I have posted 1,386 answers on Stack Overflow. That’s quite a lot.
A few of my answers are popular – they pop up on Google as top results. Some of these get voted every day. Even when I am not active on Stack Overflow, I still get a few dozens reputation points almost daily. Lucky me!
Recently I got an idea – are people more likely to vote if they like the user icon? Can changing my icon get me more votes each day?
Currently I have over 93K reputation points – would I have gotten to 100,000 already had I a different icon all of these years?
I picked my 30 most popular answers and tracked the votes they were getting each hour. I also tracked how many views and votes their questions were getting.
Each week on Sunday I changed my user icon. Here are the icons I used:
- The flag of the United States of America. The greatest country in the world.
- The flag of India.
- A photo of a plush fox from Japan.
- My wife Suzi wearing a hat. She volunteered.
- Me and my regular icon. Looking at it again, it is a little over-exposed, and I look a little pale.
Both flags performed poorly, while the Fox and Suzi Wearing a Hat did best.
It also makes sense to look at how many views we got each day:
The number of views is similar each week. Still, relative to the total number of views, it looks like Fox performed a little better than Suzi Wearing a Hat. This trend also appears on a daily breakdown:
I picked the flags of the USA and India mainly because of the time difference between them – there is almost no intersection of the working hours between India and the USA. I wanted to see if the icon affects the times I was getting votes. That didn’t work:
The question voting statistics weren’t constant either. This reflects poorly on the previous results – it is possible it is all random.
Is this a thing? Time for Round II!
I took the Fox and the USA flag to a second round, to see how they did:
Even after a month the fox and flag had consistent results!
I think it worked! Changing the icon seems to affect how many votes I’m getting.
- Maybe the numbers are too small for a meaningful conclusion.
- I don’t even know enough statistics to write an Hello World program in R.
- There was an outlier during the second USA flag run – one question received a lot of views. It doesn’t affects the results much, even when zeroed out.
- USA Flag re-run got a down-vote. I didn’t count it.
- In 2017-06-01 the icon Kobi stole one vote from Suzi Wearing a Hat. Its score should be a little higher.
- All times suffer from an off-by-one error: each hour at X:17 I took statistics that belong to X-1:00. I don’t really care.