Computer Science Events in Science Olympiad

[SciolyMaster]

The more specific events do not cater to everyone's interests. Speaking for myself, I found the Data Science test I took at Princeton last year very fun and interesting, whereas I found the Cybersecurity test at BEARSO to be less interesting. I didn't think Cybersecurity was run bad, I just personally was not interested in how hashing or cryptography works. I currently don't have an interest in cybersecurity/plan to pursue it as a career, unlike data science, which I might hope to advance into as a career one day. And there are people who are like me, but in the reverse direction. I think an event that removes all of these interests can cater to more people. Programming is common to both data science and cybersecurity, along with code analysis, etc.

Doesn't literally every study event either rotate out or switch topics regularly...? If "Data Science" shouldn't exist because it focuses on too specific a topic in computer science, then why should events like "Water Quality" or "Sounds of Music" still occur (which focus on a specific topic in ecology and physics, respectively)? And why shouldn't Anatomy & Physiology and Dynamic Planet cover all the topics at once, instead of switching every year?

Actually, now that I think about it, maybe it would be better to have study event topics be more general and consistent year-to-year, to appeal to more people... (for me at least, DP went from one of my least favorite events to my most favorite when it switched from Glaciers to Oceans...)

[0ddrenaline]

Alright, new event pitch time. Here's my view. Based on chalker's post from last year, the major problems are that the event must be:

1. run-able in a 50 min time block
2. score-able within about an hour
3. not requiring expensive hardware or software
4. not requiring specialized expertise by the event supervisor

There might be more but those were the issues pressing enough for him to mention. Because of point 4, a lot of CompSci trial events are hard to implement. Running a practical programming event often requires a lot of testing, time, and expertise from the event supervisor. We need to confine the event to a certain system that requires very little ES involvement. That's why Game On works relatively well.

So here's the pitch. Artificial Intelligence. Artificial Intelligence is my proposed event where teams compete in a video game, which they control only with pre-written artificial intelligence code. I imagine this would be similar to MIT's Battlecode or Two Sigma's Halite competitions. The game type doesn't matter, as long as it's sufficiently complex so that there are many different possible strategies. There would be a standard game for all competitions in a season. Like the competitions I mentioned, Artificial Intelligence would have a pre-built base game (made by people at the national SO level) and a player interface library for the competitors to interact with. Artificial Intelligence would be conducted in a tournament fashion, and scoring would be based on tournament placement. A number of tiebreakers could be used depending on game type, such as lowest score differential, or the amount of time taken to lose the game.

Here's why I like this idea:

• This could be a very pure programming event. AI programming would test all types of basic programming skills, algorithm skills, etc
• I'm personally a fan of the head-to-head competition format, which isn't done very often in SO. It's the one good thing about Sumo Bots, except the Artificial Intelligence event removes everything that made Sumo Bots a bad event.
• Chalker's first point: must be runnable in a reasonable amount of time. Now, I'm not exactly sure how Sumo Bots was run, but I assume that the tournament format caused some timing issues. I can think of several solutions for Artificial Intelligence. In a worst case, I can imagine that the competitors would be required to send in their code some time before the competition, or impound USB sticks containing their code. Then, the event could be run solely by the Event Supervisor. The tournament schedule would be posted and it would be open to public viewing so that the competitors and their team could watch the game. If a large enough screen isn't available, it could at least be open for competitor viewing.
• Chalker's second point: must be score-able within about an hour. This event would be very easy to score. I see no issue.
• Chalker's third point: must not require expensive hardware or software. This would not require any unreasonable hardware or software for the ES or the competitors as long as the game has simple graphics. One computer and the provided game would be sufficient for both running the event and competing in it.
• Chalkers fourth point: must not require specialized expertise by the event supervisor. This event should be very easy to run if it's made properly. The ES should be able to just run the base game with the two AI input files.

Okay, that's all the good, so here's the bad.
Obviously, this would probably require more work from the National Committees than most other events. Someone would need to make a game each year the event is run, or make tweaks to a game each year to make sure the same AI can't be reused. The game wouldn't need to be a masterpiece. Look at MIT's Battlecode. It isn't beautiful, but it's functional and made by student volunteers. I believe creating this event is possible, although it would require someone's dedication. Heck, I would love to do it myself, and I might try fleshing this out into a full trial event. It would help if there's some way to use open source software, or partner with an organization to assist with making the game.

Okay, now is time for you to poke holes in my idea. How could this event fail, other than that last major point?

[0ddrenaline]

Alright, new event pitch time. Here's my view. Based on chalker's post from last year, the major problems are that the event must be:

1. run-able in a 50 min time block
2. score-able within about an hour
3. not requiring expensive hardware or software
4. not requiring specialized expertise by the event supervisor

There might be more but those were the issues pressing enough for him to mention. Because of point 4, a lot of CompSci trial events are hard to implement. Running a practical programming event often requires a lot of testing, time, and expertise from the event supervisor. We need to confine the event to a certain system that requires very little ES involvement. That's why Game On works relatively well.

So here's the pitch. Artificial Intelligence. Artificial Intelligence is my proposed event where teams compete in a video game, which they control only with pre-written artificial intelligence code. I imagine this would be similar to MIT's Battlecode or Two Sigma's Halite competitions. The game type doesn't matter, as long as it's sufficiently complex so that there are many different possible strategies. There would be a standard game for all competitions in a season. Like the competitions I mentioned, Artificial Intelligence would have a pre-built base game (made by people at the national SO level) and a player interface library for the competitors to interact with. Artificial Intelligence would be conducted in a tournament fashion, and scoring would be based on tournament placement. A number of tiebreakers could be used depending on game type, such as lowest score differential, or the amount of time taken to lose the game.

Here's why I like this idea:

• This could be a very pure programming event. AI programming would test all types of basic programming skills, algorithm skills, etc
• I'm personally a fan of the head-to-head competition format, which isn't done very often in SO. It's the one good thing about Sumo Bots, except the Artificial Intelligence event removes everything that made Sumo Bots a bad event.
• Chalker's first point: must be runnable in a reasonable amount of time. Now, I'm not exactly sure how Sumo Bots was run, but I assume that the tournament format caused some timing issues. I can think of several solutions for Artificial Intelligence. In a worst case, I can imagine that the competitors would be required to send in their code some time before the competition, or impound USB sticks containing their code. Then, the event could be run solely by the Event Supervisor. The tournament schedule would be posted and it would be open to public viewing so that the competitors and their team could watch the game. If a large enough screen isn't available, it could at least be open for competitor viewing.
• Chalker's second point: must be score-able within about an hour. This event would be very easy to score. I see no issue.
• Chalker's third point: must not require expensive hardware or software. This would not require any unreasonable hardware or software for the ES or the competitors as long as the game has simple graphics. One computer and the provided game would be sufficient for both running the event and competing in it.
• Chalkers fourth point: must not require specialized expertise by the event supervisor. This event should be very easy to run if it's made properly. The ES should be able to just run the base game with the two AI input files.

Okay, that's all the good, so here's the bad.
Obviously, this would probably require more work from the National Committees than most other events. Someone would need to make a game each year the event is run, or make tweaks to a game each year to make sure the same AI can't be reused. The game wouldn't need to be a masterpiece. Look at MIT's Battlecode. It isn't beautiful, but it's functional and made by student volunteers. I believe creating this event is possible, although it would require someone's dedication. Heck, I would love to do it myself, and I might try fleshing this out into a full trial event. It would help if there's some way to use open source software, or partner with an organization to assist with making the game.

Okay, now is time for you to poke holes in my idea. How could this event fail, other than that last major point?

I actually quite like this idea. As you said, it incorporates many aspects of computer science in an event that is easily scorable within 1 block, and it can be made extremely easy to run if done properly.

However, since it's pre-built, how would parental/coach assistance disproportionally help some teams over others? I see a potential problem in adults writing a lot of code for the students. In addition, while the cost of entry might not be high, the knowledge of entry would be a potential problem.

[0ddrenaline]

I actually quite like this idea. As you said, it incorporates many aspects of computer science in an event that is easily scorable within 1 block, and it can be made extremely easy to run if done properly.

However, since it's pre-built, how would parental/coach assistance disproportionally help some teams over others? I see a potential problem in adults writing a lot of code for the students. In addition, while the cost of entry might not be high, the knowledge of entry would be a potential problem.

Right, that's a good point. However, it's a valid point in any pre-built event. We would mainly have to operate on the honesty policy. Just like other events, I could imagine the event supervisor asking several questions about the code's design. It's pretty easy to tell if someone doesn't understand their code if you just ask them some good questions. Although this also relies somewhat on Event Supervisor expertise, you could say the same thing about events like Sounds of Music.

Edit: I just realized that Battlecode and Halite are open source, under the MIT License / GPL. Huh, I think I'll really look into this.

[RobertYL]

I actually quite like this idea. As you said, it incorporates many aspects of computer science in an event that is easily scorable within 1 block, and it can be made extremely easy to run if done properly.

However, since it's pre-built, how would parental/coach assistance disproportionally help some teams over others? I see a potential problem in adults writing a lot of code for the students. In addition, while the cost of entry might not be high, the knowledge of entry would be a potential problem.

Right, that's a good point. However, it's a valid point in any pre-built event. We would mainly have to operate on the honesty policy. Just like other events, I could imagine the event supervisor asking several questions about the code's design. It's pretty easy to tell if someone doesn't understand their code if you just ask them some good questions. Although this also relies somewhat on Event Supervisor expertise, you could say the same thing about events like Sounds of Music.

Edit: I just realized that Battlecode and Halite are open source, under the MIT License / GPL. Huh, I think I'll really look into this.

The main point that AC is making is not the fact that it has a pre-built component, but rather the event is scored 100% from the pre-build. Events like Sounds of Music and other hybrid events is that a portion of the event still requires competitors to demonstrate their knowledge, notably with a written exam (and arguably the portion of SoM where the competitor has to play their instrument). If the game is designed to be more robust, with the event supervisor adjusting certain parameters on competition day (kind of like the target distance in vehicle events) or by taking or removing game mechanics, teams could then be given time to adjust their "pre-built" code to better suit the new conditions. This would require more work on the ES end, but could be done with proper development of the game design.