Free Allegiance Forums

Hello,

My name is bastard and as some of you may not know I am in The SystemX. I have started this thread to be a discussion towards ELO. Nothing shall be said in a hateful manner. Just mature discussion about opinions and suggestions. I give Pook and the Admins full right to delete any post towards this behavior.

Personally, I think ELO is flawed, well, because I am a (5). I have seen others that are the same as myself, and newer people that are 14 and 15s. I'm not sure exactly how ELO works because i've never looked into it. i would love to hear what other people think on this subject and how they feel it can be improved.

No, I did not use a spelling or grammar editor for this post.


bastard

Bastard

I was a 15 on Friday logged off and came back today as a 5 Im sure something went haywire in the last few days. As in this is a glitch rather then new Elo in IMHO

MrChaos

i am not talking old or new, I'm saying it's flawed in general and would like to have a civil discussion about it with members of the community, Veterans and Nooblets both are welcome to comment.

How's about before I state any opinions here, if I do at all, lemme ask a question before we get into the whole Elo is b0rked v. Elo will work if you give it time discussion...

I'm rather curious to know what Elo means to you all. I've heard stuff from personal skill, to kill factor, to global badassness, to the degree to which one contributes to a team... so I think we should all get on the same page as to what it is we're talking about here.

I'm curious to see if the different Elo camps have uniformed ideas about what Elo represents.

There are many many hundreds of threads on elo here, I realize you haven't kept up with it, maybe using the search function to go through the highlights?

Elo is constantly being worked on, whether or not somone agrees with it, elo is being improved. Bastard, have you even played a 100 rated games since the last reset? Playing games is what gets your elo closer to your skill. To get even close we need at least 10k games total to get an idea of players true skill levels, all this since the last reset...

jgbaxter wrote:QUOTE (jgbaxter @ Nov 21 2006, 11:02 PM) <lots of stuff>

Define "skill" please. /mrgreen.gif" style="vertical-align:middle" emoid=":D" border="0" alt="mrgreen.gif" />

10k games? Don't you guys think that's sort of.... excessive? This is why I think ELO isn't the best system for Allegiance.

I mean, I've made my views on this pretty clear, but that's a HELL of a lot of games. And it's going to take me many many months to participate in 100 ranked games, because I only play on the weekends. I'm still a better player than some newb w/ the same rank as me, however. ELO is going to have a hell of a time keeping up with situations like this, though. I've got 7 years of experience that ELO has no idea exists, and never will because I don't get to play as often as I used to. My skill level, though, hasn't declined back to newb level because of my lack of playing every day. I play enough to remain very familiar with the game and able to benefit any team I'm on (usually /tongue.gif" style="vertical-align:middle" emoid=":P" border="0" alt="tongue.gif" />). But, like I said, ELO is going to have a hard time tracking this since I don't play that much these days.

Not everybody is created equal in Allegiance, which is a fundamental idea ELO is based on. (When I say this, I'm not speaking of skill level equality persay)

But, I guess the problem I speak of is going to be present in any ranking system that gets implemented. Personally, though, I think I could come up with something that would work pretty good =)

Ok, this thread should probably be in Misc. Dev. but anyway...

What does ELO record?
ELO is based solely on whether the team you were on won or lost. Your individual actions are not (and probably never could be) recorded. This means that there will be some games that you lose that aren't your fault (and by the same token there will be games you win that you didn't help with). This may at first glance seem unfair (I did great and the team lost so I lost points wah wah wah!). However ELO is a statistical system so any single game is not really relevant to your final score.

How are ELO 'points' awarded?
You gain ELO points by winning games and lose them by losing games. The number of points gained/lost is based on a prediction made by the system of what chance each team has of winning. If you have a low chance of winning you get more points if you do win. Each player on a team gets the same number of points.

How is this chance to win worked out?
The formula is in the ELO thread in the Misc Dev forum. Basically all the ELO ratings of the players on a team are added together and then compared to the other team. ELO then spits out a % chance to win for each team (this formula is in need of work at the current time IMO).

Hang on, all this is talking about teams, I thought ELO measured each individual?
ELO will only ever look at which team won/lost each game. However, as time goes by, each player will a totally unique record of games they played in and which team they were on. Therefore this unique record allows for a unique rating for each player. If you tend to win games you will have a higher ELO rating, if you tend to lose them your rating will drop. As mentioned before there will be exceptions, games where you played well but the team you were on didn't and so you lost. However the more games you play, the more statistically insignificant any single game becomes. That's why people say that a large number of games are required for accurate rating.

How does my rank relate to ELO?
Your rank is you ELO/100 rounded down. So an ELO of 1599 gives a rank of (15). There is a modifier for new players which reduces rank based on the number of games that person has played.

Ok, I think that covers the basics, after all the only worthwhile discussion is an informed one. However Bastard this ground has been covered a number of times before and is rarely productive. I personally think that, whilst flawed, ELO is the single best rating system for achieving balanced games in Allegiance.

I think that the formula that determines the chance to win of each team needs tweaking as a current 70% stack is pretty much unwinnable. Also a factor of commander ability is needed, but I accept that this can only be introduced once we have data even games to generate a commander rating from. How a com rating should be added is open for discussion closer to the time.

ps. Tiger, if you want to use the above as the basis for a KB article feel free. I can add stuff if you like too.

I think elo's biggest flaw is anti-stacking. When you join the last 10 minutes of the game and lose, you lose a lot of elo. Maybe there should be a change with the 5 minute barrier. Instead of getting/losing 0 elo when having played less than 5 minutes can't we do something like:
-Calculating team elo
MinutesPlayedPlayer(1) / GameLength * EloOfPlayer(1) + MinutesPlayedPlayer(2) / GameLength * EloOfPlayer(2) + ....

-Calculating elo difference for a player
MinutesPlayedPlayer(1) / GameLength * EloDifferenceMax

this has probably been suggested before, but what is wrong with this?

[edit] never mind: http://www.freeallegiance.org/forums/index...ost&p=47705

ELO -- give it time. Overall, ELO ranks are getting more accurate. There might be some individual differences, like hey -- I can own that dude in a dogfight, but my rank is less than his.

it also does not reflect the skill of the commander - it is just a reflection on how many times you are on the winning or losing team. Commander skills, IMHO, have more to do with winning or losing than just about anything else, unless it is a team of newbies vs. a team of vets.

Yes, you get penalized for joining the losing team, whether you are actually anti-stacking or just want to play. Losing team is typically the one with slots open to join the game.

It is my hope that the autobalance patch will improve the quality of the games, as teams theoretically will be more evenly matched.

It would be nice to see some adjustment added to take into account the commanders win-loss percentage and add that into the mix of ELO and balancing. Perhaps a formula that weights the commander winning percentage into the overall team ELO?

I am not a statistician, nor am I a programmer. Perhaps someone with those skills could post here as well with an analysis?

Summing it up -- I am in favor to let current plans for ELO and autobalance take their course. I have not seen any credible alternative to ELO vetted here. theer have been threads on what an alternative player point system (points for nanning, killing bases, etc.), but no practical plan for how this might be implemented.

I have seen many complaints about ELO, but not much about ELO ruining game play. IMHO, games have gotten better since ELO was reset and has been operating a while. It may not be perfect, but it appears to be the best that we have.