It doesn't change the actual odds, it changes the perceived / measured odds.

Imagine 100 people contract the virus, 10 develop serious symptoms, and one dies.

In country A, with widespread testing, 40 of the 100 are identified. The visible statistics are 40 cases, one death, death rate of 2.5%.

In country B, with poor testing, only the 10 with serious symptoms are identified. The visible statistics are 10 cases, one death, death rate of 10%.

Again, this is about the reported statistics, not the underlying truth of the disease.