
I came across a paper from last December that applied to this discussion: Evaluating Modes of Variability in Climate Models, Adam S. Phillips, Clara Deser and John Fasullo, Eos, Vol. 95, No. 49, 9 December 2014. This paper describes how the National Center for Atmospheric Research (NCAR) has developed a Climate Variability Diagnostic Package (CVDP). This package uses metrics for climate variability and applies these metrics to model outputs to see how well they perform in an effort to evaluate internal variability of the models. You can go to the NCAR website and see some of the results. What they show is the models are doing pretty well. Here is an example:
This is an evaluation of 42 separate outputs for models evaluating the December-January-February sea surface temperature change for the period of 1900 - 2005. Red shows the areas where the models were the least accurate and blue the areas where they were most accurate.
As you can see, they did very well. There are numerous other examples on the website.
No, this will not get any denier to stop saying models aren't accurate, but it is one more clear example of how they simply ignore the reality.