If you came by our booth at Strata, you probably noticed the big jar of peppers in the front. We were encouraging people to guess the number of peppers in the jar for a chance to win a PS4. This was surprisingly fun, as many of our contestants spent a significant amount of time inspecting, analyzing, and refining their guess.
With over 130 guesses, we’re happy to announce a winner. There were 597 peppers in the jar, and the closest guess was 600. So congratulations to our winner, a gentleman from Cloudwick. We’ll be doing this again, with more peppers and more prizes, so keep an eye out at your next conference.
Being nerds, the distribution of guesses was quite interesting to us. Any time you have a bunch of data points, it’s time to drop them into Excel1 and see what emerges. First, some basics. We threw out one guess, as it was 3x the next largest guess so we assumed the person wasn’t really guessing. That left us with:
|95% CI||701.8 – 1,193.8|
Unsurprisingly, it looks like the wisdom of the crowd is working here, since the guess median was within 1.5% of the actual number. So congratulations to you, Strata crowd, you came pretty close!
But individually, people were all over the place. Such a high mean vs median indicates a right skew in the distribution, and indeed, that’s what we see (our charts cut off at 2,500 for readability purposes, but there were eight guesses between 2,501 and 10,000):
Another interesting point is that the most common guesses clustered well below the actual value of 597. Indeed, the most popular guesses were 300 and 500 (n=4 in each case). Also, there were a couple of peaks in the guesses, one around 150, and another around 300. What happens if we increase the resolution of the chart?
At 4x the resolution, we can see that there are several peaks that got lost in the earlier chart. In fact, it seems like people like making round number guesses. Hmmm…here’s a histogram of the last 2 digits of the guesses:
Yup. People like round guesses: 0, 30, 50, 60, and 80 to the point where nearly 20% of the guesses were of the form 100, 200, 300, etc.
Since we’ll be doing this guess-the-peppers contest again, we thought we would give out some advice based on our observations:
- Don’t guess a round number. There’s not a better chance that we put exactly x00 peppers in the jar, so guessing a round number is a mental shortcut and nothing else. In the best case that we did put exactly x00 peppers in the jar, you will be competing with several other people who shared your guess.
- Ask the right questions. The most common question we got was “Can you give me a hint?” Since we’re not going to tell you the absolute number of peppers in the jar2, there’s pretty much no way that our answer is going to be helpful. A much better question would be “What was the guess of the person before me?” That guess might be terrible, but it IS a datapoint.
- Take your time, but not too much time. Just throwing out a random guess might not be helpful, but we can’t emphasize how many times we had someone come over, spend several minutes carefully analyzing the jar, and then throw off a guess that was hundreds (or thousands) off. So time spent had little or no correlation with correctness of the guess.
- If all else fails, resort to trickery or bribes. We were shocked when not a single person offered to trade a tchotchke, other prize, or cold hard cash for the answer, or at least a better hint. (Of course, we wouldn’t have accepted the offer, but it doesn’t hurt to try!)
1. Why Excel, and not R or something else? With n=134, the type of tool makes very little difference, so we optimized on ease-of-use and turnaround time.
2. Certain of us tried to mislead you if you asked this question.