Today was the 62nd Annual California State Science Fair, and it was my 11th year of serving as a judge for the Mathematics and Software Panel (Junior Division)… and my (mumble-number) year as chair of that panel. As is my tradition, here’s a recap of our panel and my day. Here is the full list of major fair awards and the category awards.

In our panel, the unanimous winner was Saving Lives One Swimmer at a Time. This project was the development of a system in Python to detect when a swimmer is underwater. It did this by comparing images to a baseline, and included mechanisms to eliminate non-relevant items such as slightly moving lane-lines. The software was well-written and commented. The only problems this had was that it couldn’t handle multiple swimmers, or distinguish between those in distress and those not in distress.

The second place winner was Computer-based Automatic Music Creation through Analysis of Existing Music Pieces. This project took as input existing music pieces in the form of note sequences, note sequence repetition, and measures and developed a composition algorithm to develop similar new pieces. The software to do the generation and analysis was written in MATLAB. There was a fair amount of code developed, and all except the chunk that played the resulting music was developed by the student.

The third place winner (and one of my favorites) was Are Your Passwords Secure over Public Wi-Fi? This project looked at the security of unencrypted wi-fi, and also investigated the SSLstrip  and man in the middle attacks. It also looked at how some browsers have not implemented mechanisms to address the SSLStrip attack. There was less programming effort involved, but this guy knew his stuff. What hurt here was the lowered amount of effort compared to the other highly ranked projects.

The fourth place winner was Danger, Will Robinson! Life Critical Computer User Interfaces and the Science of Safety. Although the title was hokey, the subject was not: it explored what was the most effective user interface for a warning display.

Honorable mentions went to three projects: (1) Misspelled! Creating an Accurate Computerized Spell Correcting Algorithm, which was a reasonably good attempt at programming a spell-correction algorithm; (2) Wi-Fi Watchdog: Application to Observe the Indoor Mobility of Senior Citizens, a project that had seniors carrying android devices and looking at signal strength to determine their locations; and (3) Programmatic Signature Fraud Detection, a program that attempted to determine when signatures were fraudulent.

As for the other projects. We had two (#1, #2) related to the Monty Hall problem. Hint: If you are going to do this, make it original. Everyone knows one car and two goats. But what about more than three doors? multiple cars? multiple goats? the effect of different values of the prizes? the values of incentives to switch? Exploring how these change the odds would be an interesting project. The original question? It’s been done to death. However, I must admit that one of the students was a born salesman — he’ll end up hosting a game show one day, or end up being in sales!

And for those wanting to calculate π… again, figure out what would make this unique, because it’s been done to death. For example, one project related to Buffons Needle.  First, you need to understand why it works. Secondly, you need to understand the effects of random  number generator quality on your results… and perhaps you might explore how you can use such a calculation to test a RNG.

You should try not to do projects where the results appear obvious: we had a few where we couldn’t figure out why the project was even submitted. In such cases, you need to make clear what was unique about this. If you are going to be doing pure math, there needs to be something novel there … something that makes us see the hard work and the applicability (if possible) of the results. Most importantly, you need to understand what you are doing. If you are doing common subjects (especially things like probability in sports or gambling), try to find the unique angle — simple effort to duplicate what is likely known makes it hard for you to shine above the middle of the pack.

As for the other projects, we had some good ones related to cell counting and identifying dementia that just seemed to be in the wrong category — they were most likely in this category due to the fact they involved programming or modeling.  We had some others where the student just didn’t seem to think about the problem fully. For example, in a project related to developing a language for evaluating linear programming, the issue was less the program to solve the equation, and more (in my eyes) about the parser developed to address the language. This was not robust and provided no error feedback. This placed the project more in the middle of the pack.  We had another on security algorithms for attack/defend, but the student had difficulty explaining how the algorithms worked or understanding how the input and results mapped to the real world, other than they affected placement of resources. This was an example of a timely subject hurt by the understanding.

I should clarify here that, as usual, projects had some clear divisions. From the end of the first session, we knew the likely leaders because every judged liked them to some degree. We also knew the ones at the lower end of the spectrum. The great bulk of the projects were in the middle. This didn’t make them bad, but it didn’t make them outstanding. So if you are doing a project, you need to ask what you can do to make yours outstanding. From what I’ve seen, in this category, the answer is: (1) understanding what you did and all aspects of it; (2) being able to communicate that understanding; (3) being able to show that there was some significant effort put into the project; (4) if you developed code, developing the code using good techniques and making it robust.  If you want your project to be able to move beyond the category winner, I’d suggest making it have some utility that makes it stand head and shoulders above the typical engineering or scientific projects — which is something hard to do with math and software.

[ETA: When I generate the project pages, I've been adding a stamp of the form illustrated to the right. This illustrates the various dimensions of judging according to the Judges Handbook. I've done this for about 5 years now, and every year I get annoyed because it doesn't work. By the time the next CSSF comes around, I've forgotten about the problems with this and use it again... and curse again. Perhaps this year, by making this note, I'll remember. There are two problems with this form. First, using this at the state level with a scale of 1-5 does not work, because most of the projects (by virtue of the fact they made it to state) are already at a 3 level. Thus I end up with a number of projects clumped at the higher levels. I need to modify this to indicate the scale, along the lines of: "[Scale: -5 to 5, where 0 = minimal state level and 5 = exceptional]“. Secondly, the dimensions are unclear, and in some cases, difficult to judge through the interview. I need to rework the dimensions to clarify what we appear to be looking for. This would give the following table:]

As for other aspects related to the day:

First, instead of driving to USC I took Metro. It was lovely. Parked in North Hollywood, Red Line to Metro/7th Street, and then Expo line to USC/Expo Park. Easy-peezy, and something I’ll do again.

Second, I took a little time to visit the space shuttle Endeavour. Nice display; it should be even nicer when they have the new building built. The shuttle itself is both larger and smaller than I expected, and certainly looks more worn. They also have a display of the Rocketdyne command center, the tires, an engine, and various other shuttle ephemera and artifacts. Well worth visiting if you are down at the ScienceCenter. Note: The page talked about timed tickets being required, but that wasn’t the case when I was there. It may have been a special day due to the CSSF.

Share/Bookmark