Twenty hours of video are uploaded to YouTube every minute; in turn, users stream 75 billion videos per year. With all those videos, what are your chances of making a dent in the YouTube universe and garnering, say, 1 million views?

Slate.com writer Chris Wilson recently addressed just that question, concluding that, “You might have better odds playing the lottery than of becoming a viral video sensation.” Wilson gathered data on about 10,000 randomly selected YouTube videos and found that ** 1 in 2.39** videos was viewed no more than 10 times after one month while just

**reached 10,000 streams.**

**1 in 401**The odds are heavily stacked against the user, so much so that in Wilson’s sample, just one video cleared 100,000 views and none went over a million. His article was titled, “Will My Video Get 1 Million Views on YouTube?” The answer, according to Wilson’s data, was a resounding no.

And yet, anyone who has visited YouTube enough times knows that there are plenty of videos with at least 1 million views—too many, in fact, to count. Can we use Wilson’s data to estimate a video’s odds of getting 1 million views? We can, but first we must apply some corrections.

As with many data sets, Wilson’s YouTube stats are heavily skewed, with a huge number of videos bunched at no more than a few views (most commonly 0), and an ever-decreasing total as the numbers get higher. There are, for example, 2,226 videos with no views in their first month, 237 with 1 view, 158 with 10 views, and just 23 with 100. Any statistical estimation demands a more or less normal distribution, where the data resembles a bell curve rather than the power law.

This can be accomplished by taking the natural logarithm of the video streams plus 0.5 (which is there to make all numbers positive, as you cannot take the natural logarithm of 0—it also results in an almost perfect, i.e. 0, skewness). We can then find the average of that number for the sample as well as the standard deviation; from there, we can find the odds of a video getting any number of views.

Let’s go through a numerical example. The average of our “transformed” number of views is 2.56; the standard deviation, 2.26. The natural logarithm of 1,000,000 + 0.5 is 13.82. 1,000,000 page views, then, is (13.82 – 2.56)/2.26 = 4.98 standard deviations away from the mean. In non-technical terms that means it is a very unlikely occurrence—something we already knew! This, however, can tell us how many YouTube videos will garner 1,000,000 views in its first month on the site: about 1 of every 3.1 million.

Those are pretty long odds, but they may be overstated. As you can see in the graph, the statistics are very good at predicting video streams at fairly low viewer numbers but seem to over-state the odds (claiming that the odds are longer than they are in actuality) at very high numbers.

This might have to do with the theory of the fat tail. The vast majority of YouTube videos are user-generated and of little interest to more than a few people. These heavily impact the sample averages and distributions, making high view count videos seem less likely than they actually are. For example, compared to Wilson’s actual results, the statistical method under-predicts videos with 10,000 views by about 50 percent and videos with 100,000 views by about 150 percent.

It seems likely, then, that the odds of reaching 1 million views are being underestimated even more severely, though it is difficult to venture a guess as to how much. It must also be pointed out that our statistics cover only the first month of a video’s life; certainly, its odds would improve if we went out a full year.

Still, regardless of what the true odds of your YouTube video getting 1 million hits are, we can say with certainty that they are very, very low. Then, again, if video of a dance down the aisle at a wedding can beat the million mark 28 times over in just two months, why can’t yours?