determining consistency (1 Viewer)

[jswalker1981]

I'm trying to figure out how consistent a player is at scoring me fantasy points each week. Right now I'm taking the average points per week, and trying to base a range of which their actual points. For instance, based on 1 point per reception, 1 point per 10 yards receiving, and 6 pts per touchdown, Terrell Owens averaged 18 points per week. But his actual points per week went like this:

20, 5, 14, 8, 28, 22, 20, 21, 21, 11, 25, 16, 15, 24, 10, and 24

I thought of a range of + or - 5 points. With that range, he was within his average 8 times, but there was also four more times that he produced higher then the range ceiling of 23 points. Should I count the times a player goes over the range, or should I only count the times that a player doesn't make it to the range floor? In this case, it would have been 13 points. Any help you guys can give me, I will greatly appreciate it. I want to get a jump on my preseason work. Thanks again.

 

[Family Matters]

I think this is one of the more interesting perspectives on FF. I'm not big on over thinking who to draft but I do like to know my guys are consistent. More importantly, they have to be good during weeks 14-16, which is when most fantasy playoffs are.

I always like Faulk because he got better as the season wore on. And in weeks 14-16 he was money.

Don't get too hung up on this but if your looking at a tier of players that perform similar and 1 sticks out as more consistent then take him. Those bad weeks can crush a season.

 

[((Morpheus))]

two words: standard. deviation.

 

[code]

I've written a few times about an approach I tried a few years back...

Consistency alone is not enough (and I predict this thread will quickly devolve into debate about whether it's better to have a guy who is inconsistently great or a guy who is consistently mediocre)...you want players who are consistently high...so I created a model based on standard distribution analysis...I computed the average of the numbers and then computed the standard deviation...theoretically (for a standard distribution) the average minus one standard deviation is the minimum the player should score at least 80% of the time...I called this the "base" score and tried ranking players based on this factor...

The problem is that the sample sizes are so small that it is almost impossible to get usable results...over just one or two season you don't have a standard distribution and outliers impact the results to the point that they are almost unusable...

At best I'd look over the numbers and create a basic index based on my own qualitative interpretation...

 

[jswalker1981]

Uh, not really sure how use the standard deviation method. Sorry.

 

[code]

Uh, not really sure how use the standard deviation method. Sorry.

Standard DeviationExcel Stdev

Excel Stdev

 

[jswalker1981]

Uh, not really sure how use the standard deviation method. Sorry.

Standard DeviationExcel Stdev

Excel Stdev

Thanks!

 

[code]

Uh, not really sure how use the standard deviation method. Sorry.

Standard DeviationExcel Stdev

Excel Stdev

Thanks!

I don't know if you'll be thanking me once you do all the number crunching and come to the same conclusion I did, that the sample size is just too small to make real use of...but I wish you luck all the same...

 

[jswalker1981]

Uh, not really sure how use the standard deviation method. Sorry.

Standard DeviationExcel Stdev

Excel Stdev

Thanks!

I don't know if you'll be thanking me once you do all the number crunching and come to the same conclusion I did, that the sample size is just too small to make real use of...but I wish you luck all the same...

I don't expect this to be a major decision maker in my draft choices, but it is good to have for a tie breaker of very similar players. I was thinking, should I use the past three years for my sample size? It would be a lot of work, but right now, I've got plenty of time as I wait for NFL rosters to be set.

 

[code]

Uh, not really sure how use the standard deviation method. Sorry.

Standard DeviationExcel Stdev

Excel Stdev

Thanks!

I don't know if you'll be thanking me once you do all the number crunching and come to the same conclusion I did, that the sample size is just too small to make real use of...but I wish you luck all the same...

I don't expect this to be a major decision maker in my draft choices, but it is good to have for a tie breaker of very similar players. I was thinking, should I use the past three years for my sample size? It would be a lot of work, but right now, I've got plenty of time as I wait for NFL rosters to be set.

The more data you have the better, but my feeling is the other factors in the NFL (player learning curve, coaching changes, peer personnel changes, injury, etc) change enough from year to year that they will still skew the numbers more than enough to make the data hard to use effectively...Not that I'm trying to discourage you...data is data, just don't over interpret it...

 

[pk4239]

Matt Waldman at FFtoday has done a pretty good job with downside consistency.

http://www.fftoday.com/articles/waldman/05...nsistency_1.htm

But I am with code, less so because of predictibility, more so to do with low correlations between players which tends to lower the standard error (when your guy does poorly there is an almost equal chance your other guys do better.)

Good luck.

 

[Lowen1]

I've been playing around with Waldman's Crank (Consistency Rank) Scores for a couple of years. He is a real believer, and I enjoy reading his analysis. One thing I always worried about was the credibility of the data due the sample size. But, as I recall, Waldman has done some analysis of past performance and shown that his Crank Scores can predict (somewhat) future performance. Last year I marked by cheatsheet with a + or - to indicate favorable Crank Scores for deciding between similary projected players.

The counter argument to Crank Scores is that you can make the playoffs with consistency, but you need home run hitters to win. Would you rather have a guy score 10 points every week, or 17 some weeks and 3 the other weeks? If you can insert the player then weeks he'll get 17, you'd much rather have that guy then the guy that always gets 10.

I view Crank Scores as another piece of information.

PS: JOE & DAVID - PLEASE HIRE MATT WALDMAN. TIA.

 

[BroncoFreak_2K3]

Crank Scores

 

[code]

Crack Scores

I believe that should be craNk, but I like the subliminal message there...

 

[BroncoFreak_2K3]

Crack Scores

I believe that should be craNk, but I like the subliminal message there...

fixed above...sorry