Experts Tournament ELO

Finlandbuhanisson

Longswordman
May 29, 2015
374
947
108
#51
I might begin by first testing what the actual within-series correlation is, and maybe compare it to within-day correlation in normal rated games, to see if there even is a real problem there. To me it seems a desirable thing that a way longer series brings at least a bit more points than the shorter one.
 

Germanymungo__

Known Member
Dec 13, 2018
36
100
48
aoe-elo.com
Voobly
mungo_
View profile
Ladder
RM - 1v1
Rating
2665
Wins
2017
Losses
1584
Streak
4
#52
I might begin by first testing what the actual within-series correlation is, and maybe compare it to within-day correlation in normal rated games, to see if there even is a real problem there. To me it seems a desirable thing that a way longer series brings at least a bit more points than the shorter one.
If you want to play with the data and make something like a statistical model for series, you can have a clone of the match data.

I attached a reduced (secure) clone of the database as an SQL file. You can install any SQL client and import it. Let me know if you want a different format.

I will soon implement a Rest API so that data can be fetched live from the server.
 

Attachments

Likes: linetyReborn

Germanymungo__

Known Member
Dec 13, 2018
36
100
48
aoe-elo.com
Voobly
mungo_
View profile
Ladder
RM - 1v1
Rating
2665
Wins
2017
Losses
1584
Streak
4
#53
Hello again!

we finished a new version of the site:
  • Much more data: already 65 tourneys (shoutout to pumukel!)
  • Layout tweaks (tournament view, "upcoming events", ...)
  • Better mobile integration
  • Hopefully clearer series details:
    • the +/- in points to each game within a series are now displayed
    • A special note when somebody wins a series, but still loses points (for example losing 3:4 against Viper gives points to the loser if he has low Elo)
  • Some more stats ("all time highs", in the future hopefully containing "legendary players", visual Elo distribution, etc.
Let me know what you think, and what you yould like to see :smile:

Especially I would be happy if you tell me when anything is not intuitive/confusing.

mungo
 

United Statespatao

Well Known Pikeman
Apr 2, 2016
399
320
78
#54
This is great @mungo__ ! I am going to work on a new version of aoe2calendar.com soon, I'd love to collaborate in terms of integrating your ELO. Also maybe we can unify the data source for upcoming events.

Or potentially just surface a calendar view of your site on the aoe2calendar.com domain, if you want
 

NetherlandsQulala

Active Member
Jan 20, 2016
159
88
43
27
#58
it looks awesome, but i have a small suggestion to add the current normal rated ladder to the site aswell so you can compare them both in one screen instead of switching back and forth to voobly
 

Germanymungo__

Known Member
Dec 13, 2018
36
100
48
aoe-elo.com
Voobly
mungo_
View profile
Ladder
RM - 1v1
Rating
2665
Wins
2017
Losses
1584
Streak
4
#59
it looks awesome, but i have a small suggestion to add the current normal rated ladder to the site aswell so you can compare them both in one screen instead of switching back and forth to voobly
Good idea! I will check out the voobly API (and maybe email them) this weekend.
 

MoroccolinetyReborn

Well Known Pikeman
Oct 27, 2014
260
307
78
28
#61
amazing work , love the whole website layout and the graphs look amazing , only thing i noticed is that some tournament names are too close to each other that i can't read some of them , so would it be possible to have under the graph like an index list of<< tournament abreviation : full tournament name >> ?
 
Last edited:
Dec 11, 2012
2,973
1,040
128
www.youtube.com
#62
only thing i noticed is that some tournament names are too close to each other that i can't read some of them , so would it be possible to have under the graph like an index list of<< tournament abreviation : full tournament name >> ?
I agree this could be optimized. as an alternative for now, you can hover over the dots in the graph, then you get an info box including the full tournament name.
 

MoroccolinetyReborn

Well Known Pikeman
Oct 27, 2014
260
307
78
28
#63
one more thing i noticed is that all players in ur data start with a 2k rate so fro example first game between laaaan and mbl they both had 50% chance win rate which is just wrong , maybe it would be better to have the starting rate for players in your data to be their actual voobly rate of that the match date
 
Likes: Michaerbse

ArgentinaKantos

Active Member
Apr 28, 2017
22
35
28
23
Argentina
#64
The work you are doing is excellent, there is a lot of important information in only one site. Just one question: is it possible that you could do the same whit tournaments that happened before 2013? It would be awesome to see the tournament ELO comparison between the "current" pro players and the old ones
 
Mar 10, 2018
30
44
23
#65
one more thing i noticed is that all players in ur data start with a 2k rate so fro example first game between laaaan and mbl they both had 50% chance win rate which is just wrong , maybe it would be better to have the starting rate for players in your data to be their actual voobly rate of that the match date
The question is if that is a better estimate, given the fluctuation the Voobly elo shows, but it might be.
I think as long you have enough data the starting elo shouldn't really matter, but the question is if the database is large enough.

A thought I had, that goes in a similar direction: What would happen if you loop the elo, putting the resulting elo as the starting elo and repeat until it meets some sort of convergence criterion to fake a bigger database, I have no idea what that would really do, the idea just seemed interesting to me.
 

Netherlandsnimanoe

Knight
Bronze Supporter
Jan 15, 2014
2,210
1,637
138
23
#66
The work you are doing is excellent, there is a lot of important information in only one site. Just one question: is it possible that you could do the same whit tournaments that happened before 2013? It would be awesome to see the tournament ELO comparison between the "current" pro players and the old ones
I agree it would be great to see, but it would be pretty hard, since it's been such a long time that it's harder to decide which matches should be considered and which matches should be omitted. On top of that it's harder to find the brackets of each tournament as well. Also I think because the tournament settings were very different (mostly Bo1/Bo3 on random land map etc.) the ratings will be less comparable

Some typo's I found: on the homepage> Advantages of ELO it says op-to-date, which should be up-to-date.
Under tournaments: Nili's Appartment Cup 1 & 2 should be changed to Nili's Apartment Cup 1 & 2
WSVG Asia and Europe Qualifiers should be part of WSVG 2015 and I think it would be nice to group all the Top Expert Leagues together as well
Edit: Also I prefer (Grand) Final over (Grand) Finale
 

United StatesIamDiscern

Known Member
Oct 9, 2017
82
158
48
34
twitch.tv
Voobly
Discern___
View profile
Ladder
RM - 1v1
Rating
2665
Wins
2017
Losses
1584
Streak
4
#67
I agree it would be great to see, but it would be pretty hard, since it's been such a long time that it's harder to decide which matches should be considered and which matches should be omitted. On top of that it's harder to find the brackets of each tournament as well. Also I think because the tournament settings were very different (mostly Bo1/Bo3 on random land map etc.) the ratings will be less comparable

Some typo's I found: on the homepage> Advantages of ELO it says op-to-date, which should be up-to-date.
Under tournaments: Nili's Appartment Cup 1 & 2 should be changed to Nili's Apartment Cup 1 & 2
WSVG Asia and Europe Qualifiers should be part of WSVG 2015 and I think it would be nice to group all the Top Expert Leagues together as well
Edit: Also I prefer (Grand) Final over (Grand) Finale

While we're at it: Miguel is part of Heresy now. So rip team statistics! :frown:
 
Likes: nimanoe

Germanymungo__

Known Member
Dec 13, 2018
36
100
48
aoe-elo.com
Voobly
mungo_
View profile
Ladder
RM - 1v1
Rating
2665
Wins
2017
Losses
1584
Streak
4
#68
Hey guys, thanks for the input, I will work in the things you mentioned on the weekend.

I added some of the things you said (include Voobly rating, more readable graphs) to the top thread, just to keep it a bit organized. I cant keep it in my head otherwise 11

one more thing i noticed is that all players in ur data start with a 2k rate so fro example first game between laaaan and mbl they both had 50% chance win rate which is just wrong
Yes, Elo has the disadvantage that the initial estimate is difficult, and gets oly more accurate with more data. But I think in the beginning of his career, maybe this was about right at the time? Elo has the advantage, that for a 50%-50% game you get more points than for a game you were supposed to win. So the Elo should stabilize soon if a Elo is estimated way too low.

A thought I had, that goes in a similar direction: What would happen if you loop the elo, putting the resulting elo as the starting elo and repeat until it meets some sort of convergence criterion to fake a bigger database, I have no idea what that would really do, the idea just seemed interesting to me.
In machine learning (what I do in RL) that's called data augmentation 11. You are right, I think this would actually increase the accuracy of the final Elo estimate. But a problem is that this way, the Elo development gets a bit lost. For example, if Viper starts his carreer as a #1 (or "pretty high") this may not reflect his true Elo in the beginning of his career.

I think as long you have enough data the starting elo shouldn't really matter, but the question is if the database is large enough.
That's what I hope (and think) too. To say again with the example of Viper: It is probable that in the beginning of his career he was not so good. So it would be correct to estimate his Elo "pretty low", although a precise estimate is only possible after several games.

Better late than never: explicit thanks to all of you who reported errors so far <3
@Poxo @King_Marv @mrdmr @iViktorius @nimanoe @IamDiscern just to name a few
 

Germanymungo__

Known Member
Dec 13, 2018
36
100
48
aoe-elo.com
Voobly
mungo_
View profile
Ladder
RM - 1v1
Rating
2665
Wins
2017
Losses
1584
Streak
4
#69
Dec 21, 2018
7
3
8
#70
I really like that idea, but the settings of KOTD2 for example favoured Liereyy and TheViper with some clear civ advantages against MbL and Tatoh. It's hard to take things like this into consideration I guess..
 

Germanymungo__

Known Member
Dec 13, 2018
36
100
48
aoe-elo.com
Voobly
mungo_
View profile
Ladder
RM - 1v1
Rating
2665
Wins
2017
Losses
1584
Streak
4
#71
I really like that idea, but the settings of KOTD2 for example favoured Liereyy and TheViper with some clear civ advantages against MbL and Tatoh. It's hard to take things like this into consideration I guess..
Hm yes this is difficult. In the worst case we would have to exclude games with heavy disadvantages...
 

Germanymungo__

Known Member
Dec 13, 2018
36
100
48
aoe-elo.com
Voobly
mungo_
View profile
Ladder
RM - 1v1
Rating
2665
Wins
2017
Losses
1584
Streak
4
#72
The work you are doing is excellent, there is a lot of important information in only one site. Just one question: is it possible that you could do the same whit tournaments that happened before 2013? It would be awesome to see the tournament ELO comparison between the "current" pro players and the old ones
Absolutely, and this is the goal :smile: The site is really fresh, and we will try to include everything we find!

I think I would like to see the teams arranged based on something other than alphabetics. Since the players are in order of rating, I think its a bit confusing if the teams are not. But which statistic to use for ordering is of course difficult to say. By the way, ordering the teams based on win-% doesnt seem to work.
I implemented sorting of all tables by any row now. The teams are now sorted by average Elo by default. The percent-ordering works too now in any table :smile:
 
Likes: Kantos
Jan 11, 2016
9
9
8
#73
I really like that idea, but the settings of KOTD2 for example favoured Liereyy and TheViper with some clear civ advantages against MbL and Tatoh. It's hard to take things like this into consideration I guess..
That's true for any kind of game. You can't arbitrarily leave out games because you feel that they gave too big of an advantage to one side over the other. You should just go with it and rely on bigger numbers to minimize the impact of what you consider questionable matchups.

If you don't have the bigger numbers, then the estimated Elo ratings are useless anyway.
 
Dec 11, 2012
2,973
1,040
128
www.youtube.com
#74
is it possible that you could do the same whit tournaments that happened before 2013? It would be awesome to see the tournament ELO comparison between the "current" pro players and the old ones
there are reliable sources on aoezone for

2012 Grand Masters League, Masters of Arabia, Masters of Random Map
2011 CivCup, Brain Champion Cup 3
2010 Master of the Huns

these are added now.

-----

the story of team secret :smile:

secret.png
 
Last edited:

FranceTarsiz

Longswordman
Feb 27, 2017
431
1,162
108
27
Seattle
#75
That graph is a thing of beauty. You can see DauT has been around for the longest, JorDan being close to Viper at some point and quitting at his peak, DauT going low and climbing back to #2 again during the era of Escape HD events where he reached like three or four finals in a row, and eventually the rise of TaToH in recent times, with Viper playing as good as ever and increasing the gap between him and the others ever more at each event.
 

Time

Your time
G M T
Your zone

Upcoming Events

Bluewolf Showmatch ($10) Episode 2
Saturday 12:30 (GMT +01:00)
TheHappyBoar vs PaperHades
DeZu's showmatch $100 TheYo vs Lyx, full 7 games
Saturday 14:00 (GMT +01:00)
TheYo vs Lyx showmatch
ECL South East Asia 3v3 Ro8
Saturday 15:00 (GMT +01:00)
ECL South East Asia 3v3 Ro8
ECL South East Asia 3v3 Semi Finals & Grand Final
March 30th 15:00 (GMT +01:00)
ECL South East Asia 3v3 Semi Finals & Grand Final

Age Of Empires On Twitch

There are in total 29 streamers online
Click here for details
Age of Empires II 1397 viewers
Age of Empires II 201 viewers
Age of Empires III: The Asian Dynasties 104 viewers
Age of Empires II 101 viewers
Age of Empires II 72 viewers
Top