You want to do a "dependent (or paired) t-test". You will run two dependent t-tests, one on each treatment group (water, placebo). Your data should be the actual running times, not just a 1 or 0 for improved or didn't. Note that the test will be most accurate if you have a relatively homogenous set of test subjects, for example, all 14 year old girls, rather than a mix of ages and genders.
Here is a link to a calculator you might like:http://www.graphpad.com/quickcalcs/ttest1.cfm
Run this test twice, for the first one, group 1 will be base times and group 2 will be the same subjects after they had water. Make sure the times on each row are for the same subject, as these are paired to see by how much the time changed. For the second one, the group 1 column will be base times and group two will be the same subjects after they had the placebo treatment.
Depending on how much the runners times changed, you may find that the improvement in running time is significant in both groups (e.g. that both water and the placebo results have a p-value of less than .05). In this case, you could look at the actual t-values to see if one result is stronger than the other.