268 replies to this topic

[Kunae]

Rippthrough, on 18 September 2013 - 05:00 AM, said:

He's high-elo, end of discussion.

Unwarranted, but yes he is.

[DEMAX51]

"Matchmaker Breaking Badly for High Elo Players"

Wait a minute, I'm confused...

Is this thread not about high Elo players cooking meth?

[Matthew Craig]

Agent 0 Fortune, on 18 September 2013 - 12:54 PM, said:

I am glad you are looking into it, I am just surprised that the Elo and Matchmaker, which have been around forever are just now getting attention at launch. Getting into a match quickly, and having a well balanced match are crucial to attracting new customers.

We needed to wait for 12 vs. 12 to be in and stable and also for a number of backend telemetry systems to be implemented, no point in tuning if you can't clearly measure what's changing. As stated we could even use more than we have now, a lot of time goes into telemetry etc. backend metrics that isn't forward facing.

[Kunae]

Matt, could you guys please eliminate the "adding low Elo and high elo to get an average ~# value?" I know that I would much rather play with/against a team of roughly equivalent Elo folks, than some really high fitted with newbies on their 3rd match.

It's frustrating to have people in 3pv trial-mechs on your side, when you're pugging at a high Elo.

I would add, that yesterday was much better in this regard.

[Matthew Craig]

Kunae: This was essentially the goal of these changes, and as discussed they did appear to have that affect, but it means that high Elo and low Elo players have to wait much longer for games, as there are smaller populations towards the ends of the curve, so currently you could have what you want but you're going to suffer much longer wait times to get it (what we just saw). We're going to try and figure a better way of getting those results without impacting everyone.

[Kunae]

Matt: Could a solution be to move the new player starting Elo down the curve a bit? Shift them to the left, about 30% and you'd have much less of this happening.

[Chavette]

I don't know how its set up, but you could make search get near exponential at the end of the matching time, so the mm would spend 3 mins finding a close match, and go desperate on the last 1 minute, the idea is, the fallback could be similar to what we had before the fix, but it would make a good try on getting a close match.

Of course this would still be a bad idea in 12s where much more misses happen, but it wouldn't need to be implemented there, as it seems like you guys are already handling the two ques differently.

[Deathlike]

Chavette, on 18 September 2013 - 01:29 PM, said:

I don't know how its set up, but you could make search get near exponential at the end of the matching time, so the mm would spend 3 mins finding a close match, and go desperate on the last 1 minute, the idea is, the fallback could be similar to what we had before the fix, but it would make a good try on getting a close match.

I've never really liked that behavior... mainly because at times prior to this patch, I was generally one of the last people to get into the match... and the game always ends in a roflstomp going either way. It's almost as if the game was pulling a cruel joke on me.

Any time you end up "speeding things up towards the end", it has the same effect of "oh poop, I'll just take anyone on my team" and ignoring the consequence of that decision.

Edited by Deathlike, 18 September 2013 - 01:33 PM.

[Chavette]

Deathlike, on 18 September 2013 - 01:32 PM, said:

Any time you end up "speeding things up towards the end", it has the same effect of "oh poop, I'll just take anyone on my team" and ignoring the consequence of that decision.

It sounds worse than it is. My idea of the "speed up" would equate to the same matches we had before fix, not some complete newbie armageddon.

[Matthew Craig]

Kunae: New players already have that happen to them

Chavette: That's essentially how it works already, and one of our thoughts is more along the lines of altering this curve.

[Deathlike]

Here's a thought.

Since you're still trying to get averages, perhaps you need to use a "corrective/reserved" PUG slot that is always reserved on both teams. This means you will have up to as many premades and solo PUGs as can fill 11-players. The last "solo PUG" player is the corrective player that is supposed to "balance the averages". I haven't really thought this through, but this would allow teams to be more even... to a degree.

I'll have to stew on this for a bit, but feel free to challenge my idea.

Edited by Deathlike, 18 September 2013 - 01:59 PM.

[Matthew Craig]

Deathlike, on 18 September 2013 - 01:58 PM, said:

Here's a thought.

Since you're still trying to get averages, perhaps you need to use a "corrective/reserved" PUG slot that is always reserved on both teams. This means you will have up to as many premades and solo PUGs as can fill 11-players. The last "solo PUG" player is the corrective player that is supposed to "balance the averages". I haven't really thought this through, but this would allow teams to be more even... to a degree.

I'll have to stew on this for a bit, but feel free to challenge my idea.

Again essentially what it does currently, only issue being that either the range has to be quite large to pull in that corrective player, or if it's narrow then the whole bucket gets held up waiting for a player in a narrow range and potentially timing out. To combat wait time you have to take players from what you have available the only alternative is to wait longer, though perhaps as discussed earlier if you could almost see your bucket filling, and were getting good feedback.. waiting for 8 players.. waiting for 6 players etc. maybe waiting 3-5 minutes is deemed more acceptable to get a good match.

Another thought is to allow players to help determine how long they are willing to wait, so you can elect to wait 5 minutes for a good match knowing your guaranteed a tight game vs. electing for short wait times and taking the trade-off yourself that there's a chance it won't be that well balanced. At least each user can decide for themselves. Another potential was if you could ready multiple Mechs in multiple weight classes and have the one kick off into a game that finds the best match first.

[Solkar]

Matthew Craig, on 18 September 2013 - 01:09 PM, said:

Kunae: This was essentially the goal of these changes, and as discussed they did appear to have that affect, but it means that high Elo and low Elo players have to wait much longer for games, as there are smaller populations towards the ends of the curve, so currently you could have what you want but you're going to suffer much longer wait times to get it (what we just saw). We're going to try and figure a better way of getting those results without impacting everyone.

My suggestion to the devs trying to fix this is that each mid range ELO match is allowed to have 1 extra high and/ 1 extra low ELO ranked person per team. This will get players into matches more quickly, maybe only have the programming slide the high ELO player into a mid range match after the 2 minute mark of searching, to give them a shot at an all high ELO match first...

1 elite player per team will not be able to rampage through all 12 enemies to make it a lopsided game (where as 3 or 4 elite players on one team might) especially if the other team also has a very good pilot on it. Besides... even amazing pilots tend to blow up when 5 people shoot them all at once... which can happen all too often.

I assume my ELO is above average (I am not trying to brag or flex my l33tnes or anything, I am being objective) and I would MUCH prefer a match filled with average players where my team and the other team each have 1 high ranked and 11 mid ranked people (or 10 mid and 1 low) rather than wait for the queues to end over and over.

(coding would require a forcing aspect for the waiting high ranked players, where basically IF 2 people waiting more than 2 minutes, THEN push into a lower queue group (displace people that would have gotten in otherwise)... the mid level part of the ELO queue would never reach a point of "oh, only 22 people, lets grab 2 from outside our range", so it would not be a seeking as much as a forcing, at least the way i am envisioning it... hopefully I am expressing myself clearly, but not condescendingly =)

second edit: As a clarification, I am not talking about simply expanding the search radius... I mean go directly to the CENTER of the bell curve and insert 2 high/low players where there will be a match formed easily due to it being the center of the bell curve

Edited by Solkar, 18 September 2013 - 02:36 PM.

[Haakon Magnusson]

Matthew Craig, on 18 September 2013 - 10:39 AM, said:

Sorry to say we simply can't leave certain players with such long times to find a match so we are slowly returning to settings closer to pre patch for now. As mentioned we'll be using all the data gathered to find better ways to get you those match qualities back potentially even better without degrading it for users at either end of the Elo spectrum.

It would actually be really interesting to know if those outlayers are actually those high elo players (Or low for that matter) who sleep in their assault mechs (or whatever single class they have chosen to perfect in)
I'm propably somewhere there in the middle in most classes depending on how crappy variant I happen to be leveling, but in any case I am switching mechs a lot over the course of evening, while I'd propably be much better playing a mastered mech every time.
So if one mm category takes long... so what my next in a light/med/hvy/*** will take a different time and get me matched up.

This makes me wonder, do we need to really finetune the whole mm for those people who have that one damn mech they ride, every damn time? As far as I see, best experience is bound to come to those in the crowded part of elo curve. And if that works, it also works for the most people looking for a game. For the high and low end, expand search parameters more aggressively but do keep what works in the middle going as its been.

Aww whatta ****, I'm due another beer.. cheers...

[Deathlike]

Matthew Craig, on 18 September 2013 - 02:07 PM, said:

Again essentially what it does currently, only issue being that either the range has to be quite large to pull in that corrective player, or if it's narrow then the whole bucket gets held up waiting for a player in a narrow range and potentially timing out. To combat wait time you have to take players from what you have available the only alternative is to wait longer, though perhaps as discussed earlier if you could almost see your bucket filling, and were getting good feedback.. waiting for 8 players.. waiting for 6 players etc. maybe waiting 3-5 minutes is deemed more acceptable to get a good match.

Another thought is to allow players to help determine how long they are willing to wait, so you can elect to wait 5 minutes for a good match knowing your guaranteed a tight game vs. electing for short wait times and taking the trade-off yourself that there's a chance it won't be that well balanced. At least each user can decide for themselves. Another potential was if you could ready multiple Mechs in multiple weight classes and have the one kick off into a game that finds the best match first.

Well, I know weight matching does become an issue since that is a logistical factor that can go beyond elo. If it were a straight elo case, it probably would be easier to do in the whole:
4-man premade (high elo) would need to be countered on that same team with say a 3-man premade (low/avg elo) or 2-man + solo PUG (all having lower elo)... and then to finish off the last 5 slots, you'd get a 3-man premade (high elo) to be combined with a 2-man or 2 solo PUGs..

The weight issue is tricky though... I'd rather be informed through the UI in some fashion that my wait time is longer due to what I pick as a solo PUG. Same could be applied in a premade where the tonnages are likely to be all over the place, and there should be a "projected MM" difficulty value depending on how the premade is constructed.

It's kinda like my proposed MM idea... basically saying "your ETA or time to match is faster if you pick within favorable weight limits, and worse if you're unbalanced".

I dunno, I stopped trying to come up with ideas a while ago.

[Orbit Rain]

I'm going through the same thing bluescreen is going through. Seems like all my solo-pug games are filled with trials on my side and mastered ECM DDC's on the other. Stomp after stomp after stomp...

[Villz]

Matthew Craig, on 18 September 2013 - 11:03 AM, said:

Infinite time with an option to cancel is a definite possibility.

Honestly i think 95% of the players i spoke with would rather wait in que for 5mins to get a good game thats fun and doesn't have people who are so bad they don't know even know how to drive their mech. I can appreciate trying to educate players but a new player is going to learn nothing when he gets dominated in 5 seconds flat by high elo players. I'd rather just wait in que longer for a good game tbh.

All the points made on making the matchmaker being more candid about whats going on during a search would be appreciated.

Also thanks for the updates on the topic nice to know whats actually going on for a change mathew

Edited by Villz, 18 September 2013 - 03:31 PM.

[Xandre Blackheart]

I'm curious if there is indeed an Elo rating for each individual mech that a player has, or if there is a global Elo rating across all the mechs the player has?

I suppose you guys could calculate both pretty easily, maybe that's a way to get a little more granularity for matching players?

Comparing the current Elo for the players mech versus their global Elo might give a better prediction of the player's performance in a particular match.

For example a really good player who doesn't play light mechs very often would have a better chance of getting evenly matched using that type of calculation when piloting light mechs. Or mediums. Or whatever


Conversely a pilot who has terrible performance over most mechs just has the quirk of being a competent dragon pilot. If you matched him just according to his individual performance in the dragon, he wouldn't be matched evenly (he's not that good, just competent). If you compared or averaged his global Elo with his Elo from piloting his dragon he would be more likely to get a more balanced match.

I suppose that brings up issues with how you calculate global Elo for pilots who are still building their mech stable, but it should still work out. If you don't own a particular mech your win loss ratio is 50/50 which is neutral.

Maybe someone has a link to a good explication of how MWO actually does matchmaking at this point?

[Asmosis]

Matthew Craig, on 18 September 2013 - 07:52 AM, said:

We monitored the data throughout the day yesterday and made adjustments to maintain a reasonable average wait time, that said we suspect it is right that very high or very low elo players are probably suffering longer than usual wait times. Reason being that Elo distribution looks like a bell curve so naturally players in the middle of the curve have more players they can match against.

We're looking over the data today and will continue to make adjustments that try to maintain as much of the restriction while maintaining our goals for wait times.

Please feel free to post your feedback on how long you are willing to wait to find a good match, and potential failed to find a match. A quick note on that, the failed to find a match can be a good thing for the match maker e.g. say 48 players trickle into the matchmaker the buckets fill up in the order the players join, now there might be two good matches between them but both buckets have only part of the players they need (like playing a game of cards where you are holding each others cards). When the failed to find a match triggers everyone hits launch again and now the buckets can grab the right players. So bear in mind that it does introduce a dynamic that helps the match maker, that simply extending the timer, or altering ranges does not.

Interested to hear everyone's thoughts as we continue to tune.

My answer to this is "not long"

The reason isnt the duration you have to wait, its that with longer waits you alt-tab to the forums or something and the MWO client has a difficult time sometimes gaining "full" control of your mouse so you end up on a webpage halfway through the match.

As for elo itself, so long as *most* of the players (say 8 or 9) are in the correct bracket, pulling a few lower elo players into the match should be ok if its even between the teams.

Question: as its filling the buckets, does it do one for each team, or does it fill a bucket of 24 then split? I'm guessing two 12 person buckets with the 2-4man teams, but if the MM could dynamically switch players from one side to the other that might help.

(just hold everyone on the "loading map" screen until buckets are sorted instead of dropping into the map loaded waiting for match screen).

Edited by Asmosis, 18 September 2013 - 07:14 PM.

[nemesis271989]

Villz, on 18 September 2013 - 12:12 AM, said:

Spoke with a few very known players in regards who also affirmed the same experience.

Ryan Steel and I were playing together and over a 35min period earlier today failed to find game about 12 times over that period many in a row again and again check the logs on my act please if you need verification.

Pretty jokes in a 2man group only ..... amongst 2 12man teams come on.

(Nerf Villz-Steel Mechworks Breaking Match Maker)

You mean Known Cheese Users