Dear reader,
In time series forecasting, the first step before building a predictive model is to establish some benchmarks: simple, intuitive algorithms whose performance we aim to surpass.
This concept is easier to grasp with an example, I think. Suppose I asked you to predict tomorrow's average temperature. The simplest answer you could give, is to predict that tomorrow's temperature will be the same as today's. In technical terms, this is known as "naive prediction": assuming the next value to be equal to the last observed one.
While there are other benchmarking algorithms, the idea behind them remains the same: they are easy to understand, quick to implement, and inexpensive to run. Any complex model we develop should at least outperform these simple baselines. Otherwise, it would not make sense to invest time, money, and computational power into a hyper-complex model, if it can’t beat a naive algorithm that takes just a few lines of code to implement.
Here’s the twist, though: in practice, these benchmark models prove surprisingly hard to beat. Take the M5 competition, a recent edition of one of the most renowned forecasting contests. Only about 48% of the participating teams managed to outperform the naive algorithm we outlined above, and a mere 7.5% surpassed all the benchmarks set by the competition’s organizers.
When I first studied time series forecasting, this result stood out to me the most: the blatant lack of correlation between complexity and effectiveness. To be clear, I’m not saying simple models are always better than complex ones. After all, in the M5 competition, 7.5% of participants built models that significantly outperformed the benchmarks. What intrigued me was the practice of comparing complex solutions to simple ones; and realizing that, often but not always, complexity is just wasted effort.
Since then, I’ve started noticing this everywhere.
Take finance, for instance. It’s a well-known fact that most day traders fail to beat the market. Despite all their effort, predictions, and technical analysis, stock picking often underperforms a simple buy-and-hold strategy over an index fund. In other words, all that complexity fails to beat the benchmark.
Or consider something as everyday as hosting a dinner party. You want to treat your guests to a great meal, maybe even impress them with your cooking skills. Before you know it, you’re planning a three-course menu, complete with appetizers and dessert, spending days preparing everything. But if the real goal is simply to enjoy time with friends, wouldn’t a simple yet delicious meal (say, like tagliatelle al sugo with a glass of red wine and a lemon-seared bream) be just as effective?
Sometimes, we do even worse: not only do our actions fail to achieve anything, but they end up making the situation even more damaging than before. We strive toward a goal, but our actions backfire, creating new problems on top of the old ones. In these cases, we end up toiling twice: once when we act, and again when we have to fix our mistakes. Ironically, we would have been better off doing nothing at all.
Personal finance provides another good example: constantly monitoring your investments can trigger emotional reactions to market dips, leading to panic-selling, often at a loss. By doing less (not obsessing over the daily fluctuations) you might have avoided unnecessary financial setbacks.
Or take fashion. You buy a trendy shirt, thinking it will elevate your wardrobe, only to find it looking outdated in a couple of years, compelling you to replace it with another trendy piece. This cycle continues, leading to unnecessary spending and an overflowing closet filled with items that quickly fall out of favor. In contrast, investing in timeless pieces (like a well-fitted white button-down) ensures you always have something stylish to wear, without the constant need for reinvention.
These are small, everyday examples, but I believe this idea runs much deeper, even into our emotional lives. We often react emotionally to situations in ways that create more harm than good, leaving us to deal with the consequences of both the initial problem and our response to it.
Looking at all these examples, a clear pattern emerged: it's surprisingly difficult to do better than doing nothing. We frequently overcomplicate things, believing that more effort, more detail, or more sophistication will yield better results. Yet time and again, simplicity proves to be just as effective, if not more so. The more I recognized this, the more I began to question whether elaborate strategies are truly worth the energy we invest in them, or if, in many cases, stepping back and doing less is actually the wiser choice.
At first, this realization was frustrating: it felt like so much of what we do is unnecessary, misguided, or even counterproductive. But over time, I came to see the liberating side of it. The simplest solution is often not just enough; it’s the best one. Accepting this means we can stop wasting energy on needless complexity and focus on what truly matters. And that, I think, is a powerful shift in perspective.
So here’s my advice: when faced with a problem, don’t overthink it. Try the simplest solution first: you’ll often find it works just fine. And if you really want to improve on it, proceed with caution. Otherwise, you might just create more problems than you started with.
Vale,
Davide