Get rid of ELO?

[eldryan]

This won't show skill more clearly. It will just predetermine games to a larger extent and judge each team as "equal" teams. It thus presumes each player to be "equal" all season long. If each player is equal, there can actually be no "winner" to this contest, or winner can be an empty title as not every team has played each other nearly enough times. If each player isn't equal, then ELO is a prerequisite to fair teams and actually creating "meaningful" game results.

Your methods are actually the same as ELO when the season starts and everyone is at 1000 - it creates results, but not really informative or precise ones with the speed of the ELO system. Over an infinite number of games, meaningful trends might start to resolve but in a small sample the equality of the games will be quite debatable. The main result is far more wins by players who typically win, and far less for those who typically lose.

For some reason I'm having a delay in posting my argument...I'll post a more in-depth analysis after.

[Krumme]

Considering the posibility of keeping the ELO system but basing it on your method/public game method - with random teams and being able to win less and more ELO with random teams
But then again no one wants to loose 20 ELO nor win 5 too often - start fighting for making captain mode more regular?

[godlik3]

we could do captain mode on all games on loby
problem is the ballance
for example achiles has more than 1k elo onw with 20 losses more than wins
thats sad
so imo all games should be 15/15 elo and all dc only -15 for players and always +5 for the others ( elo raises from nowhere ik )
so all games would be captain mode and always 15/15
i think this is the fairest thing we could do!

example:
brave and me did 3 captains mode yesterday
he always choosed first yolo.syle with 700 elo and me dark with 800
no one wanted fanatismo cuz he is to high elo and would f**k the elo ballance...thats unfair because any of us would choose him first if the game was 15/15!
( not telling yolo and dark r bad ok? )

[tastay]

It's already stupid that people receive the same ELO for winning/losing while being on the same team and that is the main reason distortions will arise

A 1400 elo player should never receive the same ELO for the win as a 900 ELO player even when being on the same team because the chances of that 900 ELO ever getting higher will always be harder especially in a league with such a small player pool

[Krumme]

what do you suggest then @tastay
This is a team game and the weakest link is just as important as the strongest

[tastay]

I obviously don't have the formula on hand but in most DotA2 inhouse leagues the ELO is done in such a way that the highest ELO player will win less points than the low ELO players on the same team. Currently everyone joins and the teams are randomised in a way to make everyone either win or lose 15 points. Now I believe a more fair system would be where one where the highest elo player might win 12 and the lowest elo player wins 18. You just can't use a 1v1 formula for a 4v4 game

[Krumme]

does the high elo player than loose 12 and low elo loose 18 or how is it constructed? everyone still looses 15?

[tastay]

High ELO player also loses more and low ELO player loses less

[dweiler]

This won't show skill more clearly. It will just predetermine games to a larger extent and judge each team as "equal" teams. It thus presumes each player to be "equal" all season long. If each player is equal, there can actually be no "winner" to this contest, or winner can be an empty title as not every team has played each other nearly enough times. If each player isn't equal, then ELO is a prerequisite to fair teams and actually creating "meaningful" game results.

Your methods are actually the same as ELO when the season starts and everyone is at 1000 - it creates results, but not really informative or precise ones with the speed of the ELO system. Over an infinite number of games, meaningful trends might start to resolve but in a small sample the equality of the games will be quite debatable. The main result is far more wins by players who typically win, and far less for those who typically lose.

For some reason I'm having a delay in posting my argument...I'll post a more in-depth analysis after.

Thx for clearing it up Eldryan, and I can see where you are coming from. I think the difference is that you see skill in the ELO, and I want to see it in games (not only to reach true ELO, but all season long). An example to explain what I mean:

For simplicity's sake, let's presume 2v2 games with 1 player of 1100 ELO, 2 of 1000 and 1 of 900. All players have reached their 'true' ELO (that is, the ELO that represents their skill). In the current system, if you play 10 games, it will be most likely that both teams get 5 wins, because the skill is evenly divided between the teams. Any differences will be caused by luck.In the proposed system, if you play 10 games, it is likely that the one who has 1100 ELO in the other system will win 6 or 7 games, and the one with 900 ELO only 3 or 4.

Or: if a really good player reaches 1500 ELO (which is his true ELO), he will from then on win only 50% of his games. This is in fact a huge achievement, because he stays at the level of 1500 ELO, but there is nothing to be gained or lost for him anymore, except playing 50% - and this goes for anyone who reached his true ELO. In the proposed system good players will keep winning more games than they lose, so that it in fact shows that they are better players, instead of being drawn to 50% by the system (but keeping their high ELO).

In the ELO-system then, if everyone reached their true ELO, individual games are not decided by skill anymore, but by who has the better towers and send luck (and will generally be 50% wins and losses), while in the randomized system, skill will often (but luck is a factor too of course) make the difference.

I like it more if individual games in the league become more skill-based, instead of the balancing in the ELO-system, which, I will admit that, does make people be where they should be in the rankings, but makes games less interesting, because the influence of skill is axed from them.

[eldryan]

So if a player is below 1000, it's better to just not play - will end up higher w/l. or to play bare minimum. And the teams will still not reflect elo necessarily, it will just make dodging even bigger impact.

[supersexyy]

So I've thought about this and this is what will happen.
1. Higher players will win more games and lower players will lose more games (obviously).
2. The 'spread' will be bigger. Ie instead of having players on 900, 1k and 1.1k, we will have players on 700, 1k, 1.3k. This does not mean that better players will stick out more, it will just mean the ELOs will have a higher standard deviation.

So, apart from the fact that better players may 'deserve' to win more games (60-70%), I see no advantage.

[Diablo_]

I also still can't see any (meaningful) advantage =)
"unbalanced" (the best players randomed into same team) games might be fun at times but in the long run games should be as balanced as possible imo because that's where competition shines because every game is a challenge and there will be far less "stomps".
Plus win/losses can get pretty crazy, if the best player also plays the most he will be so far ahead of every other player which will make a wrong picture of the skill difference (just imagine the first player having twice as many points as the second one, mainly because he plays a ton more games since his win ratio will never go down), whereas there will never be such extreme outliers in a system using ELO, because the more ELO you have, the harder will it be to gain more, BUT he will still be on the first place.

[supersexyy]

Btw if you only play games where you are the highest elo then the only limit on your elo is the number of games you can play in a season. The system doesn't logically make sense.

This is extremely different to a soccer league where you are constricted to a set number of games. In a system where the number of games is not set (chess, dota etc) a scaling system needs to be implemented.

[ZTX)Fiji]

Don't forget the fact that players improve and do not stick to an Elo system just because it's there to balance the game - sometimes you learn things best the hard way - and your arguments are too robottic regarding how the elo system works.. as if the elo system will decide whether or not you stay on top or keep falling.. . Don't forget that people improve during a season, and don't forget that new players has just as much a chance to become champion than the hall of famers... Nothing stays the same but the suggestion about getting rid of elo is just an idea of how to improve everyone's developements in a different more fair way aswell as staying on top if you truly belong there. - Those who drop amd keep dropping obviously don't belong in the league and will be replaced as always.. . Unless the league needs to suffer under too big of a skill level difference.

[ILOCOS_NORTE]

For simplicity's sake, let's presume 2v2 games with 1 player of 1100 ELO, 2 of 1000 and 1 of 900. All players have reached their 'true' ELO (that is, the ELO that represents their skill). In the current system, if you play 10 games, it will be most likely that both teams get 5 wins, because the skill is evenly divided between the teams. Any differences will be caused by luck.In the proposed system, if you play 10 games, it is likely that the one who has 1100 ELO in the other system will win 6 or 7 games, and the one with 900 ELO only 3 or 4.

Or: if a really good player reaches 1500 ELO (which is his true ELO), he will from then on win only 50% of his games. This is in fact a huge achievement, because he stays at the level of 1500 ELO, but there is nothing to be gained or lost for him anymore, except playing 50% - and this goes for anyone who reached his true ELO. In the proposed system good players will keep winning more games than they lose, so that it in fact shows that they are better players, instead of being drawn to 50% by the system (but keeping their high ELO).

In the ELO-system then, if everyone reached their true ELO, individual games are not decided by skill anymore, but by who has the better towers and send luck (and will generally be 50% wins and losses), while in the randomized system, skill will often (but luck is a factor too of course) make the difference.

I like it more if individual games in the league become more skill-based, instead of the balancing in the ELO-system, which, I will admit that, does make people be where they should be in the rankings, but makes games less interesting, because the influence of skill is axed from them.

Thx mick, that points it out best.
Since I always write books you can read my conclusion only

New system critics:
The problem is that we want to have a league but cant use the typical league rating system where every player fights every other player in every possible team at least 1 time per season (like it is in a football league). If we use your new "1 point per win" system we cant prescribe players to play hard and easy games in ballance. That means, if for example fiji, Brave, Pata and Fana play the majority of their games together in a 2s competition they end up with w/l rates of 45-55%. If someone else is "farming" the bot players meanwhile he will get more points with a >60% w/l rate while he would stick to the same 45-55% w/l rate if hes playing with the top players. The only way to implement the new "1 point per win" system is to set up certain games at certain dates like in a real league. But thats not possible since Legion TD is no ones profession (I hope ).

Reaching true ELO is good:
The real ELO system comes from chess (afaik). You play chess of course in 1 on 1 competitions. But the players should also reach their true ELO at some point. And if they play vs others with same ELO their chances to win the game are of course 50%. In Legion TD we also ballance the teams in that way that the winning chances of both teams are indeed 50%. But at the beginning of a season we are all at 1k ELO. So the games will be of course unballanced, wich leads to good players reaching their true "high ELO" and bad players reaching their true "low ELO". Over the time (and the closer you come to your true ELO) the winning chances in a certain game become more and more ballanced (up to this point you will have a winning chance in each game of ≠ 50%). So good players win more games then they lose. When everyone reached his "true ELO" they have a 50% w/l chance each game and the season should be over because the ranking wont change anymore. That should be after 3 months (some players reach their "true ELO" faster then others because of high activity).

What can we do to improve the current system:

1.] Change the duration of a season.
At the point where players reached their "true ELO" the season should be over. If that is less then 3 months the seasons should be shorter

2.] Change the ELO you can win per game:
If you want TOP players to win more games in a 3 months season just decrease the amount of ELO you can win in one game. Players need more games to reach there "true ELO" and will have a higher w/l ratio because they dont sit to long on their "true ELO" where the w/l rate is 50%.


Conclusion:

When all players reached their "true ELO" and games end up with 50% winning chances the season should be over or should not take much longer.

For a higher (lower) w/l ratio of good (bad) players decrease the ELO you can win per game. At the moment it is possible to become #1 in few days. Then you play 50% games after just one week for the rest of the season. To deal with very active players with up to 500 games per season we should limitate the speed of gaining ELO. During the first month you can just gain or lose 5 ELO per game (+/- ballancing). During the second month you can win 10 and during the third month you can win the originally 15. That would stop people from reaching there "true ELO" to fast. And it will lead to more imballanced games because we would be all close to 1k elo for a longer time.

[spoiler=]Just another mathematical thing we need in adition to make people with same ELO gaving the same w/l ratio (or w/l ratios would not matter anymore):
To make the w/l of players with same ELO equal:

During month #1 1 win counts as 1 win and 1 loss as 1 loss.
During month #2 1 win counts as 2 wins and 1 loss as 2 losses.
During month #3 1 win counts as 3 wins and 1 loss as 3 losses.[/spoiler]




To finish I just want to say that this is in my eyes the only way to make good players win more games. But I dont say that it is what I want because the league is more about fun and less about competition then Mickey wants it to be. And thats a good thing. If we make bad players lose even more then they do already ... idk