It "confirms" no such thing.
Since the 1994/1995 season, Spurs have won, on average, 40.3% of their matches and scored 48.9% of the points available. For ease of argument, I'm going to round the former to 40% and drop the latter, since it's a misleading/useless statistic.*
As I said in the previous post, win percentage is a binomial. Even though a match can have three results, here we care about only two: either you win the match or you don't (losses + draws). It's like flipping a coin: either it comes up heads or it comes up tails.
But while we know that, with a coin, it'll
eventually be heads 50% of the time, we don't know ahead of time what Spurs' win percentage will be. But now we do. In order to flip a coin to see if Spurs will win or lose, you need a coin that, somehow, comes up heads only 40% of the time.
However, as anyone who has flipped a coin knows, if you flip a coin only a handful of times (say, ten), then it's
entirely reasonable to assume that it won't be an even 50/50 split. Sometimes you'll get four heads. Sometimes six. More rarely three or seven. Even more rarely two or eight. And very, very rarely, you'll get one or nine or none or ten. So even though, over the course of 757 matches, Spurs won 40% of them, it's
completely reasonable to assume that, given 20 matches at random, they will have won eight, or seven, or nine, or even six or ten.
That's the problem with the chart: it takes entirely reasonable fluctuations and posits that there is something "#spursy" about what can be described more directly as near-random chance. This is the problem of sample size. To wit:
In a binomial distribution, the standard deviation is the amount that you can assume, on average, that the results will differ from the mean. So looking at August, that means that, given 69 matches, we can assume that Spurs would normally win 40% of them, or 28. But if we pick any 69 random matches from the pool of 757, we can assume that Spurs will have won, on average, between 24 and 32 of them, since the standard deviation is 4.
Even results that are over one standard deviation from the mean are considered, in terms of statistical inference, as within the realm of normal variance. Typically results have to drift
two standard deviations away for people to start taking notice.
Our "worst" month in terms of variance from the mean, April, looks rather damning. 32% instead of 40%! But looks can be (and are) deceiving. Even being 1.58 standard deviations from the mean means only that, if you picked 91 matches
at random from the pile, we'll have won 29 (or fewer) of them over 6% of the time. Still not in the realm of statistical significance. This is especially problematic since picking 91 matches
at random ignores dependencies that exist among matches played close together. In other words, if months mattered, and last week's match had an effect on next week's match, then the correlation standard is
even higher.
In short, the null hypothesis ("what month Spurs are playing in has no correlation to how likely they are to win the match") remains unassailed. When trying to "confirm" an effect, one has to work hard to disprove the null hypothesis. Your charts, in breaking down win percentage by month, have not met that standard.
* Not all point combinations from the "available points" pool are possible, so probabilities strike me as exceptionally tricky to calculate. All the statistic tells us is that draws happen with some frequency.
