Tk3 Posted March 26, 2021 Share Posted March 26, 2021 (edited) 11 hours ago, El Ramster said: @Tk3 could you explain your break down to your rankings? How Our NFL Predictions Work | FiveThirtyEight basically there is a formula based on current ratio adjusted for HFA to determine who is most likely to win the game, and by how much then the actual results are compared to the expected results and you gain or lose ELO points based on that Edited March 26, 2021 by Tk3 Link to comment Share on other sites More sharing options...
RuskieTitan Posted March 26, 2021 Share Posted March 26, 2021 2 hours ago, Tk3 said: How Our NFL Predictions Work | FiveThirtyEight basically there is a formula based on current ratio adjusted for HFA to determine who is most likely to win the game, and by how much then the actual results are compared to the expected results and you gain or lose ELO points based on that 2 Link to comment Share on other sites More sharing options...
Malfatron Posted March 26, 2021 Share Posted March 26, 2021 2 hours ago, Tk3 said: How Our NFL Predictions Work | FiveThirtyEight basically there is a formula based on current ratio adjusted for HFA to determine who is most likely to win the game, and by how much then the actual results are compared to the expected results and you gain or lose ELO points based on that Right but what are your variables? Link to comment Share on other sites More sharing options...
EaglesPeteC Posted March 26, 2021 Share Posted March 26, 2021 2 hours ago, Tk3 said: How Our NFL Predictions Work | FiveThirtyEight basically there is a formula based on current ratio adjusted for HFA to determine who is most likely to win the game, and by how much then the actual results are compared to the expected results and you gain or lose ELO points based on that Link to comment Share on other sites More sharing options...
Tk3 Posted March 26, 2021 Share Posted March 26, 2021 2 minutes ago, Malfatron said: Right but what are your variables? essentially the same as what is in that link. I basically just copied and pasted the formulas and then linked in a spreadsheet to make inputs easy. slight different for HFA based on what nacho said but I ran it starting in S20 with everyone starting at 1500, and the final results were pretty reasonably similar to the final standings/Point differentials. Enough that I liked it. Then it was normalized to the mean to semi account for roster reconstruction - now we pick up for S21 Link to comment Share on other sites More sharing options...
swoosh Posted March 26, 2021 Share Posted March 26, 2021 57 minutes ago, Tk3 said: essentially the same as what is in that link. I basically just copied and pasted the formulas and then linked in a spreadsheet to make inputs easy. slight different for HFA based on what nacho said but I ran it starting in S20 with everyone starting at 1500, and the final results were pretty reasonably similar to the final standings/Point differentials. Enough that I liked it. Then it was normalized to the mean to semi account for roster reconstruction - now we pick up for S21 At the end of this season, S20 results will still play a role in the results, right? Link to comment Share on other sites More sharing options...
Tk3 Posted March 26, 2021 Share Posted March 26, 2021 41 minutes ago, swoosh said: At the end of this season, S20 results will still play a role in the results, right? correct.. but the older it gets the less influence it has. For one thing every season the final score normalizes towards the mean (1500), which means the influence disappears year after year. plus the fact that you were so low after S20 means every game you play this season is more "favorable" to you as far as moving since you are an underdog. If everyone started at 1500 this year your relative movement would have been much smaller for wk 1 If you look at the fivethirtyeight model you can see that by the end of the season it basically reflects the season and you can hardly even know there was previous history influence 2020 NFL Predictions | FiveThirtyEight it's mostly for the benefit of the early part of the season and not being overreactionary to the results of the early weeks Link to comment Share on other sites More sharing options...
swoosh Posted March 26, 2021 Share Posted March 26, 2021 (edited) 17 minutes ago, Tk3 said: correct.. but the older it gets the less influence it has. For one thing every season the final score normalizes towards the mean (1500), which means the influence disappears year after year. plus the fact that you were so low after S20 means every game you play this season is more "favorable" to you as far as moving since you are an underdog. If everyone started at 1500 this year your relative movement would have been much smaller for wk 1 If you look at the fivethirtyeight model you can see that by the end of the season it basically reflects the season and you can hardly even know there was previous history influence 2020 NFL Predictions | FiveThirtyEight it's mostly for the benefit of the early part of the season and not being overreactionary to the results of the early weeks I see that's what you were going for, but the early weeks is where it's going to be the most inaccurate. It's essentially a S20 rankings at this point. It'll improve over time, but S20 will always be factored in. It's kind of like college football rankings to start the season, but with even more roster turnover. Instead of bringing back 3/4 of your roster, you're bringing back 1/4 of your roster. Roster turnover aside, you're also dealing with new player ratings because it's an odd-numbered season. I don't mean to bash your work, I just think it has a lot of flaws in something like this where teams are constructed differently year after year. And I always enjoy the effort people put into things like this. I should spend more time building something of my own than bashing something else. I think it would be interesting to see this same model, but start everyone at 1500. To combat small sample sizes early on, the rankings wouldn't be released until week 4, sort of like how it's done in college football. Edited March 26, 2021 by swoosh Link to comment Share on other sites More sharing options...
swoosh Posted March 26, 2021 Share Posted March 26, 2021 (edited) double post Edited March 26, 2021 by swoosh Link to comment Share on other sites More sharing options...
Tk3 Posted March 26, 2021 Share Posted March 26, 2021 2 minutes ago, swoosh said: I see that's what you were going for, but the early weeks is where it's going to be the most inaccurate. It's essentially a S20 rankings at this point. It'll improve over time, but S20 will always be factored in. It's kind of like college football rankings to start the season, but with even more roster turnover. Instead of bringing back 3/4 of your roster, you're bringing back 1/4 of your roster. Roster turnover aside, you're also dealing with new player ratings because it's an odd-numbered season. I don't mean to bash your work, I just think it has a lot of flaws in something like this where teams are constructed differently year after year. And I always enjoy the effort people put into things like this. I should spend more time building something of my own than bashing something else. I think it would be interesting to see this same model, but start everyone at 1500. To combat small sample sizes early on, the rankings wouldn't be released until week 4, sort of like how it's done in college football. By the time we are a quarter of the way through the season it will look a lot more like S21. If a 2 win team has a great offseason then in week 1 they beat a 9 win team, would you expect them to be in the top half of the league after week 1? No, you would expect them to have a bit more success before they cracked the top end of team. I guess its a snarky response, but don't be one of the worst teams of all time and expect to be rated highly right away at the start of the season. I get that there's more roster reconstruction here than there is in the real NFL, but you're still going to have to climb Link to comment Share on other sites More sharing options...
Tk3 Posted March 26, 2021 Share Posted March 26, 2021 The pwny for those who want it. @pwny using 5.3 is correct? Owner Team PWNY Wins Loss Win % Net PF PA OR DR LUCK SirA Freiburg Venom 15.131 1 0 1.000 10 24 14 10 1 0.054 TK3 Reykjavík Direwolves 14.845 1 0 1.000 13 34 21 2 7 0.072 Rammy Egypt Starfalls 14.572 1 0 1.000 11 31 20 6 4 0.089 Glem Little Rock Uni Royals 13.784 1 0 1.000 7 24 17 10 2 0.139 Swoosh Richmond Flying Squirrels 13.503 1 0 1.000 9 33 24 5 9 0.156 Ruskie Rio de Janeiro Pirates 12.890 1 0 1.000 8 34 26 2 13 0.194 Counselor Anchorage Moose 12.811 1 0 1.000 6 26 20 7 4 0.199 Malfatron Norway Knights 11.247 1 0 1.000 6 40 34 1 17 0.297 Khodder Tacoma Thunder 11.247 1 0 1.000 3 20 17 15 2 0.297 Remixxx Scranton Papermakers 9.739 1 0 1.000 2 25 23 9 8 0.391 RMIB Indianapolis Predators 6.261 0 1 0.000 -2 23 25 13 12 -0.391 Bucsfan Phoenix Rattlers 4.753 0 1 0.000 -6 34 40 2 20 -0.297 Rackcs Chicago Fire 4.753 0 1 0.000 -3 17 20 18 4 -0.297 Josh Lancaster Fighting Amish 3.189 0 1 0.000 -6 20 26 15 13 -0.199 Ravens Hanoi Viet Kongs 3.110 0 1 0.000 -8 26 34 7 17 -0.194 YoungBucs Miami Sharks 2.497 0 1 0.000 -9 24 33 10 16 -0.156 Whickok Pittsburgh Punishers 2.216 0 1 0.000 -7 17 24 18 9 -0.139 EaglesPete Rocket City Trash Pandas 1.428 0 1 0.000 -11 20 31 15 15 -0.089 Mookie South Africa Woolly Mammoths 1.155 0 1 0.000 -13 21 34 14 17 -0.072 The Orca Antarctica Penguin Poppers 0.869 0 1 0.000 -10 14 24 20 9 -0.054 4 Link to comment Share on other sites More sharing options...
pwny Posted March 26, 2021 Share Posted March 26, 2021 3 minutes ago, Tk3 said: @pwny using 5.3 is correct? Yup. If you want, when I have some free time over the next week-ish, I can move all my data to a google doc and give you write access to add this season to it. It includes playoff data for every season as well. Link to comment Share on other sites More sharing options...
swoosh Posted March 26, 2021 Share Posted March 26, 2021 57 minutes ago, Tk3 said: By the time we are a quarter of the way through the season it will look a lot more like S21. If a 2 win team has a great offseason then in week 1 they beat a 9 win team, would you expect them to be in the top half of the league after week 1? No, you would expect them to have a bit more success before they cracked the top end of team. I guess its a snarky response, but don't be one of the worst teams of all time and expect to be rated highly right away at the start of the season. I get that there's more roster reconstruction here than there is in the real NFL, but you're still going to have to climb This response is so incredibly flawed. You're better than this. Rise above it. Link to comment Share on other sites More sharing options...
Tk3 Posted March 26, 2021 Share Posted March 26, 2021 13 minutes ago, swoosh said: This response is so incredibly flawed. You're better than this. Rise above it. idk what you want from me. I'm borrowing from a very highly regarded ranking system that is used for many, many sports sry, bro. win some games and move up the list Link to comment Share on other sites More sharing options...
swoosh Posted March 26, 2021 Share Posted March 26, 2021 1 hour ago, Tk3 said: idk what you want from me. I'm borrowing from a very highly regarded ranking system that is used for many, many sports sry, bro. win some games and move up the list This is unlike any other league though. You're trying to fit a square peg into a round hole. Why should my S20 squad play any role (let alone the majority) in the rankings this year? Did you see my squad last year? A college team would've beat it. Stick to running pwny's. Link to comment Share on other sites More sharing options...
Recommended Posts