Proxy records and climate models are used to understand past climates, but comparing them can be tricky. When the error in the proxy is large, it can make the model look better than it actually is. By looking at how uncertain the proxy is compared to the model, we found that if the proxy's uncertainty is more than three times the model's variability, the agreement between them is poor. This means that only proxies with uncertainty less than three times the model's variability should be used to evaluate the model accurately.