¹⁹(Browne et al., 2012; Chaslot et al., 2008). Monte Carlo Tree Search does include clever tricks which make the search a bit more intelligent. It does not try out actions (or moves in a game) completely randomly, but gathers data on which actions look more promising. In particular, there is quite a lot of data regarding actions taken in the first steps of the search path, since any search has to always try out one of those, and their number is limited because there has not yet been a combinatorial explosion as in the number of long paths. Monte Carlo Tree Search uses such data to bias the search towards paths whose initial parts have been found the most promising.