I’ve always had a problem with using Points Scored as the primary measure of value for players in fantasy football.

This might seem like an odd complaint – after all, the team with the most points wins, right?

But that’s exactly it. The goal of fantasy football is to win, not to score points. And to win, you only need to score more points than your opponent. Playoff tiebreakers notwithstanding, any points you score beyond getting 0.01 more than your opponent are worthless. Same goes for any points scored in a losing effort.

If you’re disagreeing with me so far, you might be thinking: “I know I need to win, but I don’t know what my opponent is going to score in a given week. I’m trying to maximize points scored!”

Almost. You’re trying to ** maximize the chance that your points scored will be greater than your opponent**. Here’s why that’s different.

Imagine you’re in a league where the average team scores 100 points per week.

If you built a team where you score 180 points in half your games, and 50 points in the other half, you’ll probably lead the league in scoring with your 115 PPG average. But you’ll almost certainly only go 0.500 and miss the playoffs.

So the answer is consistency, right? Try to get your 115 in every game? Not quite. You’ll need 120 in some games. A few you’ll need 130 or 140 to win.

The perfect fantasy player is the one that scores you just enough points to win each week. And since you don’t knw what your opponent will score each week, you want to grade players players according to how many hypothetical opponents they help you beat (slash don’t make you lose to) each week.

The way we can approximate how a player helps you do this this is with two pieces of data:

- The impact a player has on your likeliness of winning (Impact)
- The likeliness you started the player when the impact was made (Reliability)

Impact * Reliability = Fantasy Game Wins

## Part 1: Impact

I’m going to use an actual league I play in to lay this out. Some stats on the league:

- 12-team league
- Half-PPR
- 1 QB, 2 RB, 2 WR, 1 TE, 1 FLEX, 1 DST, 1 K
- Average weekly score for teams in the 2020 Fantasy regular season:
**111.6** - Standard Deviation of weekly points scored:
**17.6**

Below is a histogram of the weekly scores of teams in this league in 2020. For example, the 60 bar shows that 2 times during the season, a team scored between 60 and 64.99 points for a week.

One caveat before I continue – this is definitely not the points distribution I expected for the points scored in this league. Given the average, the most common scores SHOULD be in the 105 and 110 bars. Instead, the peak is at 110 and 115 bars, with a sharp drop-off at 130 points.

The curve is probably normally-distributed-enough to continue, but this could be something worth digging into later.

The way you incorporate standard deviation into this is you can approximate how straying from the league average impacts your expected win %.

- Scoring 112 points in a week gives you about a 50% chance of winning (an average week)
- Scoring 94 points = 15.9% chance of winning (1 standard deviation below average)
- Scoring 129 points = 84.1% chance of winning (1 standard deviation above average)

This curve shows how increasing your points scored increases your chance of winning – comparing your chances if you’re approximating with the standard deviation (blue line), and your chances if you’re taking actual data (orange line).

What this means from the perspective of measuring players:

In a week where you and your opponent are roughly equal, **a player scoring 17.5 more points than the average for his position increases your chance of winning from 50% to 84%. That’s 35% of a win. **On the flip side, scoring 17.5 less than average for your position drops you from 50% to 16%.

So, what’s average by position? Here are the average points scored for starters in this league:

- QB: 22.7
- RB: 12.7
- WR: 12.6
- TE: 9.2
- FLEX: 11.1
- *RB/WR/Flex: 12.4
- DEF: 8.6
- K: 7.9

*I’ve rolled RB/WR/Flex together for simplification purposes. Tight Ends are rarely the choice for Flex, and having a “lower bar” for players that happen to be in the Flex position doesn’t necessarily make sense. A player being placed in the Flex spot is often a result of his being in a later game during the course of a week (to maximize flexibility in case of injury), rather than being a reflection of what role the player serves on a fantasy team.

## Part 2: Reliability

Here’s the nexus of factoring in reliability with impact.

When the 100th ranked WR in a given week goes off for 30 points, that would score highly on the Impact side.

The problem is, nobody in their right mind is starting the 100th ranked WR on their fantasy team, unless they have inside information that says a certain player’s ACL is about to explode, or they are literally the offensive coordinator of an NFL team.

Because we don’t want to assume the user of this metric is better than average at starting players, we’re going to take the simple step of applying % Started to the Impact score for each player.

To do this, I’m going to grab Yahoo! start % data every Monday for the rest of the 2021 season, and use that to compile results. The nice part about compiling data here is it reduces the player pool for the metric, as most players have a percent started of 0% each week.

(Note that I realized when reviewing Week 1 that League Size may need to be considered when taking Yahoo! % Start data. Haven’t decided what to do with that yet.)

## Caveats

There is one major intricacy that I’ve decided to not incorporate into this metric; it’s the fact that not all teams have the same mean likely outcome in a given week.

By using the assumption that both teams in a given matchup have equal average outcomes, the biggest impact a player could have is either 0.5 FWS (lifting you from 50% to 99.9999%), or -0.5 FWS. Yet, this doesn’t capture the full degree to which a player can impact your team in a given week. If you’re going into a week as a big underdog, your expected win percentage might be as low as 20% given typical production for the starters on both sides. A 5 TD game from your RB could boost you 40 points, maybe even swinging you from 20% change of victory or 80% by himself. That’s an impact of +0.6, which FWS will never cover.

This perspective is another way of presenting David and Goliath strategies. If you’re David, you want players with higher variance (to swing the upset). If you’re Goliath, you want players with low variance (hey, you, don’t f*ck up).

There are two reasons I decided against incorporating this nuance into FGW:

- Holy crap it would be more complex, for only slight improvements
- The metric is intended to be used to grade player performance, not make start/sit decisions

To elaborate on point 2 a bit: The point of this metric is to grade players by how much they increase you chances of winning games (“You have an A+ team thanks to this player”). Grading players by how much they help you win a *specific* matchup is to take your chance of winning a game as a given (“This player helps A+ teams win more games”). So, I’m sticking with the way the Impact portion is laid out.

As a final note before getting into the results, if the non-normal-curve of the league were to hold up with more data (as in, it’s to be expected that there’s a “wall” around 130 points scored per week), then players that get you over that threshold would be undervalued by FGW. Players that helped you, but fell short of pushing you over the edge, conversely, would be overvalued by FGW.

## Week 1 Results, 2021

Initial observations, looking at the Top 10 and Bottom 10 from Week 1:

- This is a good system for ranking across positions. The top 10 follows intuition pretty well.
- It feels a bit surprising that Cooper and Thielen had usage percentages as low as they did, and it made a difference in their rankings.
- Rodgers getting huge negative points for his bad week demonstrate the value of this scoring system. Anecdotally, I had one team with Rodgers and one with Mostert in Week 1. The team with Rodgers was dead as a doornail after he was replaced by Jordan Love. The team with Mostert still had a small chance going into Monday night.
- Herbert grading so low with 14.38 points is a bit surprising. It shows how the lower QB scores this week weren’t started at a high rate.

Focusing on the players that scored the most by raw point total:

- Jameis is going to have a hill to climb if he’s going to finish the season as a high scorer in FGW. What could easily be his best week of the year gets him effectively nothing as almost nobody started him. It’s fair because he hasn’t helped many fantasy teams yet.
- I haven’t taken this close of a look at Yahoo! percent started before. Dak was really at 87%? Hurts only 57? Are most leagues 10-teamers? If this is true, we might need to add a league size handicap to the Reliability portion.

## What’s Next?

I’m planning to update these ranks each week throughout the 2021 NFL Season. This could be a great system for calculating a fantasy MVP, as it is measuring the impact players have on (hypothetical) wins for their teams. Short of that, it’s a useful metric for quantifying the actual impact players have had on your fantasy season.

Here are the full Week 1 results:

## 12 thoughts on “Introducing the Fantasy Game Wins (FGW) Fantasy Football Player Scoring System”