From the link in the question:
These leveraged ETFs seek a return that is +300% or -300% of the
return of their benchmark index for a single day. The funds should not
be expected to provide three times or negative three times the return
of the benchmark’s cumulative return for periods greater than a day.
(Emphasis added be me)
A leverage ETF will produce a return that is a multiple of the index which in the case of 2 or 3 times the return is worth doing some simple examples:
Consider the case of an index that goes up 10% and down 10% on back to back days. At the end, the index is down 1% as if the index was at 100 at the start, it went up to 110 and then came down to 99.
Now, let's add the leverage and see what happens. A 3x tracks would go up 30% and come down 30% which in the end is down 9% which is much more than 3 times the 1% on the index after 2 days. Again, if we start with 100, then go up to 130, coming down we take off 39 points and are left at 91. Thus, we got 3 x 3 which is way more than what we wanted here.
This is just taking a couple of days. If we keep adding more time then the differences will be magnified which is why the statement is made about the deviations existing. While my example is a bit contrived in taking an up and down, one could also take an up and up to consider how that deviates as well. The index would be at 121 after 2 days while the leveraged ETF would be at 169 which is up 69% which compared to 21% for the index is a bit more than 3 times and thus this deviates as well as compounding happens here.
If you want another example, consider an index that goes down 20% each day for 5 days straight and starts at 100: 100, 80, 64, 51.2, and 40.96 for the end. Thus, it ends down 59% in the end rounded off.
Now, let's take 3 times that return and take off 60% each day for 5 days straight and start at 100: 100, 40, 16, 6.4, 2.56. In this case, we ended up down 97% which isn't 3 times the return which would be over 100%.