I’ve been doing small experiments on my balance to learn what affects it. Most research using new tools follows a progression. Step 1: you learn what people already knew. Step 2: you find new information that isn’t very interesting. Step 3: you find interesting new information. Earlier I found that I could balance on one foot longer on a wider platform — Step 1.
Now Step 2. I’ve done a few experiments comparing different footwear (sandals, shoes, barefoot). In each experiment I ran several conditions, each consisting of 12 trials standing on my left foot followed by 12 trials standing on my right foot. These trials had gaps of seconds between them. Different conditions (different footwear) were separated by at least 10 minutes and usually more.
The right-foot average was always more than the left-foot average. You can see examples of this in my earlier results. I doubt that the right foot/leg is actually better than the left so this suggests there is a substantial warmup effect, as there is in most tasks.
To make measurements more precise, it would help to have a warmup period before collecting the main, more stable data. How long should it be? The graph below shows data from many of the conditions I have run arranged by trial number, with a lowess summary line. 
The y axis is in log seconds, not seconds; I used a log transform to make the distribution of the data more symmetrical. The maximum time was 30 seconds. (Log(30) = 3.4.) If I kept my balance for 30 seconds, I stopped, and recorded the result as 30 seconds.
The graph shows an early warmup period that lasts 6-8 trials long, followed by a slow improvement that lasts at least 24 trials. Here is something new and not very interesting: details about the warmup effect.