Hacker News Reader

Comment by sillysaurus3 | Hacker News Reader

It's been an hour, and a very clear trend is emerging: http://www.wolframalpha.com/input/?i=bar+graph+13%2C+5%2C+10...

Is there any reason why 4x more people would be banned within four days of a payout than any other part of the cycle?

(This presupposes the poll results can be trusted, which seems dubious at best. But if this survey is trustworthy, then I wonder what it implies.)

View on HN

It's a voluntary poll, hence not reliable. I would never trust an HN poll for something as significant as this. If the results were true it would be worrying, but we need much stronger evidence than this.

Hmm... possibly. It seems strange that this particular poll would attract so many trolls, though. Surely there are some troll results, but most of the results are showing that people are getting banned within 4 days of payout. As long as >75% of participants are honest voters, then it doesn't seem like the trolls could skew the results to the extent that we're seeing.

There may be a selection bias too: people who were banned just before payout may be more likely to respond to the poll.

People who had their account banned in the last few days before payment may be more likely to feel victimized, and therefore more likely to answer a poll to express their grievance.

That's probably true, and in hindsight it's exactly why polls (and statistics in general) are so difficult. Thank you (and also robryk) for pointing that out!

It's why voluntary polls are meaningless. A true random poll would not have this bias.

People who were banned are probably pissed off and chose 4days.

Outside of having actual proof of suspension, I'd view 80% of these as troll results. finding that users with suspended accounts in an hour is unlikely.

Going by http://ycombinator.com/newsnews.html HN was seeing 120k uniques as of 2011. This post has been up for an hour, and attracted 105 poll respondents. (120k / 24) is 5000 users, 105 respondents represents about 2% of that.

So either 2% of all HN readers have had their AdSense account banned, or the poll results are a bit noisy :)

(Obviously this is extremely rough math, but allowing for a huge error margin, 0.5% of HN readers would still seem too high, at least to me)

On saying that, AdSense is an enormous program

HN was seeing 120k uniques as of 2011. This post has been up for an hour, and attracted 105 poll respondents. (120k / 24) is 5000 users, 105 respondents represents about 2% of that.

Actually, HN has grown exponentially since 2011. You'll have to extrapolate this chart: https://news.ycombinator.com/item?id=5253773

Also, the traffic fluctuates depending on which part of the day you measure. There are comparatively few people here during 3AM - 7AM in the US.

Thats true if the results are close, but when one of the options clearly trumps the rest, even assuming there is some fraud ,a disturbing pattern emerges.

IMO, even if it isn't true that Google has banned people for illegitimate reasons, it's entirely possible that when they do find a reason to ban you they sit on your account so they can take what you would have made and keep it along with their normal cut. It seems pretty iffy overall, but in their TOS for Adsense there is the line "If we terminate the Agreement due to your breach or due to invalid activity, we may withhold unpaid amounts or charge back your account."

I think it's also worth noting that it can be trivial to get an account banned if you just get a few people to keep clicking the same ads all day. So 'legitimate' bans could happen even when there was no foul play from the person showing the ads.

Yes, it seemed obvious to me that when I got banned, they run some kind of check on the account just before they pay it out. That much I believe is obvious.

Now, as to whether or not their check is valid - that's the real question.

Reply on news.ycombinator.com