Problems With Elo-Hard Stats
#61
Posted 12 March 2013 - 04:44 PM
Had hardly played a match in my commandos since Elo was inplemented, maybe 6-10 matches max. So my rating would be largely untouched from the default value. I decided to go for the COM-1D, mostly for lulz and to try out how a non-ECM "crappy" mech would do. Loaded it up with a test build of 2 MLAS and 2xSRM4 (3 tons ammo) and an XL200 engine.
So what happened?
First 10 matches were pretty much ROTFLOL-stomp. 9-1 W/L. Think I had a KDR of about 4.5. I could solo a 3L at this point.
The following matches it got a bit harder, but still went 15-3. Total: 16-3 after 20 matches. KDR had dropped a bit to 2.86.
After that it seemed like my rating had gone up a bracket. I started losing a lot more, and was consistently beaten by the "better" lights in any 1vs1 situation. Well aimed alphas started oneshotting me. 4 man premades started showing up more and more.
And what do you know, my win rate started to go down from those silly values in the beginning. It actually started to go more towards my general win rate (1.25).
After now having played 38 matches in the COM-1D, I have a 1.53 win ratio (still a bit inflated, but I have no doubt it will go down further). KDR is slowly dropping, to currently being 2.29 (keeping it above 2.0 will be challenge against the 3Ls).
So... that sounds pretty much like Elo is working as intended, doesn't it? Lots of wins in the beginning since I started at a level where I could pretty much solo half the other team, then continously getting harder until I'm at a bit higher rating where it's really challenging (read: I get my behind handed to me by pilots and mechs that are better than me).
#62
Posted 12 March 2013 - 06:19 PM
BigBadHarv, on 12 March 2013 - 04:01 PM, said:
dont worry mate, i have a very very clear picture of you and how you think in my head. you are extremely common on teh internetz and what you think or say has absolutely no relevance on anything substantial whatsoever. good to see you hold grudges though, paints an even clearer picture.
Sweet, it's gonna be all personal insults from here on out. Not interested.
I don't really even think the devs know I exist. I say the things I do about Elo and about consumables because it's what I believe to be true, not for any other reason. Elo has been successful in other situations, other than its original purpose. It can work here, IMHO. There's no reason why it can't be tweaked and worked on until it works. And at the very least, it's better than the alternative of random matchmaking.
#63
Posted 12 March 2013 - 06:21 PM
freak, on 12 March 2013 - 01:52 PM, said:
Ah, hubris. You have to love it.
It's truly too bad that all those years of TT and MW video game playing have nothing to do with experience in ELO.
PS. I'm not disagreeing that the implementation of ELO needs work. The next time I drop in a group of mediums with only one or two heavies, against a team with 3-4 assaults and 2-3 lights and at least 2-4 ECM vs our 0. I will scream.
Edited by Monsoon, 12 March 2013 - 06:27 PM.
#64
Posted 12 March 2013 - 06:40 PM
#65
Posted 12 March 2013 - 06:44 PM
#66
Posted 12 March 2013 - 06:48 PM
#67
Posted 12 March 2013 - 07:33 PM
#68
Posted 12 March 2013 - 07:38 PM
#69
Posted 12 March 2013 - 07:39 PM
#70
Posted 12 March 2013 - 07:53 PM
FerretGR, on 12 March 2013 - 03:34 PM, said:
I like this part in that section the most....
Wikipedia said:
I'm guessing that Blizzard, whos darn famous for MMO gaming success, Tried it and figured out what PGI has yet to realize.. IT SUCKS Donkey Doong like a Tijuana showgirl...
Maybe PGI needs to pay attention more to whats going on and what they did.. and DUMP IT..
#71
Posted 12 March 2013 - 08:13 PM
I think.... that with 12v 12 we will seeeven closer matches as the amount of lead flying will inevitably lead to more deaths on both sides.
All i know is that I dont experience the horrid mass wave stomps that were seemingly premades nearly as often , if ever anymore.
iirc, its called a battlevalue? the mech loadout? I think a modifier that takes that into consideration in conjunction with the elo would yield some very nice balanced matches..
I think with the present system, there are still FAR too many variables in a match to give us the balanced matches we are looking for..
I love the game, so . Im good.. I think pgi is going in the right direction, so we will see.
#72
Posted 12 March 2013 - 08:51 PM
8-10 pm is the time where most people go "What's the plan?" or my team splits up to 2-3 groups instantly, no one knows where to go unless they have someone to follow, etc. A few hours later people don't need to ask what 'the plan' is and scouts know what they need to scout and people know that they should follow that D-DC. I can actually spend a minute flanking the enemy team to get behind them without half my team dying instantly.
I just don't get upset anymore when I play a few games during 'that time'. I'm sure many others experience the same thing. If you think ELO is really crapping on you, play 2-4 hours before or after your usual time and see if there is any difference. If you keep losing all the time no matter what you do then it's probably not ELO.
Edited by Elizander, 12 March 2013 - 08:52 PM.
#73
Posted 12 March 2013 - 09:00 PM
Nidhoggr, on 12 March 2013 - 03:39 PM, said:
but is already broken
because when you switch assault chassis from Atlas to Awesome you will get down in ELO rating
switch back to Atlas and you're going up
the matchmaking value should be a formula that includes your global ELO rating and you per chassis rating (or per variant if taken to extreme)
what that does it places you a bit lower than you bracket when going with a brand new chassis, but still in your bracket range
also PGI should find the median rating globally and set the starting rating (noob rating) for new players about 5% lower
if they are quick learners they will sense achievement, if not ... the fiery pits of hell for him
ps: PGI should take a hard look at battle value (BV) when evaluating MM values
Edited by zmeul, 12 March 2013 - 09:16 PM.
#74
Posted 12 March 2013 - 09:09 PM
- are just plain lopsided ROFL stomps, whether it's imbalanced by disconnects, skill, tonnage, or tech (ECM, etc.)
- seem even, but turn into stomps because of an early tactical turning point (i.e.: noticing the raven and two splat-cats in the caves before they flank your team ... went from 8v8 to 8v5 much quicker than the other team expected, and probably not the way they expected)
- are super tense nail biters, where either team could win up until last shot on the fifteenth mech to die, or the last tick on the capture timer, or 740/750 resources, etc.
Today, three weeks after Elo was implemented, I'm OK with an imperfect system ... in three months, if the average number of lopsided ROFL stomps hasn't decreased, then we have reason to be upset.
About me ... pure PUG, and over the past week (since improved stats came out, and three weeks into Elo implementation), total wins: 78, total losses: 85 ... I'm OK with that (could be better, but overall, my enjoyment for the past week has been positive).
#75
Posted 13 March 2013 - 12:57 AM
No evaluation system will be perfect.
There will always be 8/0 landslides. Players who try new loadouts/mechs and fail in it, tactical errors , lucky headshot for the atlas etc...
The lower your Elo the more landslides you will see.
The lower your Elo the less the weight differences matter.
#76
Posted 13 March 2013 - 02:48 AM
In game one, I did FIVE points of damage and got one assist. This result put me in overall 14th place. And no, the two other pilots did not DC, scoring 4 and 1 points of damage. Only 3 enemy survived. I was the first of my team to go down.
In game two, I did 220 points of damage, scoring hits on a Raven, two different Catapults and an Atlas. None of them went down. In fact, none of the enemy died. This result made me tie for second in my team as far as match scores went and 5th in total damage, both teams combined. Additionally, in my team four top damage scores went to light units (Jenner, my Commando and two Ravens, in that order), next places going to a Cicada and a Trecbuchet. Last places were occupied by a Stalker and a Cataphract. I was the last of my team to go down, the Jenner blew up literally seconds earlier, then I lost a leg and went KABOOM a few eyeblinks later.
I am not a good pilot, and especially not so on light units. That second game was probably the best I have done on my Commando ever. The point is, my skill not increase by a factor of 44 in two consecutive games. In many places, I got really lucky, massive salvoes of fire barely missing me. MWO is a very much a team game and PUG matches are prone to have funny incidents.
In one memorable match about two months ago, I accidentally TKed M A L I C E, who seems to frequent these forums. He accidentally got between my last desperate alpha strike and my target, hunting for his 6th kill of the match. He blew up (ending with 5 kills and 3 assists to his credit), the enemy Catapult blew up the trial mech I was testing and a friend of mine finished him off, because he was focusing on me, ending the match in a fairly narrow victory for our team, despite M A L I C E's Kai Allard-Liao like gaming in that particular match.
Earlier today I had one stomp game against me, which had one paricularly unusual characteristic... Every single enemy got exactly one kill and almost all of them ended with two assists. Though it's a single example, it is a clear piece of evidence that some stomps (8-0 or 8-1 kills is my definition of stomp) are actually very close matches that just go bad for one team. Additionally, only 3 people scored over 500 damage and two of them were in the team that lost.
#77
Posted 13 March 2013 - 06:33 AM
freak, on 12 March 2013 - 01:52 PM, said:
I would heartily encourage more and more players to repeat my experiment and post the results to illustrate the problem, give PGI all the data they can handle, if necessary until they choke on it.
Hamm3r, on 12 March 2013 - 02:02 PM, said:
I think that is what "Freak" is getting at here. A single pilot can not collect enough data to be statisticly significant.
" I would heartily encourage more and more players to repeat my experiment and post the results to illustrate the problem, give PGI all the data they can handle, if necessary until they choke on it."
If several hundred pilots all did this same thing and sent in their results PGI could compile the data and implement a fix, or we can wait for the nextphase to be based on the flawed system we have now.
It is unreasonable however to think that the limited number of people on the Dev staff can build this data on their own as they are only a fraction of a percent of the live player base.
#78
Posted 13 March 2013 - 06:47 AM
#79
Posted 13 March 2013 - 06:48 AM
CMGrendel, on 12 March 2013 - 07:38 PM, said:
Some mechs are just simply better at being the last one killed on a loss or surviving a win than others, even if they don't output the greatest amount of damage. Add playing with people you know and things go up.
I may not pull the greatest damage in a zombie cent, but I'll last longer. Conversely other mechs I do a ton more damage in and probably have a much greater impact but I accept I'm going to die. The one with the better KDR is not the one I do better with in game. But that's just how it is.
#80
Posted 13 March 2013 - 06:52 AM
Fenris Krinkovich, on 13 March 2013 - 06:47 AM, said:
That's what I'm wondering. "Hey you stupid Devs, you forgot to check the wikipedia page! It's only for 1 on 1's!"
League of Legends, arguable the most played game in the world, uses Elo both for it's Ranked and Unranked matches. Given that, I think the Devs couldn't have chosen a better matchmaking system. Just give it time.
1 user(s) are reading this topic
0 members, 1 guests, 0 anonymous users




















