I practice typing on a site called keybr.com. I get statistics of my typing speed in graph format as well.
Here, I guess the straight green line is a line fitted to my typing speed data. There is some function called 'Smoothness'. From Wikipedia, I see that smoothing filters noise and modifies some data. I see that as I change the value of smoothness, the line changes its slope and my typing speed apparently decreases when smoothness is high.
Can anyone give an intuitive explanation of why does this happen?
What is the correct/optimum amount of smoothness so that I get the truest trend of my typing speed over time? According to the graph, does my typing speed actually increase or decrease with time?
The original plot is very noisy. If you remove the green best-fit line, then the connection between typing speed and time is not very interpretable.
The idea behind smoothing is to fit the data to a (potentially non-linear) curve that best represents your typing speed over time. However, this requires adjusting the data and what you see is just a smooth approximation of the data. In general, there is no optimal amount of smoothing, it is usually subjective (to make your data more interpretable)!
In your case, adding too much smoothness resulted in a curve that didn't resemble the initial trend very well. There is a pretty clear upward trend with 0 smoothness (as shown by the green linear regression line). In my opinion, I think the straight green line tells you what you need to know about how your typing speed has changed over time. Ultimately, it is all about whether you value interpretability or accuracy.