5 ms·
If I've understood correctly, this relates to the German Tank Problem: using statistical analysis to estimate a count. With German Tanks, you have limited data
by ronaldx 13y ago
If I've understood correctly, this relates to the German Tank Problem: using statistical analysis to estimate a count.
With German Tanks, you have limited data available to create an estimate from (a few serial numbers, in the canonical case). But in this case, you deliberately limit the data stored in order to keep it at a manageable level: hash to randomise and then store only summary stats.
Pretty cool. Actually very cool, I really like it.
However, are there any interesting use cases for this? It seems to me that it's not particularly useful until your list is long enough to make search/sort/storage difficult: that's rare even at web-scale, no?
- antirez 13y agoHello, the use case is that you can count what you usually don't store (for example different IPs connecting every day to your server), or you can count what is usually stored but it is too expensive to count otherwise (Like iterating a database table and counting unique items in a given column). Also since it is so cheap, you start counting even things that could be unreasonable to count otherwise but that may provide useful insight, for example unique users accessing the site for every hour of the day, unique words ever used by a given user in her comments, and so forth.