Welcome to Regression Alert, your weekly guide to using regression to predict the future with uncanny accuracy.

For those who are new to the feature, here's the deal: every week, I break down a topic related to regression to the mean. Some weeks, I'll explain what it is, how it works, why you hear so much about it, and how you can harness its power for yourself. In other weeks, I'll give practical examples of regression at work.

In weeks where I'm giving practical examples, I will select a metric to focus on. I'll rank all players in the league according to that metric and separate the top players into Group A and the bottom players into Group B. I will verify that the players in Group A have outscored the players in Group B to that point in the season. And then I will predict that, by the magic of regression, Group B will outscore Group A going forward.

Crucially, I don't get to pick my samples (other than choosing which metric to focus on). If I'm looking at receivers and Ja'Marr Chase is one of the top performers in my sample, then Ja'Marr Chase goes into Group A, and may the fantasy gods show mercy on my predictions.

And then, because predictions are meaningless without accountability, I track and report my results. Here's last year's season-ending recap, which covered the outcome of every prediction made in our eight-year history, giving our top-line record (46-15, a 75% hit rate) and lessons learned along the way.

Our Year to Date

Sometimes, I use this column to explain the concept of regression to the mean. In Week 2, I discussed what it is and what this column's primary goals would be. In Week 3, I explained how we could use regression to predict changes in future performance-- who would improve, who would decline-- without knowing anything about the players themselves. In Week 7, I illustrated how small differences over large samples were more meaningful than large differences over small samples. In Week 9, I showed how merely looking at a leaderboard can give information on how useful and predictive an unfamiliar statistic might be.

In Week 11, I explained the difference between anticipated regression and the so-called "Gambler's fallacy", and in Week 12, I talked about retrodiction, or "predicting" the past as a means of testing your model.

Sometimes, I use this column to point out general examples of regression without making specific, testable predictions. In Week 5, I looked at more than a decade worth of evidence showing how strongly early-season performances regressed toward preseason expectations. In Week 15, I examined evidence that the best teams were barely more than a coin flip to win any given game in the playoffs, and less than that to win the entire thing.

Other times, I use this column to make specific predictions. In Week 4, I explained that touchdowns tend to follow yards and predicted that the players with the highest yard-to-touchdown ratios would begin outscoring the players with the lowest. In Week 6, I showed the evidence that yards per carry was predictively useless and predicted the lowest ypc backs would outrush the highest ypc backs going forward. In Week 8, I discussed how most quarterback stats were fairly stable, but interceptions were the major exception.

In Week 10, we looked at how passing performances were trending down over the years and predicted this year would set new lows for 300-yard passing games. In Week 13, we discussed how most players declined slightly late in the year, but predicted that rookies would improve. In Week 14, I explained that "hot streaks" were largely just random clustering and predicted that the "hottest" players would regress to their season averages.

The Scorecard

The league had its biggest passing explosion in months, with five teams topping 300 yards in Week 15 (which was the pace it needed to maintain to prove our prediction wrong). Unless half the league passes for 300 yards next week, though, it will be far too little, far too late.

Seven of our seventeen rookies finished Week 15 with zero catches, which isn't far out of the norm (this is a very bad rookie class so far), but makes this prediction very swingy. Tai Felton, for instance, had one 10-yard catch in Week 14... which was enough to guarantee he'd top his early-season ppg average even if he never touched the football again. (He only caught one ball for nine yards through Week 12.) Had that one catch been in Week 12 or Week 17, though, he likely would have "failed" his part of the prediction.

I wish there was some way to make the prediction a bit less noisy (for instance, a way to remove Matthew Golden's zero in Week 14 as he was working his way back from injury and only played five snaps, far and away his lowest total of the season). But I fear any discretion I grant myself to clean the data would be at risk of abuse (finding reasons to trim datapoints that were fluky and unfavorable while leaving the ones that were fluky but fortunate).

Finally, our "hot" players appear to have only regressed 34% of the way to their full-season averages when I need them to regress at least 66% of the way, but this is partly an oddity with how I'm calculating the data (I'm taking an average of all the averages, which means players who have only played one game instead of two are weighted more heavily at the moment, and those players have overperformed on average). It'll work itself out of the data over the next two weeks.

Time is On Your Side

Why do I make predictions in this column about what will happen over the next four weeks? Because five is too many, and three is not enough.

If I could, I'd make every prediction for the entire season-- or better, for the rest of a player's career. But the key to this column is accountability— I know that regression to the mean works, and I want to show it in action. For accountability, each prediction needs to have an end date so it can be graded and scored.

Given that each prediction needs a stated end, when should that end be? And here we get into the power of time. The more weeks a prediction covers, the more likely it is that random noise will wash out, and regression will dominate the results.

Each extra week a prediction runs increases the likelihood that it pays off. This is also why I try to make my comparison groups as large as feasibly possible— the more players involved, the more likely regression shows up. Every observation-- every game from every player-- is a coin flip that's weighted slightly in my favor. 

If I bet heads on a single flip of a coin that's weighted 55% to heads, I'll lose that bet 45% of the time. If I bet it a million times, I'm statistically guaranteed to walk out a winner.

This is all great in theory, but fantasy football isn't iterated a million times; we choose our players and live with their results. I can note that low touchdown scorers tend to reach the end zone, but if you start a player in your league's championship as a result and he doesn't score, the fact that he's even more likely to score next week doesn't help you; your championship is over next week, the trophy is already awarded.