I found an error in my data analysis for a post that made notables, so I feel it's important to draw attention to the error. Last month, I analyzed delta statistics - after each Q post, how many minutes until the next Trump tweet. The final product was a scatterplot showing the frequency of deltas between 0-20 minutes, it was clear that zero-delta was unusually common (expected, but the graph made it easy to see)

>>5496921 (pb)

Unfortuately my analysis was incorrect, due to some bad data from a misunderstanding about how times were represented in the data. I've corrected my error, and also making these changes:

• Including all Q posts from qanon.pub (early posts without tripcode or Q name were not in the original) through Q3312

• Changing the question slightly - "how many minutes until the next Trump tweet" created a shadowing effect (Q posts contributing to bucket n were excluded from n+1, n+2 etc) which could create a slope due to smaller sample size at increasing offsets, so I'm asking a different question that under the null hypothesis of no correlation shouldn't produce a slope: For each offset of n minutes, how many times do we find a Q post and Trump tweet separated by n minutes?

• Extending the graph to 60 minutes, and also looking at negative deltas (Trump first) since there was interest in that direction

Method: For Q from qanon.pub in Javascript console

for (var x in posts) if (posts.hasOwnProperty(x)) console.log(posts[x].timestamp);

… then copied results off console to text file (times appear to be Unix timestamps)

For Trump: http://trumptwitterarchive.com/archive "search through all of Trump's tweets", set a date range including all Q posts, export CSV, copy-and-paste results to text file (times appear to be GMT)

Both websites seem to display times in EST year-round (UTC-5 even during summer, when EDT is UTC-4) regardless of the user's timezone, whether DST is in effect, or whether DST was in effect on the displayed calendar date. (I ran tests as we crossed into DST to see how the websites rendered times.) Trump archive shows clock times in both HTML and CSV, but different time zones in each format (EST/GMT)

I converted Trump times to Unix timestamps. Then for all Q posts, for all Trump tweets, I calculated delta in minutes = floor((Trump - Q) / 60), and incremented a count bin. With "floor" the zero bin has only positive deltas, negative values like -0.3 go into bin -1. (Examples: #-2 bin = -2 to -1 min, #-1 = -1 to -0, #0 = +0 to +1, #1 = +1 to +2) Positive means Trump posted after Q (the direction of interest, since a LARP could choose to post after Trump but could not anticipate when Trump will tweet)

Due to the low sample sizes (e.g. zero delta had 28 events), I repeated the analysis with 5-minute bins to even out statistical noise, and pool together deltas from 0 to 299 seconds into a single bin (from Q's phrasing some intended small-delta events may have been over 60 seconds.) Limiting to +/- 1 hour due to time of day effect slope: since Trump often tweets in the day while Q often posts late, big offsets skew towards Trump first (negative delta)

Results:

• With 1-minute bins, zero delta (0-59 second deltas) is high (28) but only near 90th percentile with more complete and accurate data, and a wider view (6 higher counts: 29 at 2,4,13,15 min; 31 at 49,55 min)

• With 5-minute bins, zero bin (0-299 second deltas) had 127 counts, higher than following bins. You have to go out to the 15th bin (75-80 minute delta) before finding a higher count

I noticed something odd, anyone have ideas? In the graph of 5-minute bins showing both positive/negative deltas, there's an abrupt change at zero – setting aside the zero bin itself (Trump posts 0-5 minutes after Q), negative deltas have consistently higher counts than positive deltas. Going out one hour (+/- 12 bins), negative bins (Trump posts first) have counts 117-157, while positive bins (Q posts first, in at least bin 1, meaning 5+ minute delta) are in the range 80-111. Not time of day effect - change is abrupt, not a slope. In plain language, it sounds like Trump is LESS likely to post in the hour after Q has posted, EXCEPT during the first 5 minutes. What does this mean? Is POTUS going quiet because he's busy reading the replies to Q on 8chan? Who knows?

For anyone inclined to see that as a LARP signal (pro-Q hypothesis is elevated rate of Trump posting after Q) this drop-off is not immediate, the "zero bin" looks similar to the negative values to the left. By choosing when to post based on when Trump tweets, a LARP could elevate the count in negative delta bins, but the zero bin itself would look like the other positive offsets, and it doesn't

Conclusion: With better data the signal isn't as perfect as before, but zero delta is near 90th percentile, and signal looks stronger when also considering small deltas under 5 minutes

>>6143959

I feel one of the problems was that the cabal roped people into security clearance traps, so once they realized something was wrong, they were not able to speak except through whistleblower channels - which the cabal had captured. So there are many people who found out too late that they were in a trap, and were unable to communicate about what they had discovered. I have serious sympathy for those people. Of course I understand the importance of secrecy for national security and operational considerations, secrecy itself is not a bad concept. But it was misused. I realized something was wrong in the world before falling into the security clearance trap, and never applied for one for this reason. And I carefully guarded my independence so I would have the freedom to think or speak. (Although the world was clearly a dangerous place to speak, so I still found myself unable to do so, I appreciate the opsec and community principle here of anonymity.) At this point - I want to help, and I am comfortable keeping secrets … but waiting and watching before even considering anything more that would impose a gag order beyond common-sense keeping of secrets.