10 replies to this topic

[grendeldog]

So I'm a new player, less than a week in the game. I got a bit of a head start by watching Critical Rocket's videos on YouTube of every mech in the game and how they play - yeah, every last one. I'm on a Mac, so I didn't think I'd be able to play, hence watching the mech porn vids on YT, but lo and behold my laptop can actually run Win7 in a virtual machine and MWO on medium settings. I've got five mechs: Quickdraw with the dual PPCs that was my foolish first buy, Stormcrow prime, and the Hunchback mastery pack.

My question relates to team formation in the matchmaking process. Today I won two games, and every other game - and I played probably five or six hours of match time today - my team just failed to cooperate. Folks ran all over the map alone and got picked off one by one, or they were too timid and got shredded, and we even had a game where the enemy walked into our base on assault and won that way without killing us off because people would support eachother, cover fire, cooperate, or be aggressive enough to engage properly.

Now, I am not bitching that everybody sucks, why do I lose all the time when I'm a god in the cockpit, etc. I'm less than a week in, I don't know all the maps by heart, I am sometimes too aggressive in the mediums that I run and get killed, etc etc. I suck in other words, but I'm eager to learn and I feel I'm improving steadily.

What I want to know is whether I'm getting put in teams with other players that are also not very good. In other words, are teams formed by the matchmaking algorithm in such a way that players of a similar caliber are teamed up?

I ask because then if I keep practicing and get better at the game, I could expect to be placed with other players who are more team-oriented, because cooperation correlates heavily with winning in MWO - even I can see that.

On the other hand, if it is actual random matchmaking, then it was just a string of bad luck, and so either way I have something to look forward to: either better teams or better luck of the draw. I ask because I want to know which of those to expect, and to understand the game better.

Thanks!

[mailin]

I believe the way matchmaker works is that it tries to get it so that each player on average will have a 1:1 win:loss ratio. However, your first 25 drops are sort of like the honeymoon period where it doesn't really try to hard to figure out your ability. After that though, MM is attempting to find what level you are at. It looks at the win:loss ratio of each player on each team and predicts whether team A will win, or team B will.

They use some sort of algae rhythm or something to determine all of this. All I can tell you is that there is this thing out there called the pug lottery. Sometimes you win, sometimes you lose and this even applies to the best players when they pug. There are ways to tip the odds in you favor though. One of the best is by using comms. The game now has in-game voice comms, and even though they are flaky, they are better than nothing. Use them to coordinate with your team mates and call out targets.

Hope this helps, and don't get discouraged. This game has a HUGE learning curve, but it is a blast.

[Rogue Jedi]

the matchmaking uses an Elo system, basically you get assigned a number, at the start of a match the matchmaker guesses if you will win or loose, and by how much, if its prediction is correct your number stays about the same, if it is wrong your number changes, the further out the prediction was the greater the change.

matchmaker tries to put the same average Elo for all the players on each team, this means one team could end up with 1 or 2 excelent playrs and a bunch of below average players, while the other team ends up with a bunch of average players, in that case the average players will likely quickly take out all the below average players leaving the excellent players badly outnumbered.

another possibility is someone on the other team took the lead, and the rest of the team worked with him, in this case the more organized team will usualy win.

the most important things for victory seem to be coordination, aggressiveness and positioning, if you are better on more of those than the enemy team you will usualy win.

Edited by Rogue Jedi, 21 March 2015 - 07:26 AM.

[TheCaptainJZ]

Typing simple commands in chat can help a lot to get a match off to a good start.

"Plan?"
"Meet C4"
"Lights to gamma, everyone else theta"
"Guys, grab the caps!"
"Direwolf D4"
etc. etc. etc. Even if you die, you can rely position information to your teammates while spectating. In fact it's easier to because you can see all their views and data and relay it to the rest of the team. Just keep it to essentials and positive in tone. You can do this with VoIP now too but some people have that turned off.

[Modo44]

MWO uses an implementation of the Elo rating system. Your skill is tracked based on wins and losses, separately for each mech weight class. Matches are typically built to within a few Elo points difference between teams. However, many factors that can change the outcome of a match (chosen chassis, specific mech/build synergies within teams, maps, spawns) are not -- and in part can not be -- included in the matchmaker algorithm, so there is a big random factor in each match. This random factor is bigger when fewer people play (outside of the US and Euro evening prime times).

[Levi Porphyrogenitus]

Each player has an Elo rating assigned on a per-weight class basis, determined solely by your win/loss ratio.

When the MM creates a team, it attempts to achieve a team-average Elo rating. Once it has two teams, it predicts which one will win based on relative average Elo ratings, and adjusts each player's Elo rating after the match ends based on the result (small positive changes for winning a predicted win, large positive changes for winning a predicted loss, etc.).

The problem with this, of course, is that one team might find itself with a whole bunch of 2500-Elo players, while the other has a bunch of 2200-Elo players and one 3500-Elo player. If the single outlier cannot carry his below-average players, then the team of average players will very likely win, though the high-Elo guy might well wind up with a score equal to the entire rest of his team combined.

Ultimately, the MM is something of a crap shoot. The best you can do is use team chat and voice comms to attempt to coordinate. Sometimes a simple "don't abandon the assaults" is all you need, other times it might be necessary actually to take command (done either in the pre-game scoreboard or via the battlegrid (B on the keyboard by default IIRC) and hope that the big shiny star next to your name in the chat will get randoms to pay attention.

Also, it can be invaluable to call out target designations when you want your people to kill something. Each mech has a letter code assigned when a member of your team first detects it. Call out both the letter designation and the mech type, as well as any other pertinant information (say, grid coordinates, build info, or vulnerable components). If you're even a little bit lucky you'll have at least a few teammates try to focus the target you call out.

[Vlad Striker]

If you play Public match and your lost/win rate 1:1 it is OK. If wins more then loses you perform well if more loses you perform bad. It is because you have equal chances to win versus player of your skill level.

[grendeldog]

So first of all thanks for the prompt, detailed replies. I've found less bad behavior - not playing poorly but stuff like rage quitting, talking garbage to teammates, etc on MWO and it seems like it applies to the forums too... No 'Get a PC you queer!' yet =P! People are really cool about new players. I recall when I got my first kill I wasn't sure to mention it, but I did and the rest of the team was super curious about how long I'd been playing, being congratulatory, etc. you'd think that would go without saying but alas in gaming it is often very much not so.

I am pretty fascinated by the Elo system since it provides both random elements - selecting players from a large pool where those players totals are equally ranked and ranked elements (random composition but trying to match team total ratings). It's a clever way of going about it for sure. And I like the random composition today a lot more than yesterday, as it's going much better!

Luck of the draw: two of my three games so far today in SCR-PRIME were 2 kills 6 assists 150,000 C-bills on river city and 11 assists, 2 components, 417 damage, and 200,000 C-bills on canyon network. And on that second game it was two against two with me and a Raven facing something small I never got a good look at and a Timber. We stayed at range, stayed together, and both lived, winning the match.

I have to say calling out sighted mechs is a big part of the fun of my SCR, being 90 some kph, so I wonder if I might enjoy lights even more - the trial Mist Lynx with PPC goes faster and I always enjoyed that. I like the flexibility of the Crow with the large lasers though, and while I've seen lights carrying LL I'm not sure I have quite the skill for that yet. I suck with the Hunch SP and GI even though I love sniping and thus should enjoy the gauss because 60 odd kph is just sooo slow for a medium. With the P variant I can zip around and shoot ML galore - not all at once mind you its grouped so I get no ghost heat - but I have yet to figure out the exact build I need to shoehorn LL or ERLL into the P variant so I can engage at range too.

As for trying to direct the rest of the team - herding cats comes to mind - I'm hesitant to do so because while I can easily see on the map (spamming B key) where people are fooling around and getting caught out alone, I don't want to be the **** trying to boss everybody into good teamwork. I get the difference between suggesting something and ordering it be done, but being new I don't have the confidence to speak up.

I checked my stats just now, and I'm at 31/99 K/D, 45/73 W/L, average 80K C-bills per match, and 710 XP per match. That's not so horrible for a week-old noob, right?

[Rogue Jedi]

something everyone (including me) forgot to mention is that you have 4 Elo ratings, 1 for each weight class, so as I specialize in Lights I likely have a higher Elo in them compared to Assaults which I do not like (I freely admit that I am nearly useless in an Atlas or Dire Wolf) or Mediums and heavies which I am probably below average in.

if you are doing well in the SCR it is possible you were mostly in a different weight class and had earned a higher Elo, or the SCR may just be better suited to your playstyle than other Mediums you had tried.

for example lets say you excel in the Stormcrow and become one of the best Stormcrow pilots in the game, then you buy a Cicada (basicly a 40 ton Light Mech) or Blackjack (a mini Jagermech best suited for direct fire support), both are Medium Mech which share Elo with the SCR but have drastically different roles than the average Stormcrow build so you may be useless in them for your first few matches, but the matchmaker will still be expecting you to be this excellent pilot due to your skills with a Stormcrow, so you will be put against great pilots and most likely loose badly.
however your being an excellent Stormcrow pilot would not affect your Elo for e.g. the Firestarter (light Mech) the Dragon (Heavy Mech), or the Warhawk (Assault Mech)

Edited by Rogue Jedi, 21 March 2015 - 04:21 PM.

[grendeldog]

That's actually pretty cool! I know PGI hasn't done everything perfectly and they seem to have a real problem communicating well with the playerbase, but it seems like there was a reasonable amount of thought that went into the matchmaking system.

I wonder whether there's a threshold for player count below which it begins to ignore the scores. Like if there's only a small umber of people online, will it just make them wait until it can derive two equally matched teams? Or will it place them in unequal teams to get the game going?

I have no suspicion or opinion or evidence one way or the other, I'm just thinking from the server's position whether there is an added impetus for speed in team formation.

[Rogue Jedi]

basically the quicker you find a match the closer the team total Elo is likely to be, and the closer to 3 Mech of each weight class you will have, the longer you are in the matchmaker queue the more relaxed its requirements become, if you are waiting more than about 2 minutes the 3 of each weight class idea will have been relaxed and the teams Elo may not be as close if you found a match in 5 seconds