Warrior General Games

Here's an estimate of the numbers behind the ladder ranks that GLB uses. There are some differences (ie Seahawks) this is due to the following reasons:

1. k-factor of 32 was used
2. Playoff games were assumed to count; more accurate rankings if they do imo
3. Midseason teams dropped to 1082.933 (note midseason teams are anyone with the id between 255 and 326)
4. All new teams start at 1200 and teams that dropped to rookie were reset to 1200 on the Jan 31.

This is just the top 50, got the rest if anyone is interested. As you can see Djibouti and Maine are a decent way ahead in 1st and 2nd.

Rank Team Elo Win Loss Tie
Doinitin Djibouti 1486.513534 31 4 0
Maine Snails 1483.303913 31 3 1
OG GLB 1428.244895 26 9 0
Rocky Mountain Reapers 1427.025754 26 8 0
Seattle Seahawks 1422.660088 27 8 0
Portland Ronin 1421.709898 25 7 3
Sunday Funday 1420.818009 26 8 1
Alpine 1405.58674 25 9 0
TT's Wolfpack of One 1402.407923 25 9 1
Queen City Brewmasters 1388.48555 25 9 0
Legacy 1381.47553 25 10 0
Hub City Hustlers 1379.396314 25 9 1
Points at Dick 1373.896658 24 9 0
Iowa Rogues 1369.261747 24 3 0
New York Nightmare 1368.828623 25 9 1
Chicago Bruins 1367.574056 23 10 1
Somalia Pirates 1366.548886 24 10 1
Jebediahs Oblong Ball Club 1364.401386 23 10 1
Spokane Fighters 1359.81009 25 6 1
Annapolis Freedom 1348.32604 24 10 0
Black Ice en Fuego 1347.53203 22 11 1
Taylor Gang 1344.138722 22 11 1
Zone Mage Wangbang 1342.998315 23 11 1
Carolina Pirates 1341.163306 24 10 1
Kashyyk Wookies 1327.425184 22 12 0
Ball So Hard University 1324.17396 21 11 2
Portland Cardinals 1323.77051 19 10 3
Zorin World Order 1322.40654 23 12 0
San Diego Longhorns 1319.852058 21 12 1
New Jersey Generals 1318.172753 20 11 2
Bloodless Brotherhood 1312.940968 22 11 1
Santiago Benitos 1310.050705 20 12 1
Tucson Reptiles 1304.450695 20 13 2
Gurupie Nation Presents: Beavers 1304.298428 22 13 0
Mongolian Horde 1302.104412 20 12 1
Sunnyvale Samsquanches 1301.684095 20 12 1
South Otto Suppressors 1299.169907 19 11 1
A Good Day To Die 1297.94759 21 12 2
Philadelphia Devils 1295.256497 22 13 0
Minnesota Reapers 1293.675686 21 12 2
Northwest Storms 1293.403658 21 12 1
Las Vegas Outlaws 1292.932156 20 12 1
Albuquerque Pack 1287.238936 22 9 1
Cobden Muskrats 1286.289683 19 12 2
Undead Xombi 1286.276948 22 13 0
Boulder Blitz II 1283.889118 19 13 1
Monmouth EVIL MONKEYS 1283.555455 18 12 3
Toledo Fighters 1282.177338 19 10 1
Frontier Cougars 1279.543691 18 12 2

So calculated with mid season teams (255 to 303) having a starting elo of 1082 and iowa drops to 20 which is pretty close. Only difference now is I'm factoring play off games in. Also no clue what the late season 1 teams started at (team 303 onwards).

Nice to See Erkner should be Way higher

Originally posted by Thunderoo
1. k-factor of 32 was used
2. Playoff games were assumed to count; more accurate rankings if they do imo
3. Midseason teams were assumed to start at 1200 like everyone else. This is false as they all started with elo of the worst team in GLB at the time (wtf were you guys thinking...)

One out of three assumptions isn't bad.

realised different k-factor shouldn't have a huge impact on rankings. also just realised I forgot about team resets. dunno how I'd get that in there unless I had a list of teams to reset to 1200 at a particular point.

Doinitin Djibouti 1482.455055 30 4 0
Maine Snails 1477.025943 30 3 1
Engage With Zorp! 1437.834778 25 6 3
Seattle Seahawks 1419.325719 26 8 0
Rocky Mountain Reapers 1418.174778 25 8 0
OG GLB 1417.709049 25 9 0
Sunday Funday 1413.49512 25 8 1
Portland Ronin 1411.202177 24 7 3
Alpine 1401.081242 24 9 0
TT's Wolfpack of One 1394.25685 24 9 1
Queen City Brewmasters 1375.234926 24 9 0
Points at Dick 1373.896658 24 9 0
Legacy 1370.826871 24 10 0
Hub City Hustlers 1369.753919 24 9 1
Spokane Fighters 1368.698011 24 6 1
Chicago Bruins 1367.574056 23 10 1
New York Nightmare 1364.895644 24 9 1
Jebediahs Oblong Ball Club 1364.401386 23 10 1
Iowa Rogues 1360.571778 23 3 0
Somalia Pirates 1351.293767 23 10 1
Black Ice en Fuego 1347.53203 22 11 1
Hawaii Bulls 1346.567583 23 10 0
Zorin World Order 1345.453681 23 11 0
Annapolis Freedom 1339.932018 23 10 0
Dryden Blue Balls 1337.557379 20 10 2
Taylor Gang 1336.707715 21 11 1
San Diego Longhorns 1335.107177 21 11 1
Zone Mage Wangbang 1334.45504 22 11 1
Bloodless Brotherhood 1332.399736 22 10 1
Carolina Pirates 1329.450486 23 10 1
Ball So Hard University 1324.17396 21 11 2
Gurupie Nation Presents: Beavers 1324.094575 22 12 0
Portland Cardinals 1323.77051 19 10 3
Albuquerque Pack 1319.239036 22 8 1
Kashyyk Wookies 1317.756177 21 12 0
Philadelphia Devils 1315.835979 22 12 0
Tampa Ducks 1315.002794 20 10 1
Sunnyvale Samsquanches 1314.934718 20 11 1
Eugene Preachers 1314.805557 13 4 0
New Jersey Generals 1308.265479 19 11 2
Undead Xombi 1307.028152 22 12 0
Mongolian Horde 1302.104412 20 12 1
South Otto Suppressors 1299.169907 19 11 1
Steiner Vipers 1296.236071 20 12 1
Santiago Benitos 1294.859085 19 12 1
Hillfield Dinos 1294.550949 18 13 1
Monmouth EVIL MONKEYS 1294.091301 18 11 3
Northwest Storms 1293.403658 21 12 1
Tucson Reptiles 1292.821062 19 13 2

Rankings where mid season team were set at lowest elo (1082.933)

Nice work Thunderoo!

Teams that went back to initial rating are easy to spot for now - they are in Rookie leagues currently. Where did you get that 1200 was the initial rating? Not that it matters what is used, unless you implement a ratings floor.

Glad to see someone pulling this together.

I'm *pretty* sure playoff games counted, I thought I saw rankings changing on playoff days.

k-factor will only matter when they added in the midseason teams and when they added the S2 sophomore and rookie teams. I don't think we know what they did to those teams ratings, do we?

Also, midseason teams started *below* the worst elo, didn't they? They weren't even in a tie with the worst team IIRC.

Originally posted by Corndog
One out of three assumptions isn't bad.

When you don't publish the ratings, don't be surprised at people attempting to back-calculate them. When you get them wrong, don't be surprised at people making a more accurate version.

I would love to have a copy of the full ratings from the OP rankings, with midseason teams getting the right initial rating instead of the broken one they used. As you show, it makes a big difference in the appropriate rating for the mid-season teams like Iowa, and now that I'm in Osprey I'd love for them to see what they *should* be at.

And now Alpine will be getting affected this season because our league will be much lower rated than it should be - going to be a harder climb up the ladder and less room for mistakes.

Also, would you consider sharing the game data with me? Collecting that data is what kept me from playing around with these numbers myself.

Originally posted by NiborRis
When you get them wrong, don't be surprised at people making a more accurate version.

lol

I like you man, but that tinfoil hat is starting to show some wear.

But no, I'm not posting the code.

The ladder system as it is makes 0 sense and is not sustainable

What happens when your inserted as a rookie at the default and there are 6 levels of teams, each for the most part being lopsided - the effect only getting worse and unplayable if you try to play past 2 or more tiers

Bad dot midlevel teams never place properly because they inflate off rookie teams, the effect only getting worse the older the team

Maybe ironically, the pyramid we all hated so much worked better as a loose ladder concept over time

Even if you artificially modify elo changes postgame by the level of teams involved, A 2/6th chance that the team you get matched with is competition for you produces an insane amount of pointless games that change rating. Over 1000 games? lol

Originally posted by Laggo
The ladder system as it is makes 0 sense and is not sustainable

What happens when your inserted as a rookie at the default and there are 6 levels of teams, each for the most part being lopsided - the effect only getting worse and unplayable if you try to play past 2 or more tiers

Bad dot midlevel teams never place properly because they inflate off rookie teams, the effect only getting worse the older the team

Maybe ironically, the pyramid we all hated so much worked better as a loose ladder concept over time

Even if you artificially modify elo changes postgame by the level of teams involved, A 2/6th chance that the team you get matched with is competition for you produces an insane amount of pointless games that change rating. Over 1000 games? lol

Yeah.

What happened to the idea of keeping the ladder seperated by age groups? I thought that idea had a lot of traction.

I would be in favour of the ladder merging later, maybe ages 15 to 25, but you shouldn't play teams out of your age between ages 1 and 14.

You cant argue that you screwed the midseason Team's Rankings

Trying to understand this post, lost it when AGD2D wasn't on here.

Originally posted by Corndog
lol

I like you man, but that tinfoil hat is starting to show some wear.

But no, I'm not posting the code.

I don't really want the code; the current rating of each team posted to the team page and the rankings page would be plenty. Those who cared about details would be able to figure them out with almost no effort from there.

Based on the results that Thunderoo has, I'm pretty sure the only thing you've gotten really *wrong* is how to introduce new teams to the ratings, but that's pretty important. I would agree with Thunderoo's assessment that Iowa should be ranked in the top 5 instead of near 20th, for example - that's a big difference, and at the end of season matters for team trophies and achievements. Iowa went 21-3 last season and didn't sniff the top 10. But you guys have been pretty secretive about how you did it, and I wasn't willing to put the work in to gather all the game data like Thunderoo did and model up what you actually did, so I could only have theories based on the evidence available.

Eventually, elo works itself out regardless of starting rating, but "eventually" can be 100-200 rated games, which is 3-7 seasons...which is too long.