How reliable are lie detector tests?
I am agnostic on this question.
There is little scientific evidence to support the reliability of polygraphs. Despite claims of 90% – 95% reliability, critics charge that rather than a "test", the method amounts to an inherently unstandardizable interrogation technique whose accuracy cannot be established. A 1997 survey of 421 psychologists estimated the test’s average accuracy at about 61%, a little better than chance. Critics also argue that even given high estimates of the polygraph’s accuracy a significant number of subjects (e.g. 10% given a 90% accuracy) will appear to be lying, and would unfairly suffer the consequences of "failing" the polygraph. In the 1998 Supreme Court case, United States v. Scheffer, the majority stated that “There is simply no consensus that polygraph evidence is reliable” and “Unlike other expert witnesses who testify about factual matters outside the jurors’ knowledge, such as the analysis of fingerprints, ballistics, or DNA found at a crime scene, a polygraph expert can supply the jury only with another opinion…”. Also, in 2005 the 11th Circuit Court of Appeals stated that “polygraphy did not enjoy general acceptance from the scientific community”. In 2001 William G. Iacono, Distinguished McKnight University Professor of Psychology and Neuroscience and Director, Clinical Science and Psychopathology Research Training Program at the University of Minnesota, published a paper titled “Forensic “Lie Detection": Procedures Without Scientific Basis” in the peer reviewed Journal of Forensic Psychology Practice. He concluded that
Although the CQT [Control Question Test] may be useful as an investigative aid and tool to induce confessions, it does not pass muster as a scientifically credible test. CQT theory is based on naive, implausible assumptions indicating (a) that it is biased against innocent individuals and (b) that it can be beaten simply by artificially augmenting responses to control questions. Although it is not possible to adequately assess the error rate of the CQT, both of these conclusions are supported by published research findings in the best social science journals (Honts et al., 1994; Horvath, 1977; Kleinmuntz & Szucko, 1984; Patrick & Iacono, 1991). Although defense attorneys often attempt to have the results of friendly CQTs admitted as evidence in court, there is no evidence supporting their validity and ample reason to doubt it. Members of scientific organizations who have the requisite background to evaluate the CQT are overwhelmingly skeptical of the claims made by polygraph proponents. 
Polygraph tests have also been criticized for failing to trap known spies such as double-agent Aldrich Ames, who passed two polygraph tests while spying for the Soviet Union. Other spies who passed the polygraph include Karl Koecher, Ana Belen Montes, and Leandro Aragoncillo. Pseudoscience debunker Bob Park said, "The polygraph, in fact, has ruined careers, but never uncovered a single spy." Polygraph examination and background checks also failed to detect Nada Nadim Prouty, who was not a spy but was convicted for improperly obtaining US citizenship and using it to obtain a restricted position at the FBI.