November 1st —- For discussion about week 9 streaming, join the exchange here on Reddit. I’m not getting all the signs that space is bringing you the value I hoped you’d get, so definitely indicate if I should keep it up!

For this week, I’m keeping the same accuracy discussion as last week— I think it’s an important analysis and worth understanding.

As one addition, I’m including one chart I’d prepared for week 7, regarding kicker accuracy. For kicker, in particular, I care more about getting higher average scores at the top. That’s because the usual approach with correlation coefficients is especially prone to errors at the bottom of the pack: multiple random kickers at the bottom of the pack can always surprise with a boom kicking game, which makes the correlational accuracy appear worse than the results experienced.

So, as I’ve shown before, here is a view of the average scores (as of week 7) for the “Top N” kickers. The way to read this is pick a number of kickers— let’s say 6— and interpret it: “on average, the top 6 ranked kickers have produced 8.5 fantasy points over the weeks”. Kickers have sure been random this year, so it’s good to check in and ask if the model at least keeps up with the other sources. As you can see, each source has a different pattern. Source #2 (blue) does poorer with their top-most kickers, but actually has picked out better candidates after the top 4. Ignoring that source, the other 3 of us converge, while my model has given relatively more advantage in the top 5.

A look at D/ST: 2022 season accuracy vs. historical predictability

Also since we’re at mid-season, I wanted to give some update on accuracy.

Here is a look at the current status of D/ST accuracy. The week was mediocre, but all sources are also doing poorly:

The good news: Some of you won't believe it, but I'm actually kind of 1st for D/ST.... I say that meekly because of the bad news. But the meaning is: on average, my D/ST models have correlated with D/ST fantasy outcomes no worse (on average), when compared to the other top D/ST sources I trust.

The bad news: Even the best D/ST rankings this year are still crap: All D/ST ranking sources for 2022 are doing much worse than usual (mine included). Other ranking sources don't draw as much heat on this forum, but on average they're making just as many bad calls in 2022. No, it's not just Buccaneers vs. Panthers, but that's a good example.

And for the first time, I also have a view of how my rankings performed on the “Yahoo!” scoring format (where yards-allowed are not penalized in the scoring). There is little difference, but predictability is generally lower:

And for the big finale…

I have put together this overview of historical Predictability of D/ST, by assembling all correlation coefficients for all weeks in the last 12 years. On the far left, you can see how frequently each correlation occurred.

On the year-by-year line graph, you can note that 2015 and 2017 were less predictable years. Then take note of the “2022 average” line— what you see is that this current year is near the bottom compared to years prior.

Finally, you can see how things bounced around by week, on the right:

A quick look at Betting Lines returns this year

Week 7 was somewhat disastrous, almost wiping out the good gains of week 3— So that also makes it a good time to remember it’s gambling, and therefore counter any high expectations.

Before this week, you can see that the returns were almost following (slightly lagging) the average historical returns. A -50% week is well within normal, though (as I reported a couple weeks ago), so it is to be expected.

Hope you enjoy browsing the site, and thanks everyone for following along!

/Subvertadown