I just compressed one hour of verified Twitter data using these Zstandard options:
zstd -v -22 --ultra --long TW_verified_2019-11-28-22 --zstd=wlog=30,hlog=26,clog=28,slog=26,mml=7,tlen=999,ovlog=9,lhlog=26,lblog=8,lmml=4096
Compression Ratio: 5.62% (337174007=> 18940872 bytes
Conversation
If all those options seem confusing, running "man zstd" gives great documentation on what each parameter does. Mainly they adjust window size, match length matching and hash table settings.