First look at the Twitter standard sample hose appears that the advertised 1% sample rate is actually closer to 0.4-0.5%
root@es1:/data2# jq '.created_at' samplehose | cut -c 2- | cut -c -16 | sort | uniq -c
2107 Fri Jun 08 22:14
2103 Fri Jun 08 22:15