Data scientist, physicist, and fantasy football champion

Is Tom Brady bad in the fantasy playoffs?

Tom Brady is a great quarterback.

There, now that that’s out of the way, let me shit all over him.

Tom Brady is a great QB, but last week he did really poorly in fantasy football. Well, poorly for Tom Brady, so 12 points. Still, those of you banking on a classic Tom Brady 20-25 point game feel cheated. So Tom Brady sucks in the fantasy playoffs, right? Remember last year? And the year before?

This was a question on The Fantasy Footballers podcast on Wednesday and it made me wonder: does Tom Brady suck in the fantasy playoffs? Let’s take a quick look at the distribution of fantasy football scores for Tom Brady in the fantasy playoffs (weeks 14-16):

.Tom Brady playoffs vs not-1.png

For those of you not familiar, this is a box plot. The bold line is the median score, the bottom of the box represents the 25th percentile (25% of games have this score or lower) and the top of the box is the 75th percentile (75% of games have lower score than this). I’ve included all of the data from the 2010-2011 season through the present.

It actually looks like he scores a tiny bit higher in the fantasy playoffs overall. So why are we saying he sucks?

.Tom Brady playoffs vs not by year-1.png

OK, yeah, that’s why: recency bias. In 2015 he was absolutely dreadful (how dare he only get 18 points a game after averaging 25 all year?!?) and in 2016 he had two not-great games. And look at 2017! 12 points!?! Outrageous!

He’s fine in the fantasy playoffs. In some years better than fine. But just an inspection of the data like this doesn’t control for opponent or how he was overall that year. He might have had more away games or gone up against tougher opponents in the fantasy playoffs. Let’s fit a linear model that can account for this. Some notes on the fit:

  • Here I’m fitting all games in which a QB had 8 or more attempts. Sometimes relievers come in for a few plays, so we’ll toss those and we’ll toss games with early injuries. It’s not perfect, but there are far fewer of those games than regular ones so I’ll just pitch as many as I can easily and move on.
  • I threw out week 17 because: come on. Just… come on. Don’t play fantasy football in week 17.
  • I included terms for player, opponent, year, home vs. away, and whether it was a fantasy playoff game. I also included interaction terms for player by year (hey, some years are off years), opponent by year (sometimes defenses are great and some years they’re not) and the money-term: player by playoffs (is a specific player, say, Tom Brady, bad in the playoffs?).

Let’s look at the results:

## Analysis of Variance Table
## Response: fpts
##                   Df Sum Sq Mean Sq F value    Pr(>F)    
## player           140  48141  343.86  7.8726 < 2.2e-16 ***
## opp               31   3970  128.08  2.9323 1.156e-07 ***
## year               7    711  101.54  2.3248   0.02297 *  
## playoffs           1     57   57.17  1.3089   0.25269    
## homeaway           1    843  842.79 19.2954 1.157e-05 ***
## player:year      308  21909   71.13  1.6285 2.871e-10 ***
## opp:year         217  14507   66.85  1.5306 2.190e-06 ***
## player:playoffs   93   4325   46.51  1.0647   0.31842    
## Residuals       3140 137150   43.68                      
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Those asterisks on the right side are telling you whether terms are important. More asterisks means more statistically significant, meaning that the average value is different enough and the spread in the data is small enough to definitively say there’s a difference. For example, Tom Brady in 2015 (above) did significantly worse in the playoffs. Those boxes are nowhere near each other. However, Tom Brady in 2016 did NOT do significantly worse in the playoffs; yes, there’s a difference in the median (bold line), but the spread is so large and there are so few data point that it might just be part of the randomness that is fantasy football.

Keeping significance in mind, let’s inspect the results. Player, Opponent, and their interaction terms with year are all clearly important. This is very much expected: some players are great, some opponents are great, and some years some players are better and some years some opponents are better. Home/away splits are also a real thing. This makes sense so far. “Year” being significant means that from 2010 to 2017 there are some years in which fantasy QBs score more or less overall. That’s a little surprising to me. I would expect all QBs overall to do about the same from year to year. This might warrant some future study, but for now let’s just avoid the discussion but know that this term is necessary for the fit.

Now let’s discuss the two insignificant terms: playoffs and playoffs*player. These are asking, respectively, whether all QBs do better or worse in playoffs and whether there are specific QBs that do better or worse in the playoffs. The answer to both questions is no, there is no difference between fantasy playoffs and fantasy regular season, either overall or for specific players. Keep in mind that this isn’t that many games (3 playoff games per year) and fantasy football is pretty crazy sometimes. It’s perfectly reasonable for a player to do a little worse or better in the 3 games in weeks 14 through 16 but that this is still within the overall randomness of fantasy football.

But you all came here to know about Tom Brady. So, even though we know these numbers aren’t statistically significant, I’m going to give them to you anyway.

On a completely unrelated note, I like giving knives as presents.

Controlling for opponent and home/away (maybe he had to play all away games against Denver?) Tom Brady actually scores an average of 1.7 MORE points in the fantasy playoffs. It’s just tough to remember that because for the last two years (and so far this year) he’s dropped off a bit in the fantasy playoffs. For comparison, all fantasy QBs score an average of 2.56 FEWER points in the playoffs (Weather? Defenses stepping it up for the playoff push? I don’t know). But remember that these aren’t statistically significant, meaning that even though these are the average values there is still so much randomness that we can’t say anything definitively. I just gave you those numbers because that’s what you, the fans, want. Oh, who am I kidding? Nobody wants bottom lines. They want hard, honest statistical analysis with caveats.

So there you have it, a little under 2 points more in the fantasy playoffs for Tom, and about 2.5 points less for everyone, so if he gets about a half point less than the rest of year that’s expected. Once again, Tom Brady is better than other QBs. Honestly, I don’t even especially like him, and I’m not a Patriots fan. He’s just a damn good QB. If you don’t trust him, drop him and let someone like me pick him up. It’ll definitely work out well for you, Dima.

Effect of starting/backup QB on WR fantasy scores