This research analyses the influence of the outcome in the first quarter of a basketball game on the final result. The aim is to early detect alarm states for teams and to compute the probability of win from data obtained from the first minutes of the game. To achieve this goal, several statistical models based on modeling change in partial results have been tested: autoregressive, latent growth, autoregressive latent growth and path logit. Results show it is desirable to start winning the game, because the dependence of each partial outcome from the early outcome is noticeable. Therefore, to win the first quarter is essential to win, although to a lesser extent this also depends on the home advantage and the difference of quality of the teams.