A data-driven look at Cruz’s (bot?) army

For the last several days, we’ve been promising a look under the hood at Sen. Ted Cruz’s fairly astonishing Twitter performance, using the Pollchatter.org data.

As previously noted, the numbers on his Tweets have been very consistently higher than any other district in the nation, usually by quite a substantial amount. Much of this is clearly due to a concerted Twitter campaign, with easy-to-retweet “Click to Tweet” buttons on videos that allow supporters to Tweet a pre-rolled message of support (we’ll address that lovely addition to our democratic discourse later).

However, the numbers have been high enough to raise eyebrows. Today, Bayou Brief publisher Lamar White caused a Twitter storm of his own, noting the suspiciously high number of Cruz supporters apparently Tweeting during peak Russian working hours.

White’s suspicion dovetails with ours. Cruz is either running an extremely exceptional Twitter campaign, backed by a massive cadre of supporters who are spending an inordinate amount of their time tweeting for him. Or something else is going on. Or maybe both.

Today we’ll just do a first cut at this question, looking at the same issue White noticed.

RED STAR AWARD TO: @GUARDIAN_ELITE

Our first check was to see if there was an appreciable number of posters Tweeting at precisely the same time. This search didn’t immediately indicate anything amiss. In the last two weeks, there was one one-second block when six people Tweeted at the same time, nine different blocks when five people tweeted at the same second, and lots more for four and three simultaneous tweets.

+---------------------+------------------------+
| created_at_dt       | count(p.created_at_dt) |
+---------------------+------------------------+
| 2018-09-10 15:51:59 |                      6 |
| 2018-09-04 21:08:40 |                      5 |
| 2018-09-10 21:50:28 |                      5 |
| 2018-08-27 16:51:08 |                      5 |
| 2018-08-27 18:31:35 |                      5 |
| 2018-09-04 18:12:31 |                      5 |
| 2018-09-07 14:38:17 |                      5 |
| 2018-09-13 22:04:00 |                      5 |
| 2018-09-10 23:34:15 |                      5 |
| 2018-09-04 21:03:20 |                      5 |
| 2018-09-05 22:37:00 |                      4 |
| 2018-09-10 15:23:29 |                      4 |
| 2018-09-04 21:05:20 |                      4 |
| 2018-09-04 21:07:54 |                      4 |
| 2018-08-27 15:24:42 |                      4 |
| 2018-08-31 16:48:13 |                      4 |

This isn’t out of the realm of plausibility. If tens of thousands of people are tweeting on a subject every day, there’s a reasonable chance (I’ll do the actual stats later) that some will hit the button at the same time.

HOWEVER! A closer look at the data turned up something interesting: That most popular time block on 9/10 saw a user called @Guardian_Elite post twice at precisely the same second.

+---------------------+-----------------+
| post_id             | user_scrname    |
+---------------------+-----------------+
| 1039179689057824770 | ggary59         |
| 1039179689149923329 | Guardian_Elite  |
| 1039179689322004480 | Guardian_Elite  |
| 1039179690022338560 | McKenzieHarty   |
| 1039179690257276928 | Marie10778357   |
| 1039179690559381506 | Leftyissickness |
+---------------------+-----------------+

Both of these posts were retweets of the same Cruz tweet:

I’m not sure how this could be possible without automation. Unless maybe two computers running at the same time, one hand on each ….and…. go! It’s completely possible.

@Guardian_Elite, for what passes as the record, claims to be a SoCal woman, who just happens to take an intense interest in Ted Cruz. However, we would happily honor her with our first provisional Red Star Award.

TIMING OF CRUZ TWEETS

We also ran a report examining when exactly people were tweeting and mentioning Cruz’s district, TXSen.

The chart below shows a frequency distribution of posts in two-hour blocks, beginning at midnight UTC time 9/12, and going through 11:59 pm 9/13. So, the 24-hour period beginning at 7pm 9/11 Houston time (or 9/12, 3 am Moscow time). (UTC time is Greenwich time without daylight savings).

The first peak is at 9am – 11 am Houston time (5pm – 7 pm Moscow time). The very large standalone peak is 9 pm– 11 pm Houston (5 am to 7 am Moscow). And the other, larger curve on the right tops out at 1pm – 3 pm Houston time (9pm – 11 pm Moscow).

What does this show us? Nothing definite, other than that there was a huge surge on the night of the 12th, and a stronger sustained push during working hours on the 13th.

Is this strange? We can’t say yet. It’s not strange to see an evening-hours surge. People also post heavily from work, sometimes. To draw real conclusions, we’ll have to compare this chart data to other districts, and to other days, to find a baseline (stay tuned for that later).

However, if there is bot manipulation happening, it’s very likely that the bot operators are clever enough to be posting largely within plausible hours for US voters. And, as the Bayou Brief‘s Smith noted in his article, a large number of Cruz’s supporters are being tracked by the Bot Sentinel bot-spotting service.

So: Too early to say anything with certainty. But a bit of data helps clear the air.