We have two tracks in this competition with different cost function.
For this track of the competition, your hypothesis g should provide the probability estimate that whether the user will click the advertisement. Then your classifier g for an example (x, y) will be evaluated by log loss: $$L\mathrm{og}Loss=-\frac1N\sum_{i=1}^N\lbrack y_i\log\left(g\left({\boldsymbol x}_i\right)\right)+\left(1-y_i\right)\log\left(1-g\left({\boldsymbol x}_i\right)\right)\rbrack$$ The result will be larger than zero, the lower the better.
For this track of the competition, your hypothesis g should provide the binary (0 or 1) prediction on whether the user will click the advertisement. Then your classifier g(x) will be evaluated as F1 score: $$\style{font-size:20px}{\begin{array}{c}F_1=2\times\frac{precision\times recall}{precision+recall}\\[12pt]\mathrm{where}\;precision=\frac{\left|\{True\;Positive\}\right|}{\left|\{True\;Positive\}\right|+\left|\{False\;Positive\}\right|}\\[5pt]\;recall=\frac{\left|\{True\;Positive\}\right|}{\left|\{True\;Positive\}\right|+\left|\{False\;Negative\}\right|}\end{array}}$$ The result will range in [0, 1], the higher the better.