In The Black Swan Theory: The Impact of the Highly Improbable, the economist/philosopher Nassim Nicholas Taleb effectively argues against the (belief in) the predictive power of statistics to predict future developments. His argument can be summed up as this:

Statistical models can predict quantative developments, until something happens the model cannot account for. Which is inevitable.

Taleb uses the thought experiment of the turkey who, over a long period of time, takes count of when he is being fed. The turkey is fed every day, so based on the statistics he should be fed tomorrow. Unbeknownst to the turkey, the day after tomorrow is Thanksgiving. The turkey is not fed, but slaughtered. The model could not account for this.

Likewise with similar experiments in regard to war. Based on current experiment, you might be able to develop an intricate model to predict casualties in a war. The model will be useless the moment humans develop a new, more effective type of weapon.

For example, Japanese and American planners did make estimates how many casualties a hypothetical conventional invasion of the Japanese Home Isles would claim. For all we know, those estimates were based on solid statistical data and reasoning. But lo and behold, in the summer months of May 1945, what Taleb calls the Black Swan is introduced: the atom bomb.

A weapon that changed the playing field, and immediately turned all previous knowledge obsolete.

A somewhat similar example would be the common expection prior to the outbreak of the Second World War that gas warfare would be employed, much like it had in the First World War. Advances in this field had resulted in even deadlier gasses. Yet, gass warfare was never employed on a scale similar to the First World War. Instead, they got the even deadlier and destructive terror bombing campaigns. Another Black Swan.

Another difficulty in using historical data for predictions is that you can't account for the unicity of individuals involved in social matters. That is why historians are never in the Prediction Game. The belligerent countries of the Second World War were led by, one might say, exceptionally brutal leaders. The 'data' garnered by the Second World War can thus not be easily used to predict current events. Contemporary world leaders might not be as ruthless, and thus work to limit casualties in their military operations.