I feel like you're contesting my use of the word "risky", but not the substantive point I made. Risk doesn't really imply danger at all in this context, which I would think is self-evident given the usage of the term in colloquial and professional contexts (information security, finance, etc). There are obviously other forms of risk than standing under a piano suspended in the air.
To put my actual point more succinctly: yes, I agree a relevant portfolio item is generally a net positive; my concern is that most candidates without prior experience will develop a "trading algorithm" which is actually irrelevant to the process of trading in most firms. It's not about developing something past "hello world" quality, it's about developing something that resembles what the interviewer agrees is a trading algorithm.
What is the algorithm's core strategy? How was this strategy developed? Was it backtested? Did you overfit? Are you sure? Did you forward test on out of sample data? How does it weight the inputs for identifying eligible securities? Does it dynamically alter these weights in response to real time data? So then how does it process streaming data in a performant way? What time resolution does it require for the data? Okay, now that it's identified eligible securities, how does it manage risk and portfolio size? How many positions can be open and how much of the available capital can be utilized? Why and how are these specific limits enforced? But wait, how is it executing trades? How does it minimize market impact and slippage? How does it attempt to anonymize trades? Are you sure your algorithm isn't predictable? Can it be baited by market makers (i.e. your algorithm is trying to increment up to a minimum bid that will be sold to, which a market making algorithm takes advantage of by raising their price ahead of you)?
If a candidate develops an algorithm that handles considerations like these and can reason about them intelligently (if not necessarily correctly) in an interview then yes, it's impressive. They don't need it to be complete and ready to deploy, they need to understand why all of these things matter. If they don't, they haven't developed a trading algorithm, they've just slightly automated point and click trading on an Interactive Brokers GUI.