Jump to content

Paging Karl Berg...karl Berg, Please Pick Up The White Courtesy Phone...


1911 replies to this topic

#261 Modo44

    Member

  • PipPipPipPipPipPipPipPipPip
  • Bad Company
  • 3,559 posts

Posted 13 April 2014 - 11:54 PM

What you are describing, is a matchmaking problem, only partially related to skill tracking. The stats are there, but the matchmaker has tolerances so big, it breaks itself for many outliners (both very good and very bad/green players seem to experience stomps often). This discussion revolves around adjusting the matchmaker to 1) better track actual skill and 2) build matches with players of closer skill levels.

Replacing Elo (it is not an acronym, really) would not help that. It already tracks all skills combined. Any attempt to emulate that by weighing various statistics is guaranteed to miss some skills that are not easily tracked, e.g. PUG herding or tactical decisions (like when to cap, and when to fight). If you miss some skills and skill weights, the tolerances within the matchmaker automatically become bigger, breaking it further. A similar effect happens currently when you switch between your very best and worst mechs within a weight class -- they are tracked as one, so you have very easy/very hard mode mechs.

Edited by Modo44, 13 April 2014 - 11:57 PM.


#262 Tekadept

    Member

  • PipPipPipPipPipPipPipPip
  • Knight Errant
  • 1,290 posts
  • LocationPerth, Australia

Posted 14 April 2014 - 12:08 AM

View PostModo44, on 13 April 2014 - 11:54 PM, said:

Replacing Elo (it is not an acronym, really)

I know but you try telling my phone that :o

Please explain how it currently "tracks all skills combined" I fail to see how it does that based on how it has been explained to us, it is based on win/loss and whether based on whether matchmaker thought you would win or not? based on the Elo (damn that was harder then it needed to be)

Edited by Tekadept, 14 April 2014 - 12:10 AM.


#263 p4r4g0n

    Member

  • PipPipPipPipPipPipPipPip
  • Knight Errant
  • 1,511 posts
  • LocationMalaysia

Posted 14 April 2014 - 12:19 AM

@Tekadept - Having separate Elo ratings for solo and group drops would partially address the issue you raised. Having said that, what would you suggest in place of Elo or are you considering a hybrid using Elo and something else?

#264 Kageru Ikazuchi

    Member

  • PipPipPipPipPipPipPipPip
  • The Determined
  • The Determined
  • 1,190 posts

Posted 14 April 2014 - 01:08 AM

View PostTekadept, on 13 April 2014 - 11:40 PM, said:

But my point is, the current ELO system doesnt seem like it would track any kind of individual skill whatsoever. ...

That is absolutely correct ... Elo doesn't measure skill ... particularly when you're talking about one individual's Elo score in a 12v12 game.

While the scenario you present is one way to "game the system" in the short term, I don't think the individual match adjustments should be large enough the artificially inflate (or deflate) a player's Elo so much that they're affected signficantly.

I might be wrong, though ... I wonder how long it would take before a high-Elo player's Alt account would level out with the same Elo as a primary account? 50 matches? 500? ... I hope it's closer to 500.

#265 Modo44

    Member

  • PipPipPipPipPipPipPipPipPip
  • Bad Company
  • 3,559 posts

Posted 14 April 2014 - 01:11 AM

View PostKageru Ikazuchi, on 14 April 2014 - 01:08 AM, said:

I might be wrong, though ... I wonder how long it would take before a high-Elo player's Alt account would level out with the same Elo as a primary account? 50 matches? 500? ... I hope it's closer to 500.

About as long as it takes to finish Elite skills on the first mech. 100 should do it. Why do you hope for 500? That is crazy long, and such a result would show Elo as not actually working.

Edited by Modo44, 14 April 2014 - 01:12 AM.


#266 Naduk

    Member

  • PipPipPipPipPipPipPipPip
  • Overlord
  • Overlord
  • 1,575 posts
  • LocationAustralia

Posted 14 April 2014 - 01:34 AM

View PostBrian Buckton, on 13 April 2014 - 12:52 PM, said:

I'm always thinking about ways to improve the social interface and user experience and I would consider adding extra alert indicators, it's something that's been bugging me as well :o


lots can be done to improve social interface
first step would be a nice msg sound, it should be ignorable but also grab attention if you deep in mechlab dry tongue mode
it should also not play more than a few times per second, if you have 1 or 6 people messaging you at the same time with high frequency the sound would be enough to ruin lives (likely those in close proxy to me)
sound should also play to people who are alt tabbed ... likely looking at Reddit.... :)

the social icon flashing is not good enough
either the flashing needs to be stronger, the button larger
but what i would prefer to see is a lite animation come from off screen to the right and hit the social button
a pulse of light, a string of LED's, a glowing envelope, a sci-fi esq data flow packet, any of these would fit i think
this should only play if the social tab has not been active for some time (a minute or so?)

when in a group getting ready should be ALOT EASIER!
especially for the group leader !
the primary launch button is awesome and MASSIVE ... so use it
for players in a group change the launch button to a ready switch
for the group leader, give the same option, however with a few more advanced options
when all players in a group are ready (excluding group leader) make group leaders ready button flash and play a tone or msg
when his/her ready button is flashing , make it operate exactly how the launch group button currently works (ie launch the game)

when a group is not ready and the leader clicks ready, ping players in the group with a ready up request
DO NOT USE A POP UP BOX THAT NEEDS TO BE CLICKED!!!!!!!!!
make their ready button flash and play a sound (get betty politely request their mech ready status or announce drop imminent, or just a hurry up beep)

in the group status window , have an option for "show group stats" (each grp member can toggle separately)
this would put an overlay in the top right corner of the screen (unused space) with all members of a group and a light next to their name indicating their ready status, this overlay would be transparent so it would not interfere with any current or future screens
if room permits include what mech each player is piloting

not social related but
get rid of the damn ok/confirmation boxes
ie, skills
after clicking on a skill you want
then clicking on what currency to buy it with
then clicking on yes i am sure i want to do this
we do not need to click OK on yet another window saying the item was received
the fact that it changes colour is enough, play a sound or throw particles if you must
but enough with the clicking ok to proceed stupidity

ill stop here before i write 500 pages

#267 Kageru Ikazuchi

    Member

  • PipPipPipPipPipPipPipPip
  • The Determined
  • The Determined
  • 1,190 posts

Posted 14 April 2014 - 02:27 AM

View PostModo44, on 14 April 2014 - 01:11 AM, said:

About as long as it takes to finish Elite skills on the first mech. 100 should do it. Why do you hope for 500? That is crazy long, and such a result would show Elo as not actually working.

I just picked two numbers that were an order of magnitude apart ...
... 50 = way too quick ... I can do that in one day, easy (well, one day on the weekend when my wife is out of town)
... 500 = thinking about it, you're right, probably too long

Maybe I should have picked 25 and 250 as examples ... or put a time frame on it ... I'm OK with it taking a week to dramatically affect your Elo, but not a day.

#268 Modo44

    Member

  • PipPipPipPipPipPipPipPipPip
  • Bad Company
  • 3,559 posts

Posted 14 April 2014 - 02:39 AM

Again, why would a quick convergence be bad? The very goal of big early movements is to quickly put you close to your converged Elo value, and then refine it. The faster you arrive at your actual skill value, the fewer bad matches where your skill is too high/low compared to your score. In fact, the recognized matchmaker problems stem from the perceived lack of skill convergence -- players ending up in impossible or incredibly easy situations even after hundreds, even thousands of matches.

Edited by Modo44, 14 April 2014 - 02:41 AM.


#269 East Indy

    Member

  • PipPipPipPipPipPipPipPip
  • The Hammer
  • The Hammer
  • 1,245 posts
  • LocationPacifica Training School, waiting for BakPhar shares to rise

Posted 14 April 2014 - 04:09 AM

View PostKarl Berg, on 13 April 2014 - 09:20 AM, said:

So while we could shoot off a pile of rapid-fire projectiles, that would have knock-ons that would require additional investigation at the very least. Increases in network traffic, increases in processing time and collision detection costs. Or, we might consider turning AC's into trace-fire at the cost of simulation fidelity. The shells would then fire with effective infinite velocity, no longer have gravity falloff, etc.. That's not necessarily a tradeoff we would like to make.

If it's not a direction you'd like to go, and imperfect convergence is out, do you guys have any interest in separating cannon trajectories in some way? Two suggestions include recoil (moving the reticle between shots) and tenths-of-second firing delays (allowing the target to move between shots).

#270 Heffay

    Rum Runner

  • PipPipPipPipPipPipPipPipPipPip
  • The Referee
  • The Referee
  • 6,458 posts
  • LocationPHX

Posted 14 April 2014 - 04:26 AM

View PostTekadept, on 13 April 2014 - 11:40 PM, said:

But my point is, the current ELO system doesnt seem like it would track any kind of individual skill whatsoever, unlike Match Score which has some basis on individual skill ( I know its not as comprehensive as it could be)?
A person can join a match, rush in and die in the first 30 seconds and his team wins. so his ELO would go up if it was determined so. That person did absolutely nothing, yet his ELO goes up just for appearing in that drop and doing a suicide charge?


You don't understand how Elo works then. It's not an instantaneous measurement, but an average over time. The results of one match are meaningless to your Elo, but if you suicided 100 times, your Elo *will* go down.

#271 Tekadept

    Member

  • PipPipPipPipPipPipPipPip
  • Knight Errant
  • 1,290 posts
  • LocationPerth, Australia

Posted 14 April 2014 - 06:10 AM

View PostHeffay, on 14 April 2014 - 04:26 AM, said:


You don't understand how Elo works then. It's not an instantaneous measurement, but an average over time. The results of one match are meaningless to your Elo, but if you suicided 100 times, your Elo *will* go down.

I understand it perfectly in relation to it being ranking an individual based on individual performance. You are failing to grasp what I am trying to say, what I'm trying to say is it *WONT IF* you happen to luck out on a large portion of those 100 matches and get a good team made up of good teammates/premades who carry your team to victory. It wont go *DOWN* as much as it really *should*.

Edited by Tekadept, 14 April 2014 - 06:12 AM.


#272 Viken

    Member

  • PipPipPip
  • Moderate Giver
  • Moderate Giver
  • 58 posts

Posted 14 April 2014 - 06:17 AM

Unrelated to the whole Elo thing... I do have a question for Mr. Berg to at least hopefully point me in the right direction, since he was talking about the servers.

Why will private, custom matches require Premium time? This was very lightly answered by Paul in a previous section (I think Dev Video Blog 2, can't check because I'm at work) saying "Server load." The following "spoiler" is simply my opinion on the matter as a whole as opposed to something you can do anything about. Feel free to ignore.

Spoiler


Now, for the part of the question you can actually answer:
  • How does a private match take up so many resources that would require gating it behind a financially compensated method?
  • At what point would PGI need to get to, to be able to have free custom games?
If point 2 is not realized and the private matches remain gated behind premium time, I have 2 follow up questions:
  • Would we be able to "turn off" our premium time so that we can save it for when we're actually playing?
  • Can we get purchasable "Wins"? For example: Buy 10 wins worth of premium time. They stay on your account until you've won 10 matchmade (Play Now! button) games. Alternatively, you can use a "Win" as a Premium Private Match instead.
Even if you don't answer, I still appreciate you reading this. Thanks for the open discourse Karl!

#273 Heffay

    Rum Runner

  • PipPipPipPipPipPipPipPipPipPip
  • The Referee
  • The Referee
  • 6,458 posts
  • LocationPHX

Posted 14 April 2014 - 06:24 AM

View PostTekadept, on 14 April 2014 - 06:10 AM, said:

I understand it perfectly in relation to it being ranking an individual based on individual performance. You are failing to grasp what I am trying to say, what I'm trying to say is it *WONT IF* you happen to luck out on a large portion of those 100 matches and get a good team made up of good teammates/premades who carry your team to victory. It wont go *DOWN* as much as it really *should*.


Statistically speaking, what you are saying is completely irrelevant.

#274 Tekadept

    Member

  • PipPipPipPipPipPipPipPip
  • Knight Errant
  • 1,290 posts
  • LocationPerth, Australia

Posted 14 April 2014 - 06:39 AM

View PostHeffay, on 14 April 2014 - 06:24 AM, said:


Statistically speaking, what you are saying is completely irrelevant.

In that extreme case yes but you are the one that went extreme, but you missed my point anyway sigh. Back to playing elite.

BTW this thread needs to be renamed, it should be Karl berg & friends Hour of Power or something ;)

#275 Heffay

    Rum Runner

  • PipPipPipPipPipPipPipPipPipPip
  • The Referee
  • The Referee
  • 6,458 posts
  • LocationPHX

Posted 14 April 2014 - 06:44 AM

View PostTekadept, on 14 April 2014 - 06:39 AM, said:

In that extreme case yes but you are the one that went extreme, but you missed my point anyway sigh. Back to playing elite.

BTW this thread needs to be renamed, it should be Karl berg & friends Hour of Power or something ;)


To answer your question, the only true metric to measure skill is your ability to deliver a win. Every other metric can be gamed, but win or loss is the goal, and that is how you determine if you helped contribute or not. And over a sufficient number of games, Elo does exactly that.

Someone who is running around with a 2200+ Elo based off of nothing but win/loss is still significantly better than someone running around with an 1800 Elo. Both are still very, very good players, but the nice thing about Elo is that it can fine tune exactly who is probably better *on average*.

#276 p4r4g0n

    Member

  • PipPipPipPipPipPipPipPip
  • Knight Errant
  • 1,511 posts
  • LocationMalaysia

Posted 14 April 2014 - 06:55 AM

@Viken

Karl previously explained here that they run about 24 instances i.e. matches per blade and talks further about it here as well. I'm personally not aware of how this would translate in terms of costs but other posters have noted that it would explain the need for premium private matches.

Edited by p4r4g0n, 14 April 2014 - 07:00 AM.


#277 Viken

    Member

  • PipPipPip
  • Moderate Giver
  • Moderate Giver
  • 58 posts

Posted 14 April 2014 - 07:19 AM

View Postp4r4g0n, on 14 April 2014 - 06:55 AM, said:

@Viken

Karl previously explained here that they run about 24 instances i.e. matches per blade and talks further about it here as well. I'm personally not aware of how this would translate in terms of costs but other posters have noted that it would explain the need for premium private matches.


I had noted that, hence the follow up questions ;) Thanks for taking the time to point me in the right direction though!

#278 dangerzone

    Member

  • PipPipPipPipPipPip
  • The Death Wish
  • The Death Wish
  • 295 posts
  • LocationSomewhere in a F14-Tomcat

Posted 14 April 2014 - 09:08 AM

View PostRoadbeer, on 12 April 2014 - 03:20 PM, said:

Oh, then the forum has about 1,000 of them, just go to Feature Suggestion sub-forum. Or we could get Victor back...

Do you want 3PV
No
Hell No

Unbiased poll is unbiased.


This made me laugh more than it should have.

Reading through these replies from our obviously good friend Karl, and I have to say....PGI sadly has not done a lot of justice to the community as much as they have wanted to, but if there is one thing they did do, it was hiring this magnificent guy named Karl Berg. You sir are a godsend to us who have been through hell and high water since this game's induction in early 2012.

People like you give me the hope required to hold out longer, knowing that so long as people like you are working at PGI, there is still hope for this game.

I believe someone else said it before, but, what do you look like? (photo), I need to buy you a damn beer. Congratulations Karl, you're now in the danjazooone! (of beer and trust, that is) ;)

Edited by dangerzone, 14 April 2014 - 09:10 AM.


#279 StuffYouFear

    Member

  • PipPipPip
  • Bridesmaid
  • Bridesmaid
  • 80 posts

Posted 14 April 2014 - 09:35 AM

View PostBrian Buckton, on 13 April 2014 - 12:52 PM, said:

I'm always thinking about ways to improve the social interface and user experience and I would consider adding extra alert indicators, it's something that's been bugging me as well ;)


Hey, I have a tiny request that may sound silly, but can the social button that blinks be a tad taller?

One issues Ive run into with the new UI is it follows your in game screen res and windowed/full screen options.
I run two monitors, one for gaming and one for social(ts3, skype, ventilo) and since the old UI was always windowed and small it was no big deal with all the UI features remaining visible on screen when window was not focused, but unfocusing the new UI brings the windows task bar directly over and hiding the social button, also hiding the flashing notification.


You guys have no idea how much of my lunchbreaks yall have taken up these last few days, thanks Karl and Brian, you have done alot to make us like PGI again with your openness.

Edited by StuffYouFear, 14 April 2014 - 09:41 AM.


#280 Veranova

    Member

  • PipPipPipPipPipPipPip
  • Bad Company
  • Bad Company
  • 542 posts
  • LocationLondon, UK

Posted 14 April 2014 - 09:52 AM

View PostKarl Berg, on 13 April 2014 - 10:20 AM, said:


Good thoughts. I'm not sure I 100% understand your approach, so apologies if I've misinterpreted something (and correct me please if I have misunderstood!).

It definitely better reflects the biases that can impact player skill than our simple per-weight class system we currently have. However, here are some issues that would need to be considered with pair-wise tracking of Elo:

This is N^2 in memory cost o...


I think you understood what MischiefSC was saying, and I can also see why it wouldn't be feasible.

But I think it's over-complicated. If you look at how a player behaves when grouped, you'll probably find broadly two types of players:
A. Players who play with friends for fun, may have some loose communication, but generally are just out for a good time.
B. Players who play to win, communicate everything going on, and aggressively posture to have their team in the best place to win the match.

Although we all have our ups and downs between these two, I expect that the modifier would only have to estimate this "synergy" for the individual, instead of between every single friend and them.
That way it would just double the number of stored ELO values. 1 for each weight class when solo, 1 for each weight class when grouped.

I certainly find that my ELO gets artificially inflated when I'm in my light and working with 3 heavier team-mates, because when I go solo or do Lights training in a group of 3-4 lights, we consistently lose for a whole evening very often.
It feels like we're placed as very high ELO players, which are typically in Assaults carrying their teams. Meaning that we simply cannot pull our weight the same in a Light lance.
That actually brings up another issue, of bigger 'mechs ELO being worth the same as a smaller 'mech's ELO, which I don't think it necessarily is. If you looked at my ELO in periods of running with assaults in my group, compared to periods of mainly running lights regularly, you'd probably find the modifier you need to balance ELO against the weight class.





19 user(s) are reading this topic

0 members, 19 guests, 0 anonymous users