Just because a graph has peaks and troughs does not mean it is systematic. Any "two-week cycle" you are seeing is something your mind is imposing on the data. Do a Fourier analysis and I am pretty sure you will find there is no reliable pattern. I can say this with conviction because it is a random event and by definition will have no pattern.
That 10 day bad patch, clearly shown on the graph, is a certain width. Compare it with other peaks & troughs & there is something there.
However, if it's only 10 days long ( for example ) then any pool-jumping will be pointless.
If you're above average for a month that's worth a look at.
Luck-wise for the past 2 months we've been average. We gained the lead we have due to a 2 month-ish period before then where we had good luck. The gradient on the bottom graph shows the changes. Which is differentiation, not Fourier.
We had a steeper gradient 2 months back to get to the gap between green & yellow we now have. That gap has stayed fairly constant in size since, which clearly shows us shifting back to average ( or it would have got bigger or smaller ).
It still means you have to use monthly averages to get a real view of what's going on.