Young_Grasshopper 0 Posted August 23, 2011 I've been playing around with data analysis on last years RB numbers using quartiles. Basically, you order the FF scores of each running back for the year, then look at three cutoffs commonly called Q1, Q2 (median), and Q3. Q1, for example, says that player X got AT LEAST that score in the best 75% of his games. So, Q1 gives you a reasonable floor. Q3 says that player X got AT LEAST that score in the best 25% of his games. Q3 gives you a reasonable ceiling. Without further ado, here are the results for non-PPR. See revised rankings in a lower post. Matt Mueller raised a good point; I revised. Share this post Link to post Share on other sites
BufordT 377 Posted August 23, 2011 When you say "player X got AT LEAST that score", what does "that" represent. Share this post Link to post Share on other sites
TimHauck 1,888 Posted August 23, 2011 interesting...but how exactly are those rankings based off of those 3 numbers (if they even are)? Share this post Link to post Share on other sites
WhiteWonder 2,237 Posted August 23, 2011 When you say "player X got AT LEAST that score", what does "that" represent. THAT score is the number you see listed next to each players name in the 3 columns. for example, Arian Foster's Q1 score is 12... this means foster got AT LEAST 12 points in 75% of his games. hth Share this post Link to post Share on other sites
Young_Grasshopper 0 Posted August 24, 2011 Observations: 1. Michael Bush may be a steal as a consistent RB2 even WITH McFadden in the lineup. Bush's floor and ceiling were both terrific for the 8 games he played last year. If McFadden gets injured, Bush could go nuts. FF Calculator lists his ADP as 9.07 for a 12-Team draft. 2. Ray Rice may be overrated. 8-16 points a game is decent but worse than his nearby RB peers. 3. Brandon Jacobs is being punished too severely in his ADP. 7-16 points a game is not shabby, though he did only play 11 games. 4. Chris Johnson is inconsistent... but we already know that. His ceiling is very high; his floor is very low. 5. DeAngelo Williams had a baaad year even before he was injured. Will things really be different this year? 6. Javid Best and Shonn Greene did not make the top 40. Here are there stats: # Rank FullName Games Q1 Q2 Q3 43 Jahvid Best 12 4 8 10 44 Shonn Greene 13 5 6 9 I worry about them both; Best in particular who had a 3.2 YPC average. He blew up in game 2 last year and sucked for the rest of the season. Pass. Share this post Link to post Share on other sites
Young_Grasshopper 0 Posted August 24, 2011 interesting...but how exactly are those rankings based off of those 3 numbers (if they even are)? I did some massaging of the data to smooth out the numbers behind the scenes. For example, it seemed silly to rank a player who got 8-18 points per game higher than someone who got 7-28 points per game. It's not perfect but is something different. One more way to look at RBs going into the season. Share this post Link to post Share on other sites
TimHauck 1,888 Posted August 24, 2011 I did some massaging of the data to smooth out the numbers behind the scenes. For example, it seemed silly to rank a player who got 8-18 points per game higher than someone who got 7-28 points per game. It's not perfect but is something different. One more way to look at RBs going into the season. got ya. I just wasn't sure if the rankings were just a straight calculation off of the total of the three numbers or the average or something but couldn't figure it out. Share this post Link to post Share on other sites
jgcrawfish 232 Posted August 24, 2011 This reminds me of a buddy of mine high school when we were juniors. It's at the latter part of your junior year when you receive your class ranking. For us, it was essentially an average of our overall grade, then a "quartile" or "quarter" we fell into with that grade. So my buddy Steve, upon receiving his, honestly asks "How many quarters are there?" to which I respond "Um, five". So forgive me for pointing out the obvious, but you only have three "quartiles". I understand the premise of there being three dividing points for the classification into data sets, but that still doesn't negate the fact there are only 3 sets of data. I guess we can assume that all the rest fall into Q4. It also seems to me that you could further separate it by either factoring in the number of carries, or plays, or minutes or something. Guys like Arian Foster and Mendenhall logged nearly every carry for their teams last year, by factoring in the number of touches, you get a more realistic perspective of how valuable they are. But good work none the less, when I draft I try to balance one guy I believe I can count on consistently with one boom/bust guy. This is another effective way of breaking that down. I think I still prefer the "CRank" score, but like your premise here. Share this post Link to post Share on other sites
dpollaro 0 Posted August 24, 2011 thanks for the info any way you can do this for ppr or .5 per reception? Share this post Link to post Share on other sites
mccombsau 5 Posted August 24, 2011 This reminds me of a buddy of mine high school when we were juniors. It's at the latter part of your junior year when you receive your class ranking. For us, it was essentially an average of our overall grade, then a "quartile" or "quarter" we fell into with that grade. So my buddy Steve, upon receiving his, honestly asks "How many quarters are there?" to which I respond "Um, five". So forgive me for pointing out the obvious, but you only have three "quartiles". I understand the premise of there being three dividing points for the classification into data sets, but that still doesn't negate the fact there are only 3 sets of data. I guess we can assume that all the rest fall into Q4. It also seems to me that you could further separate it by either factoring in the number of carries, or plays, or minutes or something. Guys like Arian Foster and Mendenhall logged nearly every carry for their teams last year, by factoring in the number of touches, you get a more realistic perspective of how valuable they are. But good work none the less, when I draft I try to balance one guy I believe I can count on consistently with one boom/bust guy. This is another effective way of breaking that down. I think I still prefer the "CRank" score, but like your premise here. Grasshopper, good research Crawfish, not only do I love your pic but I enjoy your input...all good analysis, you never disappoint Share this post Link to post Share on other sites
steelers2101 7 Posted August 24, 2011 This reminds me of a buddy of mine high school when we were juniors. It's at the latter part of your junior year when you receive your class ranking. For us, it was essentially an average of our overall grade, then a "quartile" or "quarter" we fell into with that grade. So my buddy Steve, upon receiving his, honestly asks "How many quarters are there?" to which I respond "Um, five". So forgive me for pointing out the obvious, but you only have three "quartiles". I understand the premise of there being three dividing points for the classification into data sets, but that still doesn't negate the fact there are only 3 sets of data. I guess we can assume that all the rest fall into Q4. It also seems to me that you could further separate it by either factoring in the number of carries, or plays, or minutes or something. Guys like Arian Foster and Mendenhall logged nearly every carry for their teams last year, by factoring in the number of touches, you get a more realistic perspective of how valuable they are. But good work none the less, when I draft I try to balance one guy I believe I can count on consistently with one boom/bust guy. This is another effective way of breaking that down. I think I still prefer the "CRank" score, but like your premise here. The 4th quartile would be what they scored in 100% of their games, so it would just be their low score on the season, so I imagine that's why he didn't think it was necessary to include it. Share this post Link to post Share on other sites
Young_Grasshopper 0 Posted August 24, 2011 The 4th quartile would be what they scored in 100% of their games, so it would just be their low score on the season, so I imagine that's why he didn't think it was necessary to include it. You got it. Example: let's say a running back played 9 games and scored the following: 6, 16, 7, 11, 20, 12, 15, 3, 11 To calculated the quartiles, you put these scores in order: 3, 6, 7, 11, 11, 12, 15, 16, 20 11 is the median (Q2). You can accurately say he scored AT LEAST 11 points in 50% of his games. Q1 would be 7: he scored AT LEAST 7 points in 75% of his games. Q3 would be 15: he scored AT LEAST 15 points in 25% of his games. The other endpoints would the min: 3, and the max: 20. I guess you could call those Q0 and Q4. I like the concept of the CRank scores, but I don't like the implementation. They place a lot of emphasis on that ceiling score, the Q3. I think the floor, the Q1, is much more interesting since it tells you, barring an off week, how many points you can really count on each week. When I was prepping for last year's draft, Roddy White had a really high Q1 from 2009. The man was very consistent. I drafted him high and sure enough he really carried my team by providing a consistent 10+ points every week. That's much more valuable than a guy who will get you 18+ points 25% of the time but averages 5 points the rest of the season (do you hear me Santana Moss?). When people throw around 1000 yards / 8 TD, it matters A LOT whether he has one game with 3 TD and 200 yards because how do you predict which magical game will be the one with 3 TD? You need guys who have a high ceiling and can win a game for you outright, but the floor is where you build a consistent win patterns week in and week out. Share this post Link to post Share on other sites
Busted by the Feds 9 Posted August 24, 2011 I'm still not drafting Michael Bush with the No. 3 pick . Share this post Link to post Share on other sites
Matt Mueller 146 Posted August 24, 2011 ESPN has Bush playing 14 games. Why did you only compute 8? Good overall stats and idea! Share this post Link to post Share on other sites
jgcrawfish 232 Posted August 24, 2011 I'm still not drafting Michael Bush with the No. 3 pick . That was my premise for breaking it down by touches or plays or something. When you see guys popping into the top 30 or 40 and they only have 5, 6 or 8 games...um, that hurts the overall picture. Share this post Link to post Share on other sites
Matt Mueller 146 Posted August 24, 2011 Observations: 1. Michael Bush may be a steal as a consistent RB2 even WITH McFadden in the lineup. Bush's floor and ceiling were both terrific for the 8 games he played last year. If McFadden gets injured, Bush could go nuts. FF Calculator lists his ADP as 9.07 for a 12-Team draft. 2. Ray Rice may be overrated. 8-16 points a game is decent but worse than his nearby RB peers. 3. Brandon Jacobs is being punished too severely in his ADP. 7-16 points a game is not shabby, though he did only play 11 games. 4. Chris Johnson is inconsistent... but we already know that. His ceiling is very high; his floor is very low. 5. DeAngelo Williams had a baaad year even before he was injured. Will things really be different this year? 6. Javid Best and Shonn Greene did not make the top 40. Here are there stats: # Rank FullName Games Q1 Q2 Q3 43 Jahvid Best 12 4 8 10 44 Shonn Greene 13 5 6 9 I worry about them both; Best in particular who had a 3.2 YPC average. He blew up in game 2 last year and sucked for the rest of the season. Pass. Yeah. As high as I am on Calvin/Stafford/Burelson I'm not sure I share the optimism for Best. Talent and opportunity are there but having owned him last year he really couldn't be counted on for more than say 70-80 yds and no scores after week 2. His overall stats look good but week to week he wasn't a factor. Share this post Link to post Share on other sites
Young_Grasshopper 0 Posted August 24, 2011 ESPN has Bush playing 14 games. Why did you only compute 8? Good overall stats and idea! Good point - Bush did play 14 games according to his stats. I put a filter in the data to only include games where the RB had at least 6 carries... that's probably doing more harm than good. Now that I reran it, Bush plummeted off the list to #51. Scanning some of his games, he tended to share carries with McFadden but usually at a 2:1 ratio when both were healthy. In games where Oakland couldn't run often, this killed his value. Revised list: # # Rank FullName Games Q1 Q2 Q3 1 Arian Foster 16 12 22 28 2 Jamaal Charles 16 11 15 20 3 Darren McFadden 13 10 15 18 4 Peyton Hillis 15 9 13 20 5 Maurice Jones-Drew 14 9 14 19 6 Adrian Peterson 15 9 13 25 7 Steven Jackson 16 9 12 15 8 LeSean McCoy 15 9 14 20 9 Frank Gore 11 8 15 19 10 Ray Rice 16 8 12 16 11 BenJarvus Green-Ellis 16 8 12 16 12 Ahmad Bradshaw 16 8 13 17 13 Rashard Mendenhall 16 7 15 19 14 Danny Woodhead 13 7 10 12 15 Ryan Torain 10 7 12 16 16 Pierre Thomas 6 7 10 15 17 Mike Tolbert 14 6 10 16 18 LaDainian Tomlinson 15 6 10 11 19 Chris Johnson 16 6 15 22 20 Michael Turner 16 6 13 19 21 Matt Forte 16 6 12 16 22 Ladell Betts 5 6 7 9 23 DeAngelo Williams 6 6 6 7 24 Joseph Addai 8 6 10 15 25 LeGarrette Blount 13 6 8 16 26 Clinton Portis 5 6 6 8 27 Ryan Mathews 12 5 7 13 28 Ronnie Brown 16 5 7 9 29 Cedric Benson 16 5 8 17 30 Felix Jones 16 5 8 12 31 Knowshon Moreno 13 4 17 18 32 Fred Jackson 16 4 8 11 33 Marshawn Lynch 12 4 6 12 34 Maurice Morris 10 4 6 11 35 Jahvid Best 15 4 7 10 36 Brandon Jackson 16 4 7 10 37 Cadillac Williams 15 4 6 10 38 Shonn Greene 15 4 6 9 39 Chris Ivory 12 4 5 16 40 Brandon Jacobs 16 3 8 13 Share this post Link to post Share on other sites
c9h13no3 0 Posted August 24, 2011 Past performance does not imply future results. You're too heavily weighting the past. For example, Best is very unlikely to have turf toe again this year. Share this post Link to post Share on other sites
Bpow0407 0 Posted August 25, 2011 Past performance does not imply future results. You're too heavily weighting the past. For example, Best is very unlikely to have turf toe again this year. I don't think its meant to be a predictive scoring method, but rather a look at consistency and what you could expect from a RB last year in most (75%) of their games verses half of their games vs. 1/4 of their games. Share this post Link to post Share on other sites