You just pointed out one of the most important (and frustrating) findings from my analysis- while the price of soybeans is affected by their predictable planting/ harvesting seasons, investor sentiment plays a MUCH bigger role here.
Changes in sentiment are influenced by global supply/ demand factors as well as uncertainty, and account for much of the volatility in the price of soybeans. For example, in May 2018, there was speculation of a trade war between the US and China, one of the US's largest soybean customers. As a result of this uncertainty and potential drop in demand, the market quickly responded with a drop in the price of soybeans. A month later, in July 2018, this speculation was confirmed as China imposed a 25% import tariff on US Soybeans- reducing uncertainty, and therefore, causing the price of soybeans to increase.
Now, on to your questions!
So, based on either your domain knowledge or additional analysis, YOU actually chose the value for the period, not the algorithm. In this case, I chose a period of 360 to account for the planting/ harvesting cycle of the soybean crop.
And to answer your last question- it honestly depends. Take retail, for example. There tends to be seasonality in sales data, with sales expected to increase by X right before the holidays and perhaps, by Y in late August due to back-to-school shopping. While this may have historically been a pretty reliable means of forecasting sales, things (like covid and stay at home orders, for example) can come up that affect sales and are not explained by a trend or the seasonality. These are part of the "residual"/ "noise" component which, depending on the supply/ demand factor, can have a greater impact on sales than seasonality.