So...elo Brackets?
#1
Posted 26 March 2014 - 09:13 AM
What do we know about MW:O's elo system? For example, do we know how many elo brackets there are?
How easy is it to go up and down a bracket?
Does it make that much of a difference when you do?
Do you at any particular elo stop seeing newer players?
The reason I ask is that I've been having a particularly rough few days in-game. I usually do pretty decently. I'm almost always in the top three players on the board, which I'm sure doesn't mean much depending on my elo level.
I figure I'm somewhere in the mid-range bracket because I remember reading somewhere that new players are put in the middle elo and I see a a fair amount of champion 'mechs and it seems that there's always someone asking how to play the game or saying that it's their first match.
I haven't seen really any new players for days now, and besides missing helping new players, I'm consistently doing less than 200 damage and that's not very fun.
I'm sure it'll get better and I'm thinking I just need to take a break for a few days, but would an elo jump explain what's going on? Maybe I'm just really bad at the meta shift towards more LRMs.
Either way, I'd love any thoughts and input, Thanks!
#2
Posted 26 March 2014 - 10:14 AM
2. There are no actual "brackets" just numbers distributed on a bell curve. New players start slightly below the average line.
3. When searching for a match, the matchmaker pulls in players with roughly the same ELO score. It then averages the score of each team and matches the average, so the ELO scores of an individual player are not directly matched with another. But the team average should be the same.
Forgot to add: It doesn't just pick players all over the scale. It tries to pick player on both teams within a certain range from each other. For example, within 200 points or something.
4. As time goes on, the matchmaker gets less and less picky when trying to match opponents. After 2-3 (?) minutes, this can lead to an uneven match. It uses ELO as its first matching criteria and tonnage as its second. It is entirely possible that there is a high ELO player on each team but in different weight classes.
5. The max change that you can go up or down in any match is fairly small (50?) so one match has virtually no impact on your score. And its only based on whether your team won or loss
6. Time of day, timezones, and number of players and regions online at any time will make blocks of matches seem very different. If you are in a losing streak, take a break, come back later and you should get matched with a different population of players.
Edited by TheCaptainJZ, 26 March 2014 - 10:16 AM.
#3
Posted 26 March 2014 - 10:19 AM
- According to Paul's display about ELO matchmaking in the launch module post, there are 3 ELO tiers. "1", "2", and "3."
- These tiers have a somewhat wide range.
- From the original explanation ELO tallies premade groups to an average and then takes that average as that of a single score combined with the rest of the group to make a complete team average.
- When possible the game will always try to have at least one premade on either side. The number of players in each group do not have to match and sometimes there may be two smaller premades pit against a larger one.
- Alpha lance is reserved for a premade group whenever possible. If a rival premade cannot be found, ELO appears to draw from higher scoring players to make up the difference. (When in a match simply ask "Do we have a premade?" Almost every single time you get an answer it will come an Alpha lance team member).
- If there is a second premade on a team and alpha lance has 3 or more members, they are set in bravo lance.
- With just one or no premades, lances are filled with ELO-scored players in order of Alpha, Bravo, Charlie.
- Actual arrangement within the lance is not indicative to actual player score, however, and players are rearranged within the lance based on their Match Score.
- An independent ELO rank is kept for each chassis-class. Light, medium, heavy, assault.
- I suspect match score is what really ties into ELO.
- Statements say that win/loss is exclusively what ties in.
- However, match score did not exist before the officially mentioned 'pre-seeding' time of ELO. Also to base a player's performance exclusively by win and loss is not only illogical as victory and defeat depends entirely on the team effort but a failure to ascertain and measure the player's ability.
- You can enter higher and lower ELO tiers than your own by boosting or tanking your scores; you do this by joining a premade of either older (experienced) or newer (inexperienced) players.
- The Bell Curve model released before Paul's more recent one had new players pooling in at exactly center between 1,000 and 1,500 (tier 2).
- Later when the bell curve was shown again, new players were shifted to about the value of 1,100. That is now borderline tier 1/tier 2 and just in the range to get matches above and below the tier line.
#4
Posted 26 March 2014 - 10:34 AM
"If you find a problem, usually it's your problem." I think that applies here.
I'll check back in during some more prime time games and see if I can't wreck some stuff. then. If not, I'll need to re-think my life.
#5
Posted 26 March 2014 - 11:25 AM
elo also looks at the average for a team, not the average per player, so I imagine it is possible to always see high/low elo players depending on group cmpositio and Elo average. I'm pretty sure I still drop into matches with rookies, but I also get dropped into the same matches as high level players like Edmiester.
I don't see players I know to be high Elo very often, so when I do see them it's probably because of grouped players and averages and such, as well as because of a rather broad acceptable range of Elo.
#6
Posted 26 March 2014 - 01:09 PM
dont forget that loosing the game when the elo system predicts that you should loose the game doesnt make your score actually decrease all that much.
if you really want to see whats up, start playing a different weight class instead of what you would normally play and see if the quality of players you see decreases.
Edited by Fuggles, 26 March 2014 - 01:10 PM.
#7
Posted 26 March 2014 - 02:56 PM
#8
Posted 26 March 2014 - 05:00 PM
Most fun ever.
Mainly because I have no idea what I'll be up against instead of being able to look at my team and say "5 assaults, 5 heavies 2 lights. Okay so the enemy has probably 2 banshees, an Atlas and 2 Stalkers. Probably Cataphracts or Jagers. Firestarters. Damn I hate being right."
#9
Posted 26 March 2014 - 06:19 PM
#10
Posted 26 March 2014 - 06:39 PM
Takashi Uchida, on 26 March 2014 - 06:19 PM, said:
11 Assault + 1 Heavy vs 6 Assault + 4 Heavy + 1 Medium + 1 Light
I've seen WAAAY worse tonnage mismatch.
Tonnage isn't the be all end all factor. Its the current average Elo Teams that ruins MM. Lets hope the buckets and 3+3+3+3 helps to fix this. I'm not holding my breath tho.
#11
Posted 26 March 2014 - 06:40 PM
#12
Posted 26 March 2014 - 07:43 PM
Amsro, on 26 March 2014 - 06:39 PM, said:
I've seen WAAAY worse tonnage mismatch.
Tonnage isn't the be all end all factor. Its the current average Elo Teams that ruins MM. Lets hope the buckets and 3+3+3+3 helps to fix this. I'm not holding my breath tho.
I agree, the match posted above is not really as bad as it seems. There had been matches with one team with maybe 8 lights and 4 mediums while the other team had 8 assaults and 4 heavies. The 3-3-3-3 thing could solve major mismatching, but still could end in mismatches. Like one team could be stuck with 3 Locusts, 3 Cicadas, 3 Dragons, and 3 Awesomes, while the other team gets 3 Jenners, 3 Shadow Hawks, 3 Cataphracts, and 3 Highlanders.
#13
Posted 26 March 2014 - 07:47 PM
Amsro, on 26 March 2014 - 06:39 PM, said:
I'd personally prefer an overall tonnage limit when it comes to community warfare, where with a dropship and assuming 660 tonnage you might have one team trying to min/max with 6 Atlases, a Cicada and a Flea (8 players, 660 tons) and the other side might have two Victors(160), one Banshee(90), four Jagers (260), one Blackjacks (45), one Hunchback and a Shadowhawk (55) (10 players, 660 tons) or another team might have 12 Shadowhawks (12 players, 660 tons) or even 3 waves of 11 Locusts (11 players, 3 mechs each).
Don't think the last one would happen, but hey the dropships gotta have something to show for being 216 million cbills each.
#14
Posted 26 March 2014 - 07:47 PM
Phobic Wraith, on 26 March 2014 - 10:34 AM, said:
"If you find a problem, usually it's your problem." I think that applies here.
I'll check back in during some more prime time games and see if I can't wreck some stuff. then. If not, I'll need to re-think my life.
I like you. Very refreshing to see someone say it might be me instead of whining.
If you are looking for a unit, remind me of this thread and I will give you my thumbs up!
PS: My sig is a link if you want to check out the DHB...
#15
Posted 26 March 2014 - 08:48 PM
Nick Makiaveli, on 26 March 2014 - 07:47 PM, said:
Unfortunately it's not likely "him". The game is broken as of today, and the fix is questionable. I hope for PGI that it works out.
But currently the Matchmaker and Elo struggle hard.
Edited by Amsro, 26 March 2014 - 08:49 PM.
#17
Posted 26 March 2014 - 09:55 PM
luxebo, on 26 March 2014 - 09:23 PM, said:
I haven't slept well the past few days. Thanks btw. Lower both blackjacks to Urban mechs and bring in a Locust on the side. Ups the player count to 11.
After trading the Banshee for a Highlander.
#18
Posted 26 March 2014 - 09:59 PM
Koniving, on 26 March 2014 - 09:55 PM, said:
Koniving, on 26 March 2014 - 09:55 PM, said:
I think the lack of sleep is showing... that or my lack of decent rest (sleep? plenty. rest? hah!) is catching up to me.
#19
Posted 26 March 2014 - 10:17 PM
Shar Wolf, on 26 March 2014 - 09:59 PM, said:
Hey, sooner or later Urban mechs will show up.
Technically though that leaves 10 tons left over. Okay screw the Locust, bring in a Spider on the side. Or if we must trade the Urban Mechs for a Firestarter + Raven 3-L duo.
#20
Posted 27 March 2014 - 09:47 AM
Thank you all.
4 user(s) are reading this topic
0 members, 4 guests, 0 anonymous users