Currently, I create my samples through GetMatchHistory and GetMatchDetails. This is because I consider the skill bracket information very important, and these are the only API calls that give me access to skill data. Unfortunately, this also restricts me to only getting 500 matches per day. This can be done either on a general or hero specific basis. I use hero specific grabs for skill information. General for everything else.
Sample sizes should usually be listed. For character examinations I’ve decided 5,000 is usually acceptable. For general examinations I prefer to make them as large as possible, but I am limited to 500 per day or 15,000 per month, give or take. All samples will be exclusively created from a specific major patch period(i.e. 6.74/6.75/6.76c/6.77).
The biggest single weakness is that I cannot vary the time of day used in the sample creation. This might create a regional bias, but for now I have no other options. It’s also difficult to estimate how much of a regional bias there is since I don’t know the baseline numbers for games played per region, but from what I’ve seen the majority of matches I look at come from Europe West, US East, and US West in that order.