Global Rank via Score, not Win/Lose

[codepants]

The subject says it all: I think global ranking would be more meaningful if it was based off of game score, not whether your team won or lost.

Consider the following scenario: You are an 1100 ranked player (not unlike myself). In almost every game you play, you are the last player left standing and you have by far the highest score, mass income, and energy income (next to the air players on maps with air spots). Yet your allies die, and that's game. You can't fill three spots in a four player game.

What happens?

You never get above that rank. As you go down, you play against lower ranked players who you can run over, ensuring your rank doesn't fall too low. But it never gets too high, either. Not because you aren't a good player, but because you always end up with allies who get themselves killed.

On the other end of the scale, there are the folks who "ride" the wins of other players, joining games stacked towards them to increase their rank.

I find this all very frustrating. In short, you can have good players who always lose because of their allies, and bad players who always win because of their allies. So gaining/losing points based on wins/losses is meaningless.

I propose a new way of gaining/losing rank that is based solely, or at least mostly, on score during game. There are two ways this could be done:
(1) Taking the median score of the game, the average ranking of the game, and appointing points on how well players "should have done" against the median based on their rank. For instance, a comparatively low ranking player in a low scoring game who scored well would gain points, a comparatively high ranking player in a low scoring game who scored well would gain fewer points, etc.
(2) Comparing the score of a player to the scores of other players who played that position based on the time they died. For instance, take all the scores of players at 20:00 +- 2 minutes in position x on map y. If you die at 20:00 +- 2 minutes and score above the average and you are below average ranking, you gain points. Etc.

I acknowledge a few shortcomings with this system.

For starters, if you know what makes your score go up, you could rig the system. However, I don't think it's easy to do both that and stay alive long enough for it to make a difference in what your ranking would be.

Second, it doesn't take into account you doing stupid things, which is part of the original problem. You could have the highest score but overexpose your ACU -- classic mistake of a lower ranking player -- and gain points because of it. I'd like to say that this is accommodated for because no player who can score high enough to raise their rank would overexpose their ACU, but I can't substantiate or un-substantiate that claim.


I know this isn't a perfect proposal, and I expect there to be some discord, but I wanted to put it on the table and see what thoughts there were. So, ready, go.

[Ze_PilOt]

The goal is to win the game. Not to have the best income or having the chance to be the last one to be targetted. If you are the only one left in your team, it can be your mistake. So it's normal you are "punished".

With your idea, as score only represent mass/energy income (and ACU kill), you don't need to fight to gain rank, only to eco. That's not a good way to rate people. Also, it's overly complicated for, most likely, a worst relevant rating in the end.

The rating should be based on how well you eco, help your teammate, fight, expand,... And the only variable that tell that is .. You winning or losing. That's because if you do everything as it should, you will probably win.

If you think people are "riding on other people wins", that's wrong. Because if they are rated less, they will make their team (and them) win less in the end too.

[codepants]

The goal is to win the game. Not to have the best income or having the chance to be the last one to be targetted. If you are the only one left in your team, it can be your mistake. So it's normal you are "punished".

I agree that the point is to win the game. But I disagree that being the last one alive is synonymous with "having the chance to be the last one to be targeted." It can be your mistake. It can also mean you are better than your allies. I don't think either of us can win here without any data -- I'd be curious to know if FA collects any data that could substantiate either of our arguments here. Without that, then on this point specifically, I think we're shooting straw men.

With your idea, as score only represent mass/energy income (and ACU kill), you don't need to fight to gain rank, only to eco. That's not a good way to rate people. Also, it's overly complicated for, most likely, a worst relevant rating in the end.

This isn't true. Eco too hard too early and you lose. The game rewards aggression. Because my suggestion compares to other people in the game, you have to eco better than that other guy. And the best way to do that is to be aggressive.

The rating should be based on how well you eco, help your teammate, fight, expand,... And the only variable that tell that is .. You winning or losing. That's because if you do everything as it should, you will probably win.

I agree that the rating should be based on how well you eco, help your teammate, fight, and expand. But I disagree completely that the ONLY way to tell that is if you are winning or losing. If you think that's true, you have never played a team game. You can do everything right and still lose because your teammates make mistakes. Furthermore, you can do everything right except one thing and still lose -- for instance, failure to build flak or TMD.

If you think people are "riding on other people wins", that's wrong. Because if they are rated less, they will make their team (and them) win less in the end too.

1) Substantiate your claim.
a) Your "because, will" statement is a statement of causation. I agree there's a correlation between rank and skill. But having a good rank does not cause someone to play better.
b) You don't believe it's possible at all to ride on other people's wins? Not in the least? You believe the ranking system is 100% perfect and everybody is exactly the rank they should be regardless of the people they play with?
2) To be clear, I am not saying that everybody rides on everybody else's wins. Obviously that's impossible because nobody would do anything. I am not saying that's the root of the problem or even that it's happening often enough we aught to do something about it. I'm saying it's a symptom of a problem that we should fix.

As I said, I don't have any data to back this up, but since you think the ONLY way to tell someone's rank is whether they win or lose, then watch this replay and tell me what I could have done differently. DaNoob overextended and Doobers suicided. I'm not saying I played it perfectly. I did not. Obviously I made mistakes. And I am not saying I am special, this is simply the most recent example I have seen that is convenient for me to post. There are plenty of other replays and casts where this happens to someone.

By disagreeing with me you are essentially saying that even if your allies do nothing you will win a game -- that the magic of rank will kill the other players. Even if you join a 5v5 and your four allies don't click the mouse a single time, you should be able to win "because of your rank." We both know that statement is ridiculous.

[Serafijn]

I don't like this idea, nor do I see the use of it for the ranking. Do you feel like you deserve a higher rating?

[IceDreamer]

Take the best of both worlds. If you win rating rises, if you lose it falls, but factor each player's score/minute into how MUCH those ratings change. If you're the last one left, AND you were gaining score faster than your allies, and you lose, you'll lose less rating than another, worse player would have in identical circumstances. Voila!

[Reaper Zwei]

Why should people gain and lose rating differently amongst the same team? its a team game. Also the game you posted is desynced and the people you are complaining about are lower rated then you one of them being half your rating. Why would you expect someone that far below you to play as good as you? If it bothers you that much why not just host 1000+ games? Might take you bit longer to get going but could be less headaches for you.

[Serafijn]

Take the best of both worlds. If you win rating rises, if you lose it falls, but factor each player's score/minute into how MUCH those ratings change. If you're the last one left, AND you were gaining score faster than your allies, and you lose, you'll lose less rating than another, worse player would have in identical circumstances. Voila!

I like how you win or lose with the team. I fear such things might induce score-whoring..

[codepants]

I don't like this idea, nor do I see the use of it for the ranking. Do you feel like you deserve a higher rating?

1. Explain.

2. Not necessarily. If the ranking system is truly only intended to be a measure of how good you are at ending up on the team that wins the game, then I'm probably right where I should be. If it is a measure of skill, then I don't know. I haven't played with all the players on FAF in all situations, and even if I could I couldn't possibly synthesize that amount of information accurately, at least not before it had changed. No ranking system is perfect, but I think this one could be better, which is why I want to have this conversation. And my idea may not be the best idea, or even better than the way it is. But nothing gets better if you assume it is perfect. I see a problem so I am calling it out and suggesting a solution.

Take the best of both worlds. If you win rating rises, if you lose it falls, but factor each player's score/minute into how MUCH those ratings change. If you're the last one left, AND you were gaining score faster than your allies, and you lose, you'll lose less rating than another, worse player would have in identical circumstances. Voila!

Consider the following scenario: A 600 and an 1800 play two 1200s. According to FAF, this game is perfectly balanced (the sum of the ranks equal each other). I would be willing to bet with some certainty that in 80% of the games they play, one 1200 will hold off against the 1800 until the other 1200 kills the 600, then the two 1200s will double team the 1800. Perhaps this is a facet of there being more subtle differences in the higher ranks, perhaps I'm wrong, but let's say that's what happens. Does it mean the 1800 is a bad player? No. Does it mean he/she is worse than the 1200s? Probably not. But play this scenario enough and the 1800 will be ranked below them. That is the problem I am calling out.

I like the way the current system factors score into points gained/lost, but it does not save our friend the 1800.

Why should people gain and lose rating differently amongst the same team? its a team game. Also the game you posted is desynced and the people you are complaining about are lower rated then you one of them being half your rating. Why would you expect someone that far below you to play as good as you? If it bothers you that much why not just host 1000+ games? Might take you bit longer to get going but could be less headaches for you.

Apologies, my copy seems fine...

The one that suicided was only 100 below my rank, I believe.

I don't expect everybody to play at the same rank, but I don't think players who play with people of a lower rank should be punished for doing that. IMO the fun of the game is playing with a variety of ranks, learning from each other, etc., and the way the system is currently set up, it unfairly either penalizes or rewards players who choose to play with other ranks (see the 1800-600 example above).

I appreciate your suggestion, and it might very well be a plausible solution, but I don't think the resolution should really be to restrict the way the game is played. Making the game less accessible should not make it better, making it better should make it more accessible.

I like how you win or lose with the team. I fear such things might induce score-whoring..

Perhaps. As I said, this is not a perfect solution. I'm not experienced enough with the scoring system to say how accurate it really is, but I guess I'm suggesting it's a better representation of player skill than whether or not one wins or loses. I can't substantiate or un-substantiate that claim, however.

[IceDreamer]

codepants, your error in that scenario example you explained is in assuming that 1200 + 1200 = 600 + 1800. I admit it looks like it should be the case, and it is conventional, current player's wisdom, but that's not actually correct. Any automatic team matchmaker will have to take into account what I'm about to say about this for this very type of scenario.

Supcom has an extremely steep learning curve. From what I've seen, there's a point at which the playing field begins to level out, and that point is at around 1000 rating. Players below this point are, in a word, appalling. This is the zone of people who either have no clue what to do at all, or are just incredibly bad at doing it. There is a second plateau at 400 rating too, which separates people who are absolute beginners still struggling with the UI and recognising units from players struggling with game concepts, strategy, and getting everything to flow properly.

In essence, there is an enormous gulf of a difference in play ability between a 300 rating player (Meaning 300 rating after many games, so the system knows them) and an 800 rating player. There is also an enormous gap between an 800 and a 1300, despite it still being a gap of 500, though the gap is smaller. After this though, everything constricts and the game's metrics become more subtle. The difference between 1300 and 1800 is there, but it's not an overwhelming gulf of a difference. It's more about efficiency than knowledge.

You are right. A 1200 rated player can hold off an 1800 for some time in most scenarios, and 1800 vs 1200 is not enough of an advantage for the high rating player to defeat both of them. On the other hand, as an 1100/1200 rating player myself, I can comfortably take on two 600 rating players single-handed, and do in fact find myself doing so fairly often in my friendship group. It is not unheard of for me to take out two teams, each consisting of a 300 and a 600, by myself, even when they have an unofficial truce and are focusing me, much to their frustrations on skype It's the same gap, and yet it's not.

1200 + 1200 VS 1800 + 600 is not a fair team setup. The 600 is significantly outclassed by the 1200, and effectively a waste of space and mexes. A fairer matchup would be 1200 + 1200 VS 2200 + 600, as those are the players I'd say could feel confident about taking on the pair. So yeah, the 1800 player in your scenario is going to lose out on his rating if he keeps on playing in such a biased scenario.

I do think, however, that he should lose out at a slower rate than his partner, because there's no doubt he will be performing better in each battle.

[belatedcube]

I think it more depends on which strategy you use on what player at a certain time and what he does to fight itnot really how much you eco, or kill. for example jester snipes: if they scout and built aa (specifically inties) your jesters BETTER be denied. if they don't build any aa or air factory then they die (just look at the facepalm casts BRNK has with jesters). TMLs: if they scout and build tmd your tac missiles will fail. if they don't scout and/or build tmd then you can kill everything you want.

so a score based on wins isn't exactly accurate (in teamgames) because you can do terribly and still win because your teamates did something even though you didn't (which is why 1v1 ranks are more dependable).