Print

What’s the Better Indicator for This Week’s Performance?

You are viewing free content provided by FantasyGuru.com. Why not consider subscribing today?

by Mike Horn, Staff Writer

Published, 11/15/13

 

Last week, there was a message board discussion of whether year-to-date (YTD) fantasy numbers or the last several weeks are a better indicator of future performance. So I thought I’d take a look at this.

 

I examined running back fantasy scoring over the last 10 years (2003-2012) in a PPR format (TD = 6 FP, 10 yards = 1 FP, Reception = 1 FP). In order to minimize the impact of injuries on the data, I restricted my data set to only RBs who missed no more than 2 games in the first 16 weeks. Since I just wanted players who were relevant for fantasy at least once during the year, I only included RBs who managed 50 FP or more for the season with at least one week of 10+ FP. This left me with 505 RB seasons.

 

Most discussions of using recent performances focus on the last four weeks of play. I tweaked this a bit to include the last four games rather than weeks – often four weeks will include a bye or injury-missed game and so using weeks would mean some of the comparisons would be for three games and some with four. I standardized it at four games.

 

For each RB season in my data, I calculated the YTD average through Week Five, Week Six, Week Seven, etc. I also found the average for the last four games for each running period. Starting in Week Six, I then looked to see whether that RB’s FP for that game was closer to his Week 1-5 YTD FP average or the average of his last four games. For example:

 

   

Weekly FP Scores

Avg FP thru Wk 5

Avg FP Last 4 Games

Player

Year

1

2

3

4

5

6

Priest Holmes

2003

37.3

32.8

31.6

14.3

15.1

26.0

26.2

23.5

 

In 2003, Priest Holmes averaged 26.2 FP through Week Five. In the four games Weeks 2-5, he averaged 23.5 FP. Then in Week Six, he scored 26.0 FP. The YTD average was closer to that score than the preceding four game average, so that counted as 1 point for the YTD average.

 

If a player missed a game in the preceding four weeks, whether due to an injury, suspension of some reason besides a bye, I threw that data out. So if Holmes had sat out his team’s game in Week Four, I wouldn’t have scored the Week Six outcome. But if he’d had a bye in Week Four, then I would have calculated his last four game average from Weeks 1-5. Since that would have meant his YTD average and last four game average were the same, the result would have been a tie and not counted anyhow, but I think you understand how I treated byes differently than injury-missed games.

 

So I did this for every player and every YTD vs. four game average for all my RBs through Week Sixteen in the 10 years in the study. I ignored Week Seventeen as it would be too much work – and subjectivity – to go through every game and decide whether both teams had something to play for in Week Seventeen. A few meaningless Week Sixteen games were probably included.

 

Here’s the results:

 

FP Score in Given Week Was Closer to the Average FP in the Specified Period (RBs Ending Year with 50+FP)

Week

Year-to-date

Last 4 Games

6

126

101

7

198

169

8

180

182

9

210

177

10

199

205

11

225

201

12

227

214

13

222

212

14

225

208

15

221

208

16

225

204

Total

2258

2081

 

For example, for all Week Six games, 126 times the year-to-date average going into that week was closer to the actual Week Six score than the preceding four game average was. In 101 cases, the four game average “won.”

 

The total number of games scored in Week Six is lower than the other weeks because any RB who had a bye in Weeks 1-5 would necessarily have the same YTD and preceding four game averages, so the Week Six comparisons were ties and not counted.

 

Most weeks, the YTD average is a slightly better indicator of what an RB will score than his last four game average. Over all weeks in the 10 seasons, about 52% of the time the YTD average was closer than the four game average.

 

Some data stuff: some of the results would be skewed by partial games where a player posted a score but didn’t finish. My screen for missed games wouldn’t catch that. This would affect the validity of the four game averages more than the YTD averages and would also impact the week being examined – Week Six for Holmes in the example. This might cause a slight undercount in the last four game “wins.” However, since it would favor the lower of the two averages when it happened in the week under consideration, theoretically it would just about even out. I don’t know this for a fact, though. On the other hand, ties went to the four week average (other than as discussed above). For example, YTD = 23 FP, last four = 21 FP, week being considered = 22 FP. Both averages were 1 FP off – the formulas I used added that to the four week average column. So those numbers might be a little high. Since I was using decimal scoring and averages, I doubt this happened in more than a few cases.

 

I also looked at RBs who totaled at least 100 FP and then 150 FP to ensure that the more marginal RBs for fantasy weren’t overwhelming the better players in the data.

 

FP Score in Given Week Was Closer to the Average FP in the Specified Period (RBs Ending Year with 100+FP)

Week

Year-to-date

Last 4 Games

6

87

72

7

140

122

8

133

126

9

146

127

10

141

147

11

167

136

12

153

158

13

167

142

14

158

147

15

160

143

16

166

133

Total

1618

1453

 

FP Score in Given Week Was Closer to the Average FP in the Specified Period (RBs Ending Year with 150+FP)

Week

Year-to-date

Last 4 Games

6

65

52

7

101

91

8

96

91

9

101

95

10

104

106

11

123

96

12

111

115

13

124

102

14

116

109

15

118

109

16

128

95

Total

1187

1061

 

They weren’t. Actually the YTD average was closer 53% of the time in those two data sets, a 1% better.

 

With all that work, I don’t think there is much difference between the two averages in helping you evaluate players. YTD numbers are usually based on bigger samples and therefore would seem to be better than four week averages skewed by fluke games or a couple of easy (or hard) opponents. On the other hand, the more recent numbers would seem to better reflect the current state of a team’s offense – it’s current efficiency, players in or out due to injuries, rookies improving, veterans wearing down, etc. It turns out, I don’t think either number is really better or worse for helping you figure out what will happen in one game.

 

These numbers might help in estimating what will happen in a series of games; I’ll look at that next.

 

You're going to score 1,994 points this week!

Back to the top