It expects death rate to be constant over the year, which is wrong.
You need to take the total number of deaths per capita, for 2020, and compare that to the same number for 2019, 2018, 2017 etc. And look at the standard deviation.
You can clearly see bumps due to flu season in previous years, but because flu season happens every year since before 2013, this becomes a regular seasonal effect and don't show up as 'excess' except in late 2018. We also happen to know that flu was greatly suppressed by anti-Covid measures these last two years.
But I'm still not sure what I'm looking at, overlaying a single year over an average is always going to look "excessive". I want to see previous years as well, why not just extend the graph? That's the one single most important number that you never get so see. Here they have buried it in a complex algorithm which looks honestly suspicious at first glance, upper bound for excessive death is basically the same as average. Is standard deviation really that low?
EDIT: I just want to see how many people died per capita in 2020, 2019, 2018, 2017 and so forth, why is that not shown anywhere?
Because that statistic makes the whole thing seem very mellow. Same thing as average life expectancy by year; the change from 2019 to 2020/21 is very slight.
In other words, Is the emperor actually naked, or are you simply fishing for the conclusion you want to hear?