1 of 1 people found this helpful
I've ran into this before and I don't have the official Tableau response, but isn't 500 colours enough? Best practices suggest we use no more than a handful of colours, else our mind shuts down and just sees a blur of colours rather than discrete ones. Further the colours repeat (I think after 20) depending on the colour palette you've chosen. I am not sure what you are trying to build here in terms of visualization but I am quite certain there has to be a more succinct and cleaner way to showcase the data you are attempting in some sort of visualization.
Sorry for being the messenger of bad news.
Thanks for the reply. I wanted to show the total NFL Running back production as a sum of each player, with each player with a different color. so when you click on the player name you can see their results with the rest of the league in the back ground.
I figured out of a way bypass the problem: basically sorting the color legend by production and player name (with extra columns created in the data) so my legend is only showing players who achieved certain production in the NFL, and omitting the players with low total yards.
Take a look and let me know what you think!
If you got it working, then great! Critiquing your visual would be a different issue entirely. But seeing as you're a Raiders fan I'll feel sorry for you and assist you here. As a whole limit the colors, define the terms, and if you state something about your p-values, etc make it indisputable.
Open with your most visually compelling piece or at least something that tells a story.
1) I have no clue what is happening in this slide. It's pretty informative if I only select one person at a time, but it does nothing as an aggregate. It's just a kaleidoscope of colour. I would consider limiting the colours maybe to only those who've had X carries and receptions or at lest play in Y many years. Or using action filters to select by conference, anything. I can barely even hover for a tooltip currently because the individual lines are so small.
2) Neat graph, but I cannot tell where the % change increases or decreases because in no way is it normalized. This seems less a graph about production and more a graph depicting how dangerous and short-lived most NFL running backs careers. What if to 'normalize' this you showed what the average starting running back, or at least back that got X many snaps, season yardage was based on his years of tenure?
3) I can barely make out that years 3-5 are the most productive for RB's as an avid fan I can agree that these tend to be some of the more productive, based on this is when guys get their big chance, their bodies are stronger, change to get on a winning team instead of the losingest team that drafted them etc,
4) This seems geared toward fantasy foootball, but when? RBs with at least 9 years of experience are not usually drafted in the first round. Not because those with 9 years aren't deserving by sheer statistics, players don't play long enough to be that productive (usually). Also can you define Elite RB here?
5) Only Stephen Jackson's graph is remotely close to a good fit to an exponential/polynomial 2 trend line. Don't force a visually appealing trend line to fit your interpretation of the facts. Jamaal Charles' line has a R-squared of barely .17 and a p-value of over .62. Usually anything greater than .05 is considered statistically insignficant. The values for Marshawn Lynch are even worse. I like the ability to add individuals to the chart. My suggestion would be to remove the trend line entirely because tableau doesn't automatically try to find a good fitting trend line it uses the rules you've given it. I would instead use maybe a average reference line based on selection and I would maybe show the high and low values, or **** even reference lines for standard deviations if you like statistics enough. But only show for the user selected not for everyone its too busy.
6) You demonstrate an understanding here of p-values. However a sample size of 15 just isn't quite large enough because we can't prove a normal distribution and n>=30, further our last 3 points are n=6,3,2 which drastically reduce the accuracy proof you're demonstrating. Maybe if you want to hold to true principals as I pull this slider to 12 years you only show those whom were playing for 12 years, as I pull it to 10, only show those who played for 10, so that my N in the pre-12 or pre-10 years is not unduly influences by those young sparks that cannot last. Again, define what an RB1 is please.
6 & 7) As with number 6, I dig graphs and p-values maybe change the filter. You're arguing that longetivity is a predictor of success. I would suggest changing how the filter works such that when I say 3-9 years it graphs me everything that players with 3-9 years of experience did including year 1 to year 9 of their productivity. This way it more accuracy represents the hypothesis in question.
9) Go Chiefs
Finally, good work. I love stats, I love Tableau, and I love Football. But I think its best to enter analytics without trying to prove something but instead keeping a completely open mind and letting the numbers speak. RB numbers peak at a certain point as do all physical aspects of human performance, except shuffleboard somehow. I cannot remember where I've seen a certain visualization but it showed every position and based on the number of games and/or seasons they;ve played and tracked their yearly or game production. It showed the humps in all positions where statistically individuals had their best years, not compared to their own years but compared to the average person playing in their X year. Such that Urlacher in his 8th year was compared to all MLB playing that year. And whatever his Wins above replacement was, it was recorded for all year 8 MLB. It was fantastic and I think that something like this would be a great place to go. Except in terms of fantasy points, it would be points above average player, etc.