November 3, 2025

Time to reflect again on the predictability of the 2025 season: month #2 now, following up on the report for month #1.

If you’re new to these: Examining predictive accuracy has been a long-standing tradition, underpinning a key purpose of Subvertadown: To give us an understanding of how predictable things have been. We want to know "Was this season more or less predictable than normal?"

Compared to Other Seasons

So as usual, here’s a look at how each individual model is doing, compared to other seasons.

This does not tell "how good the models are". It only tells us how predictable the current season is, compared to the historical norm.

Whereas in week 4 I said the pattern is different, during weeks 5-8 the accuracies are mostly matching expected level at the blue bars.  The main exceptions are: Very high predictability of Game scores (Vegas), and exceptionally low predictability of Kickers. As a 3rd aberration, TE is also still lagging behind.

Just as in weeks 1-4, Kickers and TE are the positions that are continue to significantly lag behind, in predictability this season. This is not what we normally see.

Comparative Accuracy Assessment

Reminder / for newbies: I’m not trying — and don’t expect to be— “#1”. That’s plain unrealistic to expect that all the time, against very competent expert analysts I track. My goal is rather to check that the models are still performing at a similar level to others, specifically sources that have been consistently good for at least a few years. Many sources are great one year but then poor the next. My chosen experts are good in a more consistent way. We naturally trade places at the top, year to year. Knowing that the models perform at least at a "similarly" to top sources lends confidence and gives reason to trust forecasting, when we extrapolate models to future weeks.

Defensive Maneuvers

D/ST accuracy started the first month amazingly excellent. The second month has been acceptably better than the historical average, but but not as clearly predictable as the first month. My models still stand at #1 on the season, however their accuracy has been average compared to competing sources, for weeks 5 - 8.  

The biggest slip was in week 8, where the model had its worst call in 5 years: Ranking the Bengals D/ST in 3rd (4th in ESPN). And in week 6, the model was too high on the Commanders, while it missed the Broncos D/ST and was especially low on Raiders D/ST compared to other rankers. I did not expect others would be that high on Raiders D!

My conclusion from month #1 was that the model was excelling and had not need for an update. My conclusion from month 2 is that there’s room to improve the projections when the defense is relatively poor in real football.

On a better note, the ESPN model has been doing exceptionally well. In fact, using the ESPN recommendations for ANY other setting would have produced better than average results. I expect this has something to do with the very high accuracy of Vegas this year in predicting game scores. ESPN scoring of D/STs is even more sensitive to Vegas lines than is the Yahoo default scoring setting.

Here’s the Kicker

It’s time to acknowledge it: The year 2025 has turned out to be the worst we’ve seen for predicting kickers, since the terrible low-point of 2022!

All models have struggled. The BEST ranking source I see out there has a correlation (ranking vs. outcome) of only a mere 0.10. That’s the best that I see out there. The real point is that everyone has been similarly poor at predicting kickers.

As for my own models have been exactly average among the other top ranking sources that I track. Unlike last year (where I was clearly leading), this year we are all quite similarly poor.

Here is the plot I usually show for Kickers, which is the “Accuracy Gap” method invented for FantasyPros.  During weeks 5-8, my projections have eerily matched the accuracy of other sources— even though we recommend different kickers. It seems like nobody really knows what’s going on!

One key purpose of these posts has historically been to reflect on areas for improvement. 

When I look at the patterns of kicker upsets this year, it seems to me that there are more of the “PAT busts” than usual.  In other words, teams that are expected to win, or at least have a lead, end up having more TDs instead of FGs. In years past, these kickers would usually get 1 or 2 FGs, but that seems to be happening less. My guess is that there’s a connect to the very high accuracy of Vegas this particular year. When it’s easier to guess winners and losers, it’s worse for kickers because the game script is too polarized. But that’s just a theory.

Even that explanation isn’t very satisfying to me, though, and I will need to dig into it more (in the off-season). I have some upgrades in mind.

Two Cents for a Quarterback

My QB model boomed to start the season. And in weeks 7 and 8 it exceeded comparative accuracy again.

However, something happened in weeks 5 and 6, that was confusing.

Week 5 is forgivable, whereas Week 6 is puzzling. The week 5 dip in accuracy was mostly due to ranking Trevor Lawrence especially low— and to ranking Kyler Murray higher than anyone else (and that was Murray’s last game, when he got injured, before his long absence).

Week 6 could use some more digging into. I ranked Bo Nix and especially Jordan Love higher than anyone else, and they had mediocre 15 point games. I was also highest on Justin Fields that week, in the London game against the Broncos. While I already factor International games, perhaps other rankers put a bigger discount on the two QBs in London.

Overall, it confirms what I usually see, that the methodology seems to suit QB well. I don’t see major needs for correction.

Survivor

The current results are still better than normal, although we saw 3 losses in 3 different pathways. A total of 4 losses out of 24 games in the 3 pathways.

In the second month, the losses were:

• Pathway 1: Cardinals in week 5 (replaced by the Colts backup)

• Pathway 2: Eagles in week 6 (replaced by the Rams backup)

• Pathway 3: Falcons in week 8 (replaced by the Chiefs backup)

 Betting Lines

Our baseline expectations should normally be to lose about -40% of the weekly pot, by this time (after 8 weeks). So, if you routinely bet $100 per week, then the sum of your 8 weeks of losses would be $40.

That means, a normal “coin flip” betting process would lose us about -5% per week, from the target weekly bet amount. Flipping coins pays the bookkeeper. Yes, that’s if we were just monkeys shooting darts.

During weeks 5 - 8, we did NOT maintain the high earnings that we saw in the first month. However, the net ROI was about break-even, for the month. The adjusted win rate was 52.5%, and there was almost no change since week 4.

After I announced a change to the website’s display of betting tables, in the last Accuracy Report, I did mange to make one error after that launch. In week 6, there was a bet of 49ers vs. TB ($700), which was meant to be canceled but it was not. I am, of course, including the loss in the above picture and in the below diagrams. I’m mentioning it because it supports that the model itself resulted in exactly 0 loss, from weeks 5 - 8.

You would probably note that the betting model has suggested lower amounts, during the last few weeks. That is the new normal, for my revised models. Later in the season, the total suggested amount will increase again, most likely.

Here is the list of bets recommended weeks 5-8:

/Subvertadown