*Positional Value in the DSFL: A Study - Printable Version +- International Simulation Football League (https://forums.sim-football.com) +-- Forum: Community (https://forums.sim-football.com/forumdisplay.php?fid=5) +--- Forum: Media (https://forums.sim-football.com/forumdisplay.php?fid=37) +---- Forum: Graded Articles (https://forums.sim-football.com/forumdisplay.php?fid=38) +---- Thread: *Positional Value in the DSFL: A Study (/showthread.php?tid=22383)

*Positional Value in the DSFL: A Study - Modern_Duke - 05-23-2020 This idea took a few different forms over time, but the basic premise that I wanted to explore was to get a more mass-scale idea of the relative value of each position. How important is having a 250 TPE quarterback? Can a team get by with a low-earning DE? What is the difference in performance between an average player and a bot player? So how to best explore this? I could have just picked a team and messed with player attributes one at a time, but that would only give an answer for the value of each position specifically to that team. For example, testing for team performance with a 50 TPE running back may make less of difference to a team with maxed QB and WRs than it would to a team with a weak passing game. I decided that rather than working with a specific, existing team, I'll just make my own team. So how does one try to make a context-free team? Well, with the help of the TPE tracker, I took each player in the DSFL, and calculated the average attributes for each starting player at each position. So from each team I pulled the following: 1 QB (I just took the top TPE player, even if a team plays more than 1) 2 RB 3 WR 2 TE 1 OL (here is where I will mention, if a team did not have a "starting" player, I substituted a 50 tpe player to calculate the average) 2 DE 2 DT 2 LB (some teams do play other formations, but I just made enough for Nickel) 3 CB 2 S 1 K And here we have the average starting player at each position (as pulled after the 5/9 update) Now, after manually adjusting a roster to create this average team, it was time to start simming to get the baseline results. The average team posted a 56.43% win percentage and +2.613 average point differential over 2100 games at home against the Grey Ducks. Why Minnesota? Well, the original plan was to play the average team against itself, but somehow I never knew the sim doesn't let you do that. So I kind of randomly picked Minnesota to be the opponent for all tests. Before you say it, yes, that does spit in the face of that "context-free" stuff I was saying before. But, look, I didn't want to manually make a whole second team, alright? Alright, so anyway, after getting the results for the control group, its time to test out these positions. What I did was, for each position, in addition to the average players shown above, I made a player at 50 TPE (to represent a baseline, a replacement bot player) and one at the max 250 TPE. So for each round of testing, I replaced the average player at that position with the 50 tpe, and then with the 250 tpe player. For example, with the first test, I used the same team full of average players that got the 56.43% win percentage, but just replaced the Average Starting QB with a 50 TPE QB to find out how much impact that makes. Disclaimer: For most positions I only have second- or third- hand knowledge of how to best build a player. I tried my best? Quarterback Attributes used: Results: 50 TPE: 39.72% win percentage, -3.317 point differential Average (237 TPE): 56.43%, +2.613 250 TPE: 56.86%, +1.869 Well, nothing surprising there. A 50 TPE QB will really hurt your team. Just to note, as you can see there are a lot of maxed QBs in the DSFL this season, so the average was already close to 250, and obviously got similar results. But since the average and the 250 were so similar I threw in another test with a 150 TPE build to get a picture of a middle ground, and it got 55.71% and +0.749 Running Back Attributes used: Results: 50 TPE: 51.14% win percentage, -0.103 point differential Average (173 TPE): 56.43%, +2.613 250 TPE: 54.86%, +1.249 Ok so let me explain, I actually did all this work weeks ago, but held off on posting it because some of these results just don't make sense (oh just wait, this is nothing). As you can see, the team did worse with a maxed out running back than an average one. I'm still willing to write this off as normal fluctuation as a result of not running many sims (only 350 per), but still, this isn't what I was expecting to see. Wide Receiver Attributes used: Results: 50 TPE: 50.57% win percentage, +0.789 point differential Average (183 TPE): 56.43%, +2.613 250 TPE: 56.57%, +3.529 That's more like it. Oh and just to be clear here, this is showing the difference of only one WR out of the three in the lineup. So that 50.57% is with a 50 TPE WR starting at LWR. The RWR and Slot receiver are still the average 183 TPE. Tight End Attributes used: Results: 50 TPE: 63.14% win percentage, +3.783 point differential Average (166 TPE): 56.43%, +2.613 250 TPE: 55.43%, +3.229 ...um 50 TPE: 63.14% Excuse me, what the fuck? Did I input something incorrectly? Let me run that again... 50 TPE (re-do): 58.57%, +3.140 point differential *visible confusion* Do tight ends not matter in DSFL? I looked at the individual stats on this run, and like you would expect the 250 TPE TE got about double the catches and yards as the 50 TPE, and picked up more pancakes too. I guess in theory, a 50 TPE TE might see fewer targets than a 250 TPE TE, and if a 250 TPE TE is a worse receiving option than an average WR, fewer TE targets is actually good for the team? Offensive Line Attributes used: Results: 50 TPE: 57.43% win percentage, +2.029 point differential Average (127 TPE): 56.43%, +2.613 250 TPE: 55.43%, +2.629 More seemingly backwards results. At least the defensive position tests mostly make sense. Defensive End Attributes used: Results: 50 TPE: 52.29% win percentage, +0.760 point differential Average (143 TPE): 56.43%, +2.613 250 TPE: 60.86%, +3.206 I found this very interesting. Going into this, my guess was that Defensive End would be on the low end of positions that make a difference, but these results show otherwise. Defensive Tackle Attributes used: Results: 50 TPE: 53.14% win percentage, +1.494 point differential Average (137 TPE): 56.43%, +2.613 250 TPE: 60.00%, +3.911 Very similar to defensive end. Makes sense. Linebacker Attributes used: Results: 50 TPE: 49.43% win percentage, +0.717 point differential Average (170 TPE): 56.43%, +2.613 250 TPE: 57.71%, +2.494 From the looks of it, linebacker is one of the more important positions, or at least important to get better than a minimal earner. Cornerback Attributes used: Results: 50 TPE: 56.00% win percentage, +1.814 point differential Average (128 TPE): 56.43%, +2.613 250 TPE: 60.57%, +2.791 Cornerback performed well on the high end, but I was surprised to see the floor being pretty high. Safety Attributes used: Results: 50 TPE: 50.86% win percentage, +1.169 point differential Average (194 TPE): 56.43%, +2.613 250 TPE: 56.00%, +2.049 Very interesting to see the difference in cornerback and safety performance, which wasn't what I was expecting Kicker Attributes used: Results: 50 TPE: 58.86% win percentage, +2.160 point differential Average (194 TPE): 56.43%, +2.613 250 TPE: 58.57%, +3.123 Much like offensive line, not much conclusions to draw there. Which, based on what we know about kickers, may be right. And here's a nice unreadable graph for visualization So based on the data points we can come up with formulas to calculate win percentage based on TPE amount: QB: Win % With Average Teammates = 0.00081 x TPE + 0.38231 RB: Win % With Average Teammates = 0.00021 x TPE + 0.50854 WR: Win % With Average Teammates = 0.00032 x TPE + 0.49374 TE: Uh, I'll say, N/A, Insufficient Data OL: Minimal Effect or Insufficient Data DE: Win % With Average Teammates = 0.00043 x TPE + 0.50204 DT: Win % With Average Teammates = 0.00034 x TPE + 0.51546 LB: Win % With Average Teammates = 0.00043 x TPE + 0.47828 CB: Win % With Average Teammates = 0.00024 x TPE + 0.54267 S: Win % With Average Teammates = 0.00028 x TPE + 0.49774 K: Minimal Effect or Insufficient Data And from there, you know WAR, the baseball stat? Well we can do something similar here (well, no, but just bear with me) here to compare different TPE levels to a "replacement level player", i.e a 50 tpe bot, and multiplying by 14 to represent a full season. WAB (Wins Above Bot) for a Max 250 TPE Player: QB: 2.27 RB: 0.58 WR: 0.90 TE: Minimal? OL: Minimal? DE: 1.20 DT: 0.96 LB: 1.20 CB: 0.67 S: 0.79 K: Minimal? Going into this, in terms of positional value, I was anticipating something along the lines of: QB, RB, DT, LB, CB, WR, S, DE, TE, OL, K. So needless to say, I was surprised by the low impact of max RBs and CBs No but seriously, this was just for fun, there are so many variables at play here to make this whole study completely meaningless. Actually wait no, I recommend everyone other than the Royals accept this as fact.