The best time to post on Hacker News is just the emptiest time

Every guide says post Tuesday to Thursday, 8–10am Eastern. We measured six months of Hacker News and that window is the worst named window in the data: 5.72% success against a 6.49% baseline. The effect people are chasing is real, but it is not about when readers are awake. It is about how many other submissions you are competing with, and the popular advice reliably sends you to the most crowded hours of the week.

What the data says about the day you post

Mon6.4%Tue5.8%Wed6.0%Thu6.2%Fri6.5%Sat7.2%Sun8.2%Sunday posts succeed 41% more often than Tuesday postsOrange/grey = P(score ≥ 50). Pale = submissions per day. n = 186,890 stories, 182 days.
Success is P(final score ≥ 50). Weekends are the good days and they are also the quiet days — Sunday sees 748 submissions against Tuesday's 1,218.
DaySubmissions/dayP(≥50)vs baseline
Mon1,0926.39%0.98×
Tue1,2185.82%0.90×
Wed1,1776.03%0.93×
Thu1,1676.18%0.95×
Fri1,0386.47%1.00×
Sat7467.20%1.11×
Sun7488.22%1.27×

Competition explains it, not attention

If the weekday effect were about when readers show up, volume and success would be unrelated. They are strongly, negatively related: across 182 individual days, the correlation between how many stories were submitted and the share that cleared 50 points is r = −0.70.

The busier the day, the worse your odds — r = −0.70 Each dot is one day. 182 days. stories submitted that day → P(score ≥ 50) quiet days crowded days
Each dot is a single day. The front page holds roughly the same number of slots however many people submit, so the denominator moves and the numerator does not.

This also explains why the folklore is not merely wrong but backwards. "Tuesday to Thursday morning, before standup" is advice that everyone has read, so everyone follows it, so those hours carry the heaviest submission load of the week. The advice destroys its own edge.

How much is this worth? Less than you would hope

Best day over worst day is 1.41×. Real, but small next to what people imply when they tell you to reschedule a launch. And for Show HN specifically — the category most people are optimising when they ask this question — it is smaller still, because Show HN plays a harder game:

All storiesShow HN
Posts measured186,89025,671
Reach 50+ points6.49%2.87%
Best daySunday 8.22%Saturday 3.46%

And the number nobody publishes: the median Show HN scores 2

Half of all Show HNs finish with 1 or 2 points Sample of 4,980 Show HN posts. Median 2. Mean 10, carried by a 2% tail. 52%1-2 points39%3-9 points10-49 points — 5.6%50-99 points — 1.2%100+ points — 2.0% final score, left to right →
Sample of 4,980 Show HN posts across six months. Median 2 points, mean 10 — the mean is carried entirely by a 2% tail.

We think this is the most useful number in the piece, because it reframes what a quiet launch means. If you shipped a Show HN and it finished on 2 points, you did not fail at launching. You landed on the median. Slightly over half of all Show HN posts finish with one or two points, and 92% finish under ten.

We know because ours did. That is what sent us looking.

What we would actually do with this

  1. Stop optimising the clock. The whole timing lever is worth about 1.4×, and for Show HN closer to 1.2×. Any effort beyond picking a quiet day is better spent elsewhere.
  2. If you are going to pick, pick a weekend. Sunday and Saturday win for the unglamorous reason that fewer people post then. Not because a different audience is reading.
  3. Distrust any "best hour" claim, including one you compute yourself. There are 168 hour-of-week cells; the maximum of 168 noisy cells always looks impressive and rarely replicates.
  4. Judge your launch against the median, not against the front page. Two points is the middle of the distribution, not a verdict on your product.
  5. Spend the saved effort on being present. The one thing you fully control is answering questions in your own thread in the first hour. That is worth more than any day of the week.
Method, so you can check it. Every number comes from the public Hacker News Algolia API. We queried daily counts for the 182 days ending 27 July 2026: total stories, stories finishing at ≥50 points, and the same two counts restricted to show_hn. Weekday figures aggregate those daily counts (n = 186,890 stories; 25,671 Show HNs). The correlation uses the 182 daily pairs of submission volume and daily success rate. The score distribution is a sample of 4,980 Show HN posts drawn from six one-week windows spaced a month apart, to avoid a single period dominating. Success is defined as final score ≥ 50, which is roughly the level at which a story has had real front-page exposure. Counts come from Algolia's nbHits rather than from paginating results, so they are complete for the window rather than sampled — except the distribution, which is explicitly a sample.

This is the first of a short series. We are a two-person project trying to get a developer tool in front of people, and we keep finding that the received wisdom in growth is either unmeasured or measured once and copied forever. When we check something and the answer is useful, we will write it up like this — including when the answer is that we were wrong.