International Simulation Football League
*ISFL Elo ratings - 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: *ISFL Elo ratings (/showthread.php?tid=33101)

*ISFL Elo ratings - Gragg9 - 06-26-2021

Apologies for typos. Against my better judgement I’m posting this near midnight before I go to bed.

League boomers may remember a media series about each team’s Elo rating from week to week. For y’all new people, Elo rating is a way to measure the relative skill of competitors. Here’s the short explanation: After each game teams gain or lose rating points based on the results (win/loss/tie) and the rating of their opponent. If a high rated team beats a low rated opponent, they only gain a modest amount of rating. The low rated team also loses a small amount or rating. This is because the high rated team is ‘expected’ to win. However, if the low rated team beats the high rated team, the rating gained and loss is more significant. Fun tidbit for you nerds, but ELO stands for Enot Lan Oacronym.

For a more detailed and probably more easily understood explanation, check out the wikipedia article https://en.wikipedia.org/wiki/Elo_rating_system or Ithica Hawks introduction to the Elo system https://forums.sim-football.com/showthread.php?tid=21466&pid=287518#pid287518

That’s also a good segue to why I’m talking about Elo. For most of the season I’ve been tracking ratings for every team in the ISFL. It’s been a fun way to track franchise success across seasons and with more detail than a simple win/loss record. Below is an image of a chart showing historical ratings back to the sim switch at the beginning of season 26. All teams started with 1200 Elo. Knowing the forum the image will be too small to be legible so hopefully this link will work and be interactive as well:

Link Here

[Image: Capture.PNG]

Gotta say it sucks to post that after how the last few games went for OCO, but more on that later.

Let’s take a brief closer look at these ratings!
  [Image: Capture.PNG]
Here’s a table view of the current, peak, and lowest rating for each team since the sim transfer. Note once again that 1200 is where each team started. The table and chart start after the week one games at the beginning of season 26 so anything before that doesn’t exist.

Anyway, we see that one team is currently at it’s peak rating. It also happens to be the number one ranked team headed into the playoffs according to Elo. Berlin has actually pulled out quite a lead in rating after Arizona, Sarasota, and Colorado had mediocre ends to their seasons. (don’t sue me, that’s what the data says). In fact, it’s the second highest Elo all-time, only 5 points behind Arizona’s peak earlier this season. With a win to start S30 they will have the new record.

Throughout the past 3 seasons (excluding the first few weeks) 7 teams have held the top spot.

ARI : 13 weeks
SAR : 12 weeks
CHI : 5 weeks
BER : 4 weeks
COL : 4 weeks
HON : 3 weeks
YKW : 1 week

Arizona and Sarasota are no surprise here as they’ve been the dominant team for as long as I’ve been in the league (maybe if I leave the league balance will be restored). Chicago may be a bit of a surprise given how the last couple seasons have gone. Crazy that at that time both them and Honolulu were battling for the best team. Berlin and Colorado both had their time in the spotlight this season. Yellowknife is a bit of a surprise with two championships during this time and only one week (tied) for top Elo.

Looking at the bottom teams is a little less interesting. Big F for Baltimore. Look forward to seeing this chart in a few seasons tho <3

Now let’s look at each team individually. Starting with the highest currently rated team and moving down.

[Image: Capture.PNG]
Rating: 1425

Gotta be optimistic if you’re Berlin looking at this. One game away from the new all-time rating (playoff games won’t count) and no real ‘bad’ seasons in the mix. Even if the have a rough start to the next season they may hold on to the number one spot for a few games due to their significant lead over the current number 2 Colorado.

[Image: Capture.PNG]
Rating: 1360

Oh lordie, yellow Yeti does not work for this. Sorry ya’ll. Colorado appears to be slightly more consistent in their rating than Berlin but wasn’t quite able to finish this season out as strong as they would have liked. Still plenty to be happy about here.

[Image: Capture.PNG]
Rating: 1341

While they only clock in 3rd to end this season, AZ has been *the* most consistently dominant team since the sim switch. They've been top Elo in each of the 3 seasons and their lead at the beginning of the season was record setting and left the rest of the league in the dust. Though as they say the bigger they are the harder they fall. Which is exactly how Elo rating works. They were brought back down to the earth after losing several games they should have one. Ratings are unforgiving like that.


[Image: Capture.PNG]
Rating: 1333

The only other team that can argue their case against Arizona in this time frame is Sarasota. If you look at playoffs then their ultimus trophy could do the talking, but we’re looking at regular season dominance. The story here might be very different were it not for the losing streak at the end of last season. They did hold the lead for one week this season and only by a single point before being knocked down by OCO. I feel that is very important and worth mentioning in this brief analysis.

[Image: Capture.PNG]
Rating: 1327

Yellowknife is the king of consistency. That’s how they’ve brought home multiple championships in this time frame. Others may like the big peaks of Arizona or Berlin but this is what impresses me. Unfortunately for Yellowknife they are the highest rated team left out of the playoffs this season. Their only lead during this timeframe was a single week, week 9, and there was very nearly a four way tie.

[Image: Capture.PNG]
Rating: 1267

Yuck, another color that really doesn’t work. Oh well, bear with me. At least according to rating NYS is a significant jump down from the teams already talked about. In this time frame they’ve been all over the league standings. There’s been a loose general trend upwards though and it’s good enough to secure them a playoff spot. This is also the first team listed that has never held the top spot.

[Image: Capture.PNG]
Rating: 1267

Actually tied with NYS is NOLA. They must like drama because somehow they drop to the bottom and then turn it around and make a deep playoff run every season. They’re one of two teams to hold the dubious honor of second-to-last every season in this time frame. In that time they have 2 playoff appearances including a remarkable come from behind qualification this season.  Both NYS and NOLA hold the...honor of being the lowest rated team in the playoffs.

[Image: Capture.PNG]
Rating: 1178

Pain.

[Image: Capture.PNG]
Rating: 1157

Austin is one of the most intriguing teams in this time frame. Clearly second worst the first season, ultimus appearance last season, and back to the basement this season. They did end the season to hop up the rankings quite a bit. Goes to show how little there is between the teams, particularly towards the bottom.

[Image: unknown.png]
Rating: 1137

Chicago is another really interesting team but in kinda a macabre way. They’re one of two teams to make it to the top and then fall all the way to the bottom. Last season especially was historically bad and is certainly in the conversation for one of the most sim screwed seasons of all time. I believe this season was largely fallout from that. I believe Chicago will be able to turn it around though. Question is how quickly and will it be enough to win championships?


[Image: Capture.PNG]
Rating: 1126

It surprised me to see how similar Chicago’s and Honolulu’s trajectory has been. The big difference is they had a lot of success at the end of last season. Since then they’ve had one of the biggest single season drops. Honolulu really likes their winning and losing streaks it seems.

[Image: Capture.PNG]
Rating: 1126

Philly has been surprisingly consistent but not in a great way. They are a young team though so they’ll be looking for a breakout season soon. Many seemed to think this season would be that breakout but clearly that didn’t quite come together. They had some big wins though so we’ll see if they can break out the basement next season.

[Image: Capture.PNG]
Rating: 1117

Well damn I thought the end of my season hurt. SJS quickly went from playoff contender to (neaer) the bottom of the pile. Admittedly only a few points separate 5 teams here but still. I have no analysis to offer here but I can offer my sympathy. Not too much sympathy cause I’m in the conference with y’all, but a little sympathy.

[Image: Capture.PNG]
Rating: 959

Finally the team that’s in a league of it’s own. Baltimore knows that I have faith in their rebuild, but let’s talk a little about how remarkably low their rating has been.

Their peak rating was week 1. Because I artificially started everyone at 1200. Making them the only team to ever peak with a loss.

Their lowest rating is 158 points below the next lowest

The distance between the current top team and SJS is nearly as much as the Baltimore and next worst team at the time of Baltimore's lowest rating. Given how Elo is designed that’s insane.

Towards the end of their record setting losing streak they were losing 3 points per game because the outcome was what the rating system expected. In their first win they gained 47, near the theoretical maximum.

Sorry to y’all Baltimore peeps out there. You know I love you.




Wrapping up then! If you have any thoughts or advice on this please let me know. I might use this rating system for...something...later so I want to make sure it’s solid. For you nerds out there I have the K factor set at 50, which is quite high. I think it works though but I could be convinced otherwise. Also if you’re a tableau pro I could use some tips on the data viz (obviously).


Anyway, thanks for reading and thanks in advance for your thoughts and feedback!

RE: ISFL Elo ratings - The_Kidd - 06-27-2021

I can't remember if this accounts for opponent strength like the SRS metric in basketball? Because some teams seem to get punished or rewarded too much for "good losses" or "empty wins" --- but most ratings seem on par for the trends and strengths of the teams.

One ask though: on the sliding scale legend --- can the week and season be added to the elo rating? So it would show like: "AZ S27 W11 ELO 1107"

RE: ISFL Elo ratings - Modern_Duke - 06-27-2021

Well done.

I found a K of about 35 to be pretty good. Also I know FiveThirtyEight does some adjustment for each team regressing back towards the mean each offseason, or something like that. Doesn’t look like you did, but not even sure if it’s necessary.

RE: ISFL Elo ratings - Gragg9 - 06-27-2021

(06-27-2021, 01:29 AM)The_Kidd Wrote: I can't remember if this accounts for opponent strength like the SRS metric in basketball? Because some teams seem to get punished or rewarded too much for "good losses" or "empty wins" --- but most ratings seem on par for the trends and strengths of the teams.

One ask though: on the sliding scale legend --- can the week and season be added to the elo rating? So it would show like: "AZ S27 W11 ELO 1107"

It does account for opponent. I could tune down how much rating changes from game to game however

I could probably do that for the legend with some time and learning. If that viz is something I keep updated I'll look into it.

RE: ISFL Elo ratings - Gragg9 - 06-27-2021

(06-27-2021, 06:21 AM)Modern_Duke Wrote: Well done.

I found a K of about 35 to be pretty good. Also I know FiveThirtyEight does some adjustment for each team regressing back towards the mean each offseason, or something like that. Doesn’t look like you did, but not even sure if it’s necessary.

I didn't do any of the fancy adjustments that 538 did *yet* in the future I likely will

RE: ISFL Elo ratings - Gragg9 - 06-27-2021

[Image: Capture.PNG]

Here's a K-factor of 35 for reference

Below is what I had been using, a K-factor of 50
[Image: Capture.PNG]

RE: ISFL Elo ratings - 37thchamber - 06-29-2021

Yeah, I would say a K of 50 is far too high personally (especially if you're starting at 1200) and is definitely part of the reason you see massive fluctuations. I don't think I've ever implemented Elo with a K above 32 and base rating of 1450. I had an implementation that adjusted K in bands like FIDE do as well, and increased K for playoff games, which worked okay.

There was an attempt to adjust rating change based on margin of victory but it led to some wild changes and I couldn't balance it... also had one where I tried to tweak based on normalising pt differential based on strength of schedule, but that didnt work because id either have to create a moving range of games to calculate SOS from, or the formula would be basically broken for half of every season.

Elo is a good framework for an objective rating system though. I think it would probably work best for calculating power rankings here, because it would give you an easy way to gauge form as well (just sum rating change over a fixed period), meaning you wouldn't necessarily have to rank based entirely on elo rating, but you could incorporate delta into it and rate a team at 1200 who just jumped 35 pts in their last game higher than a team at 1210 who dropped 25 to end up there on the grounds that the first team is "hotter" right now.

All-time Elo would be very interesting, but a ton of work to put together.