A book that presents difficulties in isolating relevant signals from the noise. Drawing on his experience in fields as diverse as politics, sports betting, poker, chess or purse, the author shows the difficulties of distinguishing reality of a signal. Personally I liked the tone quite different works on Big Data that promise almost absolute lighting models. Nate SIlver shows the need to mix data and human analysis, some problems are not modeled and logically enough it is necessary to measure the relevance of predictions. On this last point it shows that the human leads to biases ... to be heard when one is unknown ..he atypical predictions do .. and when we reached the fame, it is more urgent in trend. The author sets forth an interesting theorem in economics .... if a forecaster had flair .... so that he will have great difficulty in having a second time. The author makes an apology Bayesian techniques ... and calls out the educational framework from Mr. Fisher, and on this point I can only join him ... in the world of Big Data are found .. correlations between data ... which does not mean causality: it's not because sales of ice are correlated with forest fires must be banned Miko ice.