Didn't devs plan to make intentional big probability to dropping with same people? I thought they already implemented it. And if so, you cannot judge anything by such statistics.
If the whole matchmaking and elo-system would work with a good algorithm i wouldnt often have high damage/kill-scores with my bad mechs like the superfragile Dragons (a real pain to play compared to other heavies). E.g. three consecutive rounds with one of my Dragons, which i really don't like to play because they die tooo fast: [img width=700 height=437]http://i42.tinypic.com/dy0z94.jpg[/img] [img width=700 height=437]http://i40.tinypic.com/1z4fhaw.jpg[/img] Look at the scores, i'm not playing very well, having not very high scores compared to what i get with other mechs, with Dragons i can't brawl very aggressively, but need to skirmish more (in Round 2 of the three you see, that i was the only one brawling!? with a dragon?! wtf!!). But still with my "average" score i am one of the topscorers. wtf? Playing with random teammates, which still are training aiming, or with random Enemies, which often run toward me and overheat too fast but without dealing high firepower - crap builds? running shs? running no extra heatsinks? don't know how to fill those Hardpoints properly? don't know how to setup your Mech with optimal Balance? I guess, many many players out there, really still don't know Smurfy or Mechspecs.com. They just run builds, which seem to be ok. I just posted today a screenshot of a Orion-Pilot with only Double ERLLas and STD350. Wtf? No wonder, if he doesnt achieve good teamcontribution or personal stats. Why am i mixed with such dudes again and again - forcing me to suffer many Team-Losses? Coincidence or Matchmaking-Algorithm Failure? IMO this shows, how bad the system work, mixing people randomly together, which play on totally different level. Maybe if the Elo- and Matchmaking-Algorithm would work properly i wouldn't get high scores with bad mechs against Teammates with much more powerful mechs. Here the rule works: "A skilled fighter can kill a troop of noobs with only a knife." Give him a bunch of equal skilled fighters and he really needs some BFGs or Tactics. And this is where the real fun begins. Atm i'm playing game after game, looking at all the dudes experimenting with their mechs+builds+gameplay and so leading to inconsistent statistical results. After so many hundreds of played battles i expect to get drops with players, who already know how to build their mechs and how to play tactically. But it doesnt work atm. Simply spectate your teammates after Death, and look at their Weaponloadouts and how they play. You will be shocked, with which players you are mixed! Everything seems to be Random. Have some Drops with Noobs in your Team VS some veteran Premade Troops, add regular Gamefreezes to it, and you really have broken Data for the Elo-System. Goodbye Match-Making-System. As said, i don't care anymore. As long i can experiment with creative builds after builds (and that's still the most fun for me ) i am happy with MWO. Give me some Multiplayer-Campaign with an epic story in the future, give me some community warfare, give me some more Mech+Weapon-Options and you can make me explode in a Mech Porn Orgasm. Arrgh, MW1 was a joke, but MW2 was best Mech-Sim IMO. Maybe we have to wait for MWO2 to get what we now expect!
You know what happens when they put you and a good player on the same team Blagg...the enemy gets steamrolled like you and I did on Arctic lol so maybe your ELO is so high they think they can add you to these crud teams and you will single handedly win it. You are a victim of your success
I dont know how they calculate the elo, but it should be based on daily and weekly stats and also by weight range. For instance, I do 400-800 damage in a medium, I am generally a good medium pilot, but i couldn't trip over someone and do damage to them in an assault. average round in an assault would be maybe 100-200 damage, horrible. So while dropping in an assault, I would like to drop with other people who basically suck as hard as i do. But on the flip side of that, I really hate carrying 11 window lickers when i pilot my mediums. It would be interesting to see on the dashboard, the amount of connected players. At least then when i get unfilled drops, I could think to myself, there are only x people online right now, its understandable (how hard would it be however to change a player from one side to the other? ie 10 vs 12, hell give me a locust from the other side, anything is better than nothing). And I always seem to get put into the side that is filling up. the other team always seems to have the full 12 when I join. weird.
There are actually 4 ELO brackets per player; Light, Medium, Heavy and Assault ELO is currently calculated STRICTLY by wins and losses. If you win then your ELO, for whatever weight class you are in, goes up a little and if you lose your ELO, for whatever weight class you are in, goes down a little. Very basic rudimentary and ineffectual system. I've spoken to one of the devs about this and he agreed that it needs work. It functions but it needs more tuning to be able to function PROPERLY. My contribution to this discussion was that ELO needs to be representative of kills, damage done, capture points acquired, mechs countered (ECM), mechs TAG'd etc. I suggested a much more robust system that would reward us for showing up and not just whether the pugs managed to fuck it all up for us or not. The problem is that they have a gazillion other things going on in order to keep we, the mob, happy and they are falling behind.
Even with what we know (just wins/losses) we still don't know how Elo is calculated. Without knowing what the K-factor is we can't know much about the system. Is k a constant? Does it change based on the player's Elo, as is common in chess? Does it change based on number of games played? etc.
I didn't want to have to break out the math but here you go... Figure 1 In Figure 1 you will notice the highest score a player can have is 2800. The lowest score a player can have is 0. When a player first starts playing MWO, they are assigned an Elo score of 1,300. The maximum amount a player’s score can change in a single match is ±50. One of the key calculations that needs to be done is to determine the probability of a lower ranked player beating a higher ranked player. An example of this is, if two players of the same ranking have a 50% chance of winning each. The bigger the gap between the two player’s scores, the less probability of the lower score player beating the higher scored player. http://i.imgur.com/Ttx2q.png[/img] Figure 2 A player’s new rating, after a match is complete, is calculated using the probability value determined in Figure 1’s formula. In Figure 2, we take a look at a sample of how a player’s score changes between matches. In the example, one player has a rating of 1,350. Another player has a rating of 1,410. The first step is to calculate the probability of the 1,350 rated player beating the 1,410 rated player. The rating difference between the two players is found by subtracting the lower player score from the higher player score. In this case it’s 1410 – 1350. There is a difference of 60 points between the two players. Filling in the formula from Figure 1, we find that the 1,350 rated player has a 0.41 probability of winning (or 41% chance). This is where Elo starts to determine the outcome of a match and the result of player scores after the match is complete. http://i.imgur.com/J5ZiG.png[/img] Figure 3 We now use the probability value of 0.41 to determine the change in player ratings. Case 1: 1350 Player WINS over 1410 Player Since the lower rated player won despite the odds against him/her, they are rewarded a much higher change in score than the higher player is. The higher player score will actually be reduced. One variable that is set by winning or losing is the WinFlag (as seen in Figure 3). The WinFlag is a binary (true/false) value of either 1 (Player has won) or 0 (Player has lost). In this case, the WinFlag value is 1. Let’s begin the calculation: Old Rating = 1350 Maximum Change Allowed = +50 for a win, -50 for a loss (as seen in Figure 1) WinFlag = 1 Probability of Winning = 0.41 1350 Players new ranking = 1350 + 50 x (1 – 0.41) = 1380 1410 Players new ranking = 1410 – 50 x (1 – 0.41) = 1381 A player’s rating will only go down if they are beaten by a player who has a lower rating than theirs. In this case, if the 1350 player lost, their score would not change since the Match Maker was correct in its prediction. So how does this affect Match Making – Phase 3? This formula and scoring system is run against match data that the current dedicated servers are giving us. Basically it’s being tested on real world numbers generated by you, the community. Once we get a full understanding of how accurately the Match Maker is working, we are going to add some additional parameters to the mix. These include a more defined player skill rating and a Mech weight class balancing system. More info on these when the first pass of Elo testing is done. CLARIFICATION: I've been asked to clarify the Elo system above. It makes sense for a single player but not for teams. Does a player on 1 team get rated against a player that he beat on the other team? Here's how a team works: At the start of a match, all player's on ONE team have their Elo ratings totaled and divided by 8 (max players). You may realized that this is simply the AVERAGE of a team's total Elo. Team 1's average and Team 2's average are then used to calculate the probability of win (as per the formulas above). If Team 1 beats Team 2, then the appropriate math as above is applied to each player using the probability score calculated by the team averages. Some people also asked to have the description simplified. Here's the summary: The Match Maker uses a scoring system to determine if your team is more likely to win or lose based on your team's average Elo rating. If the Match Maker determines that you're going to lose, but you actually win, then your Elo score is going to go up and the enemy's score is going to go down. If the Match Maker determines that you're going to win and you actually win, then your Elo score isn't going to change very much (if at all). The same applies to a prediction of loss and you actually lose, your score may drop but it will be slight. The more games you participate in, the more accurate the Match Maker becomes and you will start seeing that you are playing against people of relative equal skill. About Elo and Competitive Play As you know, Elo is now a functioning part of Match Making. We are aware that this is going to cause problems with tournament/community organizers who rely on sync-drops to play against specific teams. While not officially sanctioned by PGI, we do appreciate the effort these organizers are putting in to run these events. That being said, we are not intentionally trying to disrupt your ability to do so but at the same time, this side effect of Elo helps curb the problems associated with sync dropping. Elo is meant to balance the Match Making mechanic for the entire community. As mentioned before in other avenues, there are numerous competitive players here at PGI, in all types of game genres, and we fully understand the need for features and tools that support this style of gameplay. We DO plan on bringing in the functionality to run PGI controlled tournaments/special events, and Community Warfare will fill in the rest of player's needs. ELO Tuning (February 2013) We just wanted to notify the community that we are aware the new match maker can take a while to find a match. We're actively monitoring the data coming in from the match maker and we are planning to tune it to ensure that the average time to find a match is acceptable. Even with pre-seeding there is a degree of balancing that needs to occur over the next day or so. Please bear with the match maker and don't Alt-F4 etc. as that will only make the issue worse for now. We will keep you posted as we make updates. Data taken from this thread: http://mwomercs.com/forums/topic/66079-matchmaking/page__view__findpost__p__1626065
All I can say it, I rather take the old unbalanced matchmaking that would give me a fast game. Sure it is unbalanced, but at least I get to ride my mech and shoot at stuff. Right now it is like this: Load matchmaker > long wait > may even find no match > wait again > wait for game loading > wait for names to fill > wait for everyone to click ready > wait > then game. The game lasts less than 10 minutes, but the wait time and load/unload tim between games is at least 5 minutes. And after all these IT IS STILL UNBALANCED MATCHMAKING!. Just give me a fast game ffs. I rather get slaughtered. At least I got to play, and got to get some cbills and xp.
I think some odd UI behaviors cause some people to misunderstand how match making is working. The entire match is "filled up" (all 24 players chosen) before anyone gets launched into the game. When you actually connect, your client immediately displays the enemy team's names (because that's all it knows about them), but does not display each of your teammates until they've actually connected to the server and the server has their Mech loadout (since you're standing next to your teammates, it actually needs to load their Mech's graphics too, which may play into it). As a corollary to that, if any of your teammates disconnect (or fails to initially connect, perhaps they crashed while loading, or alt-F4ed when they saw "Terra Therma" pop up, or got bored while waiting for "Searching..." to finish and they were part of a pre-made lance) before loading in, it looks like you just have one less player on your team, even though the match maker did assign you someone. The other team, however, sees that player name there (and sees they're not marked ready). Because the match maker presumably just sends the match off to the server running the game, the game server itself doesn't do any balancing, which means if a couple people from your team disconnect on their way in, it's going to look like it's a 10 v 12 game. That's not really the fault of the match maker, and by the time you're connecting to the game server and know your allies loadouts, that's a little late to do any last-minute rebalancing to handle disconnects. Disclaimer: this understanding of how match making is working is based primarily on observing what's happening while talking with players dropping into the other side of a match I'm dropping in, so I may be slightly misunderstanding some of the specifics, but I'm pretty confident on the generalities =).
Intersting, Jimbery - I didn't know that and always questioned why the other lance was full when I got in and why PGI said they'll never just drop those disco's and backfill them while we wait 2 minutes for the game to start...
If the disconnects are happening prior to the match it is still a match making failure whether or not it sends that information to the "play" server is irrelevant simply because if they did not put any code into the match making system to prevent the match maker not knowing there was a disconnect because of a failure of communication with the play server its still a match making problem. They should have designed in redundancy to prevent occurrences of players dropping before joining (Play server to MM server - "hey MM, player dropped send me another"). You have to wait at the start anyway for 60 seconds and that's 60 seconds they could back fill the short team. In a game with 30 to a side dropping 1 or 2 on occasion is no big deal but when 2 players is nearly 20% of your team that's a huge deal. If it only happened once in awhile I wouldn't mind but it happens over and over and over again. I've played lots of these games and haven't seen that figured out perfectly yet so it isn't a dig at PGI. In my opinion if we are going to be waiting up to 60 seconds for the match to start that would be an ideal excuse for them to put in this redundancy because there is ample time to just simply take the next available player and plonk him onto the short team.
They wanted me to lose this one; Them: 4 assaults, 6 heavies, 1 medium, 1 light Us: 1 assault, 1 heavy, 3 medium, 7 lights (8 really, as cicadas are only pretend mediums) I'll post up a screenshot when I am awake and work out how to embed. 1am and i should have gone to bed hours ago... but i was on a win streak.
That light swarm... I'd say this match is very unpredictable and nearly 50/50, depending on skill allocation. Those lights can easily maul all those assaults and meds one by one, even if they are in close formation. But if those lights would be too untactical, then assaults should win. Unless if lights just outcap. P.S.: I have matches pretty close to this about 1 from 10, not very rare.
Another thing to remember about the pre-game lobby is that any enemy player with 0 ping has not yet connected, or has DCed. So it's just the way things are displayed that makes it seem like the other team gets everyone connected at once.
I'm seeing exactly this too. Often, I'm doing top damage if solo-in a jenner/spider...I should not be out damaging a Battlemaster for example. One notable game I was in my Spider 5v, which has ONE ER large laser. The next guy down was 200 damage below me-and I'm frankly, an awful player.
I know what ya mean bud...It is giving me a swelled head thinking I am a much better light driver than I am when I come top damage match after match. Check out this new video to see what I mean... That is as soon as Michael decides to post it He may not though because It relies on modules and thats not really part of the mech build is it....ahh screw it! Michael this ones not gonna be for your mech builds section this ones just to show off a couple modules so I am gonna just post it....soon
