This is a bad article:

• Goes into almost no detail on how any of the algorithms work
• Large match window is not the only notable difference between deflate and LZMA
• If you care about compression time then it’s very likely that you care because you care about total round trip time (compression + transmission + decomp time), so that is what you should be measuring (there are some funny cases where compression speed matters because you’re actually CPU bound - like MMO servers with 10k clients)
• No way to download their test data. Different compressors are strong on different kinds of data, and I imagine they have a funny dataset because most real world data doesn’t compress as well as whatever they’re using
• Snappy is obsolete and should never be used
• I expect bzip2 is obsolete and should never be used
• Deflate is obsolete, it is not “a reasonable \"middle of the road\” algorithm or implementation that reaches reasonable compression at reasonable CPU utilization", the only reason to use it is because everyone has a decoder for it

If you want to see some serious compressor comparisons, some good places to start are:

• Matt Mahoney’s large text compression benchmark
• lzbench
• Charles Bloom’s blog
• This GDC talk linked to from the top of Charles Bloom’s blog