Jump to content

Why Elo Doesn't Work Here


633 replies to this topic

#321 RussianWolf

    Member

  • PipPipPipPipPipPipPipPipPip
  • Legendary Founder
  • Legendary Founder
  • 2,097 posts
  • LocationWV

Posted 24 January 2014 - 11:08 AM

All done here.

No one has been able to give plausible explaination of how ELO is accurately working in this environment (because it isn't).

Quote

[color=#959595]Example. Lets say we were to set up a hypothetical new account. We drop in random pug matches with this account. But instead of fighting we simply run off to a quiet corner and hide for the duration, never DCing. Hypothetically, the enemy never finds this account and it never fires a shot. Now the 11 other members of the team fight it out with the enemy and sometimes they win and sometimes they lose. [/color]

[color=#959595]This account would then have an ELO rating of x while having demonstrated no skill what-so-ever. [/color]

[color=#959595]How has ELO rated this account? How can it?[/color]

[color=#959595]ELO was not designed for this environment, so any data it generates is flawed. Using the data that ELO generates in any way will then give flawed results as well.[/color]

[color=#959595]So if the ELO data is flawed, and the matchmaker uses that data, then the resulting matches will be flawed as well.[/color]

[color=#959595]Does the ELO system work? yes. Does ELO work here as it has been applied? NO! [/color]

Edited by RussianWolf, 24 January 2014 - 11:08 AM.


#322 RichAC

    Member

  • PipPipPipPipPipPipPip
  • 661 posts

Posted 24 January 2014 - 11:12 AM

View PostRoadkill, on 24 January 2014 - 10:58 AM, said:

They aren't. Back before it was changed, people would farm wins/losses. Not damage. They could make more c-bills by simply joining game after game, waiting for it to launch, then either dying swiftly or simply exiting. Switch mechs, drop in another game, rinse, repeat.


Why give someone a win bonus if they weren't connected at the end of match? I mean it says it right there on the scoreboard...shouldn't be hard to implement.

Also, this does not change the fact people consider damage done = skill. It doesn't change the "scoring".

And so they discourage people form farming wins, but rank peoples based on it? lol

Quote

That's no longer effective because more c-bills are awarded for damage (i.e. participating) than just for the win or loss.


Again they could just not reward the disconnected player anything. And again, cbills should = match score. And most players believe damge done = more skill and deserves a higher score. meaning better aim, situational awareness, and tactics.

Quote

Sure. Doesn't mean it's smart. :ph34r: It's just the only publicly visible way to stroke their e-peens. If Elo rankings were public, I suspect people would care about them.


Well you've already publicy stated how you would "game" the system if you knew yours, and you vote for them going public....lmao

But guess what, match scores at the end of the match are already public. And thats how people want to be matched up, and that is how people wnat to get paid cbills.
Farming wins has nothing to do with them. Thats PGI's problem, like Russianwolf described. What PGI should do is not reward a win bonus for people who don't finish the game.

Don't penalize the community for their problem by making winning a bad thing, but then rank people based on winning....absurd.


Quote

Sure, absolutely. There are lots of things that need fixing, and I agree 100% that rewards are one of those things. Role warfare isn't fully implemented yet. Hopefully that will help balance out the rewards for the guy who stood on a hill and directed his teammates to victory. But right now that guy's not being appropriately rewarded, despite some of the efforts that PGI has made in that direction.


I have no idea what you mean by Role warfare, or why proper rewards for winning shouldn't be implemented now. Its the root of all PGI's complaints and problems. I've made a thread how MWO is the only game i've ever played where winning is a bad thing. And how you might get called names for winning match for your team lol

Now i've gotten two revelations in the past two days. They rank peole based on wins only in a random team game, and they don't give people bonuses for winning because people can just quit a match and farm wins.... WoW.

Edited by RichAC, 24 January 2014 - 11:16 AM.


#323 Roadkill

    Member

  • PipPipPipPipPipPipPipPipPip
  • 3,610 posts

Posted 24 January 2014 - 11:13 AM

View PostRussianWolf, on 24 January 2014 - 11:08 AM, said:

All done here.

No one has been able to give plausible explaination of how ELO is accurately working in this environment (because it isn't).

Several have been given. You're just ignoring them. MischiefSC explained it in great detail.

You not understanding the math != Elo not working.

#324 Roadkill

    Member

  • PipPipPipPipPipPipPipPipPip
  • 3,610 posts

Posted 24 January 2014 - 11:23 AM

View PostRichAC, on 24 January 2014 - 11:12 AM, said:

Why give someone a win bonus if they weren't connected at the end of match?

Because it isn't always the player's fault. My MWO client is fairly stable, but I have a couple of buddies who can barely make it through 2-3 matches before it crashes.

RichAC said:

Also, this does not change the fact people consider damage done = skill.

Doesn't make people right. :ph34r:

Would you rather be rated on your actual skill, or on what other people think your skill is?

RichAC said:

And so they discourage people form farming wins, but rank peoples based on it? lol

*sigh*

Are you even paying attention? NOT FARMING WINS. NOT FARMING DAMAGE. Farming c-bills.

RichAC said:

But guess what, match scores at the end of the match are already public. And thats how people want to be matched up, and that is how people wnat to get paid cbills.

You're the first person I've ever heard say this. Who are these mysterious "people" you seem to think back you up?

Most of the people I play with and against ignore match score. It just isn't relevant to anything. So no one cares about it. I get that you think it's awesomesauce, but I just don't see anyone else out there who does.

Which is not to say that I think your idea is dumb. I don't. I think using match score - or an improved version of it - is a good idea. But it needs to be relevant, and the current match score depends too heavily on damage to be relevant.

#325 Grits N Gravy

    Member

  • PipPipPipPipPipPip
  • 287 posts

Posted 24 January 2014 - 11:23 AM

View PostRussianWolf, on 24 January 2014 - 09:42 AM, said:

How can ELO rate my skill if I'm only 1/12th of my team?

The fact of the matter is that Elo scores are the single most accurate predictor of match outcomes. An Elo score is not just a numerical ranking of players. It doesn't measure player strength or performance. It's a precise expression of how likely you are to beat someone else, based on players they have previously beaten. The only thing your Elo score tell you is how many games out of 100 you are likely to win Vs that person.

So my Elo score tells me that if I have 150 more points that my opponent, I am likely to win 70% of my games Vs him, or that I will lose 3/10 games to him. Elo scores have a high level of confidence and accuracy even when teams are close in scores. The true skill white paper posted previously indicates that even in challenged sets Elo is accurate 60% of the time.

Elo is a stat like, batting average, ERA, QB rating ect, It's just in a different form. So why would Elo alone be better at predicting outcomes than say damage? Damage is circumstantial to a win or a loss, it is not the determinant factor, same with kills, ect. There are plenty of anecdotal accounts of people doing top damage and losing, or continually doing top damage and losing. So it tends to be not as accurate at predicting the out come of matches as Elo.

What really matters in these system is their predictive outcome, this is how you judge if they work. Does the outcome the matchmaker predict represent what's occurring. Ie are the people with 150 more points winning 70% of the time? Elo matchmaking alone does have tons of room for improvement. As It has a considerably high margin of error and is not helped out by the implementation of Matchmaking in MWO. IE using team average of Elo scores, and the average match spread being 175 elo points between teams.

MWO matchmaking could be made to function much better by taking steps to match teams with the same variance in Elo scores, or build teams in with a tighter variance of Elo scores, Tighten the population of Elo scores by reducing the K factor after a given set of matches, Apply a Delta to the K factor based on Damage over team Average, Kills over Team Average, or any other stat that is circumstantial to a win. All of which would make matchmaking more accurate.

While adopting a true skill or any other Bayesian form of Elo may yield some improvements to matchmaking, True skill is really just an Elo system with a certainty variable. It only takes into account wins, losses and ranking certainty to adjust ratings. Ie it adjust the K factor based on whether you've been upsetting high ranked players or losing to lower ranked. This is why it converges more quickly than Elo scores. Though it would be a good start it still would need to be implemented correctly. If it was simply copy and pasted here it wouldn't revolutionize the matching. As you would still be stuck with wide variance and large spreads between teams.

It's not that Elo as a theory is wrong. It's that implementation of it in MWO is incomplete and not well done. This stat should rile you up. The current average Elo spread between teams is 175 points. Which means when you press the ready button, there is a 50/50 chance that you will be on either a team that wins 73% of the time or wins only 23% of the times. That's why the game can feel so streaky. That and the high Elo variance allowed between teammates, leads to the majority of peoples gripes with Matchmaking.

Edited by Grits N Gravy, 24 January 2014 - 11:39 AM.


#326 MischiefSC

    Member

  • PipPipPipPipPipPipPipPipPipPipPipPip
  • The Benefactor
  • The Benefactor
  • 16,697 posts

Posted 24 January 2014 - 11:37 AM

View PostKhobai, on 24 January 2014 - 08:06 AM, said:


Battlevalue could work the same way as elo, and have a dynamic value. For example, the more games a particular mech or weapon wins, the higher its battlevalue becomes. So youd get battlevalue that autoadjusts over time based how successful mechs/weapons are in the game. Itd be pretty difficult if not impossible to game that type of system.


So now you're getting into where the matchmaker should be going. It's not so much battlevalue though because in MW:O things don't have the same direct value as they do in tabletop. Individual player skill with a given weapon is far, far more significant here than they are in TT being a big part. Some people just can't hit {Scrap} with a projectile, some people can rock the poptart like there's no tomorrow.

It's also all built around Elo because, again, YMMV with any mech or build but how you specifically play with a given build or even sort of build, that's more consistent.

So currently Elo is only split four ways, one for each weight class. They have already discussed adding more parameters in the future but that at least is a starting ground and it makes sense.

Elo is based off win/loss because at the end of the day, that's what truly matters. Everything you do - damage, kills, using chat, scouting, drawing attention, organizing an ambush, keeping your teammates alive, driving people out of cover, it all affects the odds of your team winning. It's the supreme meta result of all your behaviors. You can get damage and kills and be a burden on your team. So it all comes back to win/loss.

However, you'll have different mechs and builds you do that better with. Not just by weight but by chassis, even buy build and weapon. In fact you'll have specific team distributions on different maps and game modes you do that better in. As the matchmaker gets more complex it should be looking at (eventually) how you rock boxxors in mechs with JJs, multiple lasers and SSRMs on high-heat maps when you've got a team comprised at least 60% of people who do well in sniper builds and playing against a team that's LRM-boat heavy.

The awesome thing is that an Elo-based matchmaker can get way, way more precise than just BV would indicate. Wispy in a Jenner is much higher Elo than Wispy in a Raven 4X. The current system can't see that. BV may not even show that.

The problem with the more complex approach is two fold though.

1. It's far more complex to set up and tweak. 2

2. It requires a large playerbase to take advantage of or all that extra data is useless; it's just happy to have people roughly within 750 points of your roughly calculated value and roughly aligned tonnage.

Right now a few things would help.

Split pug/premade Elo scores. Of all the things that divide performance how someone plays with a team is the biggest in my opinion. Someone who premades well and drops pug is going to likely represent a huge gap in how they're rated vs how they play.

Use a Gaussian distribution curve instead of a logical one. That means the curve is fatter and flatter in the middle and dips more sharply on each end instead of a more gradual curve up and down. Statistical analysis indicates that in human performance we tend to rise sharply from 'new' to about 'average', and then the very top percentage where we are distinctly above average drops off sharply as well. This gives a better representation of players actual skill and makes for thicker bands to facilitate the next part, which is the most critical -

Match to a range, not a target, in Elo. Matches would be far superior if everyone on each team was within 100 or 150 points of each other, even if the difference in value between the two teams was higher. Instead of having the 1800 player on a team with a 1400 target draw in a 1000 skilled player or two 1200 skilled players, just let the 1800 guy be there and keep trying to pull in 1400s. If you can have someone between 1700 and 1900 on the other team that's great but at the end of the day if there is one thing we've absolutely seen time and again it's that 1 or 2 or 3 or even 4 good players can not carry a corresponding pile of less skilled players against a more balanced team distribution. Quit requiring veterans to carry absolute rookies. If there's some huge Elo ranked players, drop them in the highest Elo-band match you can and largely ignore their point impact on filling the rest of the team.

Make sense? No veterans + rookies = average. That doesn't work, it's the cause of so much drama. This lets higher Elo players play how they enjoy and not have to pull twice as hard every game. It stops dropping low Elo players into a higher Elo meat-grinder as filler where they just get crushed.

#327 MischiefSC

    Member

  • PipPipPipPipPipPipPipPipPipPipPipPip
  • The Benefactor
  • The Benefactor
  • 16,697 posts

Posted 24 January 2014 - 11:52 AM

View PostRussianWolf, on 24 January 2014 - 11:08 AM, said:

All done here.

No one has been able to give plausible explaination of how ELO is accurately working in this environment (because it isn't).

[/size]


Let me clarify that for you -

Nobody has been able to explain how statistically analysis and probability theory works to you in a way that you understand.

Fair enough. Complex topics that take a lot of effort and interest in to get a handle on. Elo absolutely does work - in fact if it didn't work it would mean that modern mathematics are based on flawed principles and everything from quantum physics to rocket science to reproductive medicine would not actually work and that we've had any success with any of it is, in fact, just wild dumb luck.

Here is the simple, short, basic version for you.

Every single action of a single factor (in this case a player), no matter how small or insignificant, can be measured by observing it over time. The environment in MW:O is actually pretty static and predictable. Only a few maps, 3 game modes, set team sizes that vary only rarely, roughly matched player skill and roughly matched mech tonnage.

In the matches that you play the only real constant however is you. That constant is the same between all players, thus that constant is the only value that can be reliably charted.

How many matches to find your convergence? Depends on your skill. 200-500 is pretty generous in the current system, allowing for how many variables are a bit skewed (pug/premade mixed Elo, teams involving high/low Elo to hit a target, etc).

It's not accurate. It can't be used anything like reliably to, say, put everyone who plays MW:O in a list and rank them by skill 1-10,000.

It is as roughly accurate as is possible with the current population and limitation of tools however. More players = more samples, more matches per player = more samples. More samples = more potential accuracy.

In time you could have an Elo based on how you perform with a specific loadout on a specific mech and with a specific breakdown of team members - it could even predict how you'd play against a certain sort of enemy team on different maps and in different game modes. There are Elo-based systems like TeamSkill that can take your win/loss metrics in one game and convert them, reliably, into another whole game.

Elo works. Are there changes needed for the matchmaker? Sure. Elo however works.

#328 RichAC

    Member

  • PipPipPipPipPipPipPip
  • 661 posts

Posted 24 January 2014 - 11:54 AM

View PostRoadkill, on 24 January 2014 - 11:23 AM, said:

Because it isn't always the player's fault. My MWO client is fairly stable, but I have a couple of buddies who can barely make it through 2-3 matches before it crashes.


Thats not our problem. Could be computer issues, but most likely connection issues. They will still get all the othe rewards, they just won't get the win. Not an issue imo. IF a player shows a disconnection right on the scoreboard. It should be easy for PGI to indentiy who doesn't get the win bonus.

Quote

Doesn't make people right. :unsure:

More accurate then thinking something who gets stomped on a server, ranking last, but cause he simply got lucky on a winning team, deserves a higher ELO. Oh I know, it will eventually work out. We just need faith in the math. refer to my sig.

Quote

Would you rather be rated on your actual skill, or on what other people think your skill is?


Well I believe stats tell my story, more then your ELO based on random team wins ever will. Which will always be trying to play catchup and which might never be as accurate. I rate my skill the same way profesional athletes are rated. Just because most computer nerds who don't follow sports disagree, doesn't mean they are right either.

Quote

Are you even paying attention? NOT FARMING WINS. NOT FARMING DAMAGE. Farming c-bills.


This is no excuse to totally ruin their own game. Who encouraged this moronic idea? Oh i know, you have friends that crash every other match and they should get a win bonus for match they didn't play? but farmers shoudln't? ya....I don't think so bud. So because of people with bad connections and farmers we should ruin it for everyone? When PGI could simply just not award win bonuses for people with disconnect next to their name, but only reward them for what they earned during their time actually playing?

We should make winning the worst thing you can do in this game? Thats a recipe for a failed game! And has been the biggest root issue of this game since I've started playing.

That would be like saying a guy deserves to score a run instead of a triple, because he tripped on the dirt or base at 3rd that a field hand didn't rake up or sweep well enough he still deserves to score... Its not his fault, but All he gets is the triple, he doesn't get the run scored. sorry.


Quote

You're the first person I've ever heard say this. Who are these mysterious "people" you seem to think back you up?


Most of the people I play with and against ignore match score. It just isn't relevant to anything. So no one cares about it. I get that you think it's awesomesauce, but I just don't see anyone else out there who does.

Which is not to say that I think your idea is dumb. I don't. I think using match score - or an improved version of it - is a good idea. But it needs to be relevant, and the current match score depends too heavily on damage to be relevant.


So you really believe noone cares about match score? meaning kills, assists or dmg done on the scoreboard? ummm I don't even know what to say to that man. ALmost everybody I've ever played with on teamspeak who says good job or well done, for a high dmg or high kills in game?

All i know is if PGI feels the same way you do, its no wonder this game is dying and doomed. I'd liike to think they are not that naive.

Edited by RichAC, 24 January 2014 - 12:11 PM.


#329 Grits N Gravy

    Member

  • PipPipPipPipPipPip
  • 287 posts

Posted 24 January 2014 - 11:54 AM

View PostMischiefSC, on 24 January 2014 - 11:37 AM, said:


So now you're getting into where the matchmaker should be going. It's not so much battlevalue though because in MW:O things don't have the same direct value as they do in tabletop. Individual player skill with a given weapon is far, far more significant here than they are in TT being a big part. Some people just can't hit {Scrap} with a projectile, some people can rock the poptart like there's no tomorrow.

It's also all built around Elo because, again, YMMV with any mech or build but how you specifically play with a given build or even sort of build, that's more consistent.

So currently Elo is only split four ways, one for each weight class. They have already discussed adding more parameters in the future but that at least is a starting ground and it makes sense.

The problem with the more complex approach is two fold though.

1. It's far more complex to set up and tweak. 2

2. It requires a large playerbase to take advantage of or all that extra data is useless; it's just happy to have people roughly within 750 points of your roughly calculated value and roughly aligned tonnage.

You could just apply a modifier to the K factor based on each piece equipment and chassis. Easy to use items increasing your K factor, thus your Elo goes up, so you have to play superior teammates more quickly. Track live data in background and make changes to dial in the system before you launch.

Quote

Split pug/premade Elo scores
Doing the same thing here would help too. Apply a modifier to K factors based on group size.
The beautyf of all these solutions is u can run them in background and tweak them before the go live, or use historical data and model them too. It doesn't take a lot of effort to do so, it's just a matter of focus.

#330 RichAC

    Member

  • PipPipPipPipPipPipPip
  • 661 posts

Posted 24 January 2014 - 12:04 PM

View PostGrits N Gravy, on 24 January 2014 - 11:54 AM, said:

You could just apply a modifier to the K factor based on each piece equipment and chassis. Easy to use items increasing your K factor, thus your Elo goes up, so you have to play superior teammates more quickly. Track live data in background and make changes to dial in the system before you launch.

Doing the same thing here would help too. Apply a modifier to K factors based on group size.
The beautyf of all these solutions is u can run them in background and tweak them before the go live, or use historical data and model them too. It doesn't take a lot of effort to do so, it's just a matter of focus.


what decides what piece of equipment is easy to use. Are ppc's easy to use? They are a meta build by top players, but do you consider them easy to use?

Any sort of battlevalue is nonsense. We are going to have weight limits, which affect skirmish more then anything. Lets see if that actually even changes anything for all the whiners before we make it more complicated and take it even further. I'm not so sure it will even make a difference anymore.

The biggest problem is premades versing random pugs, more then the fact they sometimes drop with 4 atlases. public cw lobby should help this though, minus the sync droppers. we shall see.

Edited by RichAC, 24 January 2014 - 12:12 PM.


#331 RussianWolf

    Member

  • PipPipPipPipPipPipPipPipPip
  • Legendary Founder
  • Legendary Founder
  • 2,097 posts
  • LocationWV

Posted 24 January 2014 - 12:05 PM

View PostGrits N Gravy, on 24 January 2014 - 11:23 AM, said:

The fact of the matter is that Elo scores are the single most accurate predictor of match outcomes. An Elo score is not just a numerical ranking of players. It doesn't measure player strength or performance. It's a precise expression of how likely you are to beat someone else, based on players they have previously beaten. The only thing your Elo score tell you is how many games out of 100 you are likely to win Vs that person.

So my Elo score tells me that if I have 150 more points that my opponent, I am likely to win 70% of my games Vs him, or that I will lose 3/10 games to him. Elo scores have a high level of confidence and accuracy even when teams are close in scores. The true skill white paper posted previously indicates that even in challenged sets Elo is accurate 60% of the time.

Elo is a stat like, batting average, ERA, QB rating ect, It's just in a different form. So why would Elo alone be better at predicting outcomes than say damage? Damage is circumstantial to a win or a loss, it is not the determinant factor, same with kills, ect. There are plenty of anecdotal accounts of people doing top damage and losing, or continually doing top damage and losing. So it tends to be not as accurate at predicting the out come of matches as Elo.

What really matters in these system is their predictive outcome, this is how you judge if they work. Does the outcome the matchmaker predict represent what's occurring. Ie are the people with 150 more points winning 70% of the time? Elo matchmaking alone does have tons of room for improvement. As It has a considerably high margin of error and is not helped out by the implementation of Matchmaking in MWO. IE using team average of Elo scores, and the average match spread being 175 elo points between teams.

MWO matchmaking could be made to function much better by taking steps to match teams with the same variance in Elo scores, or build teams in with a tighter variance of Elo scores, Tighten the population of Elo scores by reducing the K factor after a given set of matches, Apply a Delta to the K factor based on Damage over team Average, Kills over Team Average, or any other stat that is circumstantial to a win. All of which would make matchmaking more accurate.

While adopting a true skill or any other Bayesian form of Elo may yield some improvements to matchmaking, True skill is really just an Elo system with a certainty variable. It only takes into account wins, losses and ranking certainty to adjust ratings. Ie it adjust the K factor based on whether you've been upsetting high ranked players or losing to lower ranked. This is why it converges more quickly than Elo scores. Though it would be a good start it still would need to be implemented correctly. If it was simply copy and pasted here it wouldn't revolutionize the matching. As you would still be stuck with wide variance and large spreads between teams.

It's not that Elo as a theory is wrong. It's that implementation of it in MWO is incomplete and not well done. This stat should rile you up. The current average Elo spread between teams is 175 points. Which means when you press the ready button, there is a 50/50 chance that you will be on either a team that wins 73% of the time or wins only 23% of the times. That's why the game can feel so streaky. That and the high Elo variance allowed between teammates, leads to the majority of peoples gripes with Matchmaking.


I get that ELO is a rating system (not ranking) and that it works in some environments (chess, fixed team sports, etc.) with better accuracy. My point is the application in this environment does not give good accurate data, that is then fed into a matchmaker.

You say that it expresses my likelyhood of beating an opponent is x of 100 games.

I say in chess, if I face the same opponent 100 times ELO is a good predictor. number of variables = 2
If you use ELO to predict 2 chess opponents that have never met before, but have played some of the same people, not quite as good, but will likely be close. variables = 10-20 lets say.
If you try to use ELO to predict 2 chess opponents that have never played any of the same opponents, it gets a lot less accurate. variables = 100-1000 maybe.

In MWO you are starting with 24 variables. Go to 100 matches and you are approaching 2300 variables (if you are a constant). The more variables you add, the less likely you are to get accurate resulting data. The less likely it is to be a predictor.

Statistics backs this up every time. The more variables you have, the less accurate the data.

Secondly, no one has addressed the hypothetical that I gave.

ELO can rate an individual in an individual effort, or a team in a team effort. But I see no way for it to rate an individual in a team effort with any accuracy.

There simply is no one math equation that fits into every scenario and gives a proper outcome every time. ELO has its uses, but I don't think this environment is the right one for ELO.

#332 RussianWolf

    Member

  • PipPipPipPipPipPipPipPipPip
  • Legendary Founder
  • Legendary Founder
  • 2,097 posts
  • LocationWV

Posted 24 January 2014 - 12:21 PM

View PostMischiefSC, on 24 January 2014 - 11:52 AM, said:


Let me clarify that for you -

Nobody has been able to explain how statistically analysis and probability theory works to you in a way that you understand.

Fair enough. Complex topics that take a lot of effort and interest in to get a handle on. Elo absolutely does work - in fact if it didn't work it would mean that modern mathematics are based on flawed principles and everything from quantum physics to rocket science to reproductive medicine would not actually work and that we've had any success with any of it is, in fact, just wild dumb luck.


I realize that you don't know me. But when you start off like that, you've lost your audience. I've aced Calculus, Trigonometry, Physics, Statistics and Geometry. Understanding complex equations and systems is something I also do for a living.

I'll restate it very simply.

Math works. Always. BUT math can give flaw results when placed in the wrong environment.

Simple equation

Input a constant greater than 0. X^2 = Y and Y will always be greater than 1

Simple. Easy. Always works.

Change the environment where you allow numbers larger than 0 and the equation stops being 100% accurate.

.5^2=.25 oops, that's not greater than 1

Math is wrong? No. You just put the equation into the wrong environment for what you were working on. So it gave you flawed data. If you use the data and continue, that's your problem.

#333 Roadkill

    Member

  • PipPipPipPipPipPipPipPipPip
  • 3,610 posts

Posted 24 January 2014 - 12:44 PM

View PostRichAC, on 24 January 2014 - 11:54 AM, said:

Thats not our problem. Could be computer issues, but most likely connection issues. They will still get all the othe rewards, they just won't get the win. Not an issue imo. IF a player shows a disconnection right on the scoreboard. It should be easy for PGI to indentiy who doesn't get the win bonus.

Ah, I misunderstood you. I thought you were saying that if someone disconnects they should get no reward at all. No win/loss bonus I can agree with. That seems totally reasonable to me.

RichAC said:

More accurate then thinking something who gets stomped on a server, ranking last, but cause he simply got lucky on a winning team, deserves a higher ELO. Oh I know, it will eventually work out. We just need faith in the math. refer to my sig.

You don't need to have "faith" in math. Math is fact. Math is science. Math is not a religion where you just need to belieeeeeeeeve hard enough and it will be true. Math is true whether or not you like it, or believe it.

If someone gets lucky and gets a win they don't deserve, their Elo will go up briefly. Then they'll play another game and it will go back down. It is self correcting.

The same thing can happen with match score. I played in a game last night where - god knows why - the other team completely ignored me. I could have stood still in the middle of the fight and they'd have just kept going as if I wasn't there. I ended up with almost 800 damage in a goofball Shadowhawk build that should have sucked. I only made it to test maneuverability with an XL325, but I wrecked face through no skill of my own. A 3-yr old could have wrecked face like that while being ignored.

Luck exists. It can briefly affect an Elo rating, but it can also briefly affect whatever other stat you propose as well.

RichAC said:

Well I believe stats tell my story, more then your ELO based on random team wins ever will. Which will always be trying to play catchup and which might never be as accurate. I rate my skill the same way profesional athletes are rated. Just because most computer nerds who don't follow sports disagree, doesn't mean they are right either.

You rate your skill based on an inaccurate measurement because it makes you feel good. Fine. This is a game, after all. The goal is to have fun.

But it doesn't mean that you belieeeeeeeving really hard makes match score an accurate gauge of skill. Match score doesn't rate skill - it combines a combination of stats (that we don't fully understand) and relies heavily on damage scored. But that's not skill. And it's easy to game - much easier than it is to game an Elo rating.

RichAC said:

This is no excuse to totally ruin their own game. Who encouraged this moronic idea? Oh i know, you have friends that crash every other match and they should get a win bonus for match they didn't play? but farmers shoudln't? ya....I don't think so bud. So because of people with bad connections and farmers we should ruin it for everyone? When PGI could simply just not award win bonuses for people with disconnect next to their name, but only reward them for what they earned during their time actually playing?

We should make winning the worst thing you can do in this game? Thats a recipe for a failed game! And has been the biggest root issue of this game since I've started playing.

That would be like saying a guy deserves to score a run instead of a triple, because he tripped on the dirt or base at 3rd that a field hand didn't rake up or sweep well enough he still deserves to score... Its not his fault, but All he gets is the triple, he doesn't get the run scored. sorry.

Huh? You lost me again. WTF are you talking about?

RichAC said:

So you really believe noone cares about match score? meaning kills, assists or dmg done on the scoreboard?

No. I mean what I said, not what you want to twist in into.

No one cares about match score. Match score is an actual thing you know. It's on the end-of-game screen. No one cares about match score because it isn't tracked. Match score isn't, as you say, "kills, assists, or damage done." Those are different stats, and yes people do care about those. Match score probably even takes those into consideration - along with other things of which we have no idea - but match score is NOT the same kills, assists, and damage done.

People congratulate me after a match for doing lots of damage. They congratulate me after a match for getting lots of kills. They even congratulate me after a match for getting double-digit assists.

No one has ever said a word about match score. Never. I've never seen it mentioned after a game at all. It just isn't anything that anyone cares about. I couldn't even tell you what my best ever match score was... but I can tell you my best kills, my best damage, and my best assists.

I realize that you love match score and think it's awesomesauce. That's great. But no one else thinks about it because it isn't tracked.

If PGI were to start tracking it and record your average match score or something, then I agree with you that people would care about it. But right now they don't.

#334 A banana in the tailpipe

    Member

  • PipPipPipPipPipPipPipPipPip
  • The 1 Percent
  • 2,705 posts
  • Locationbehind your mech

Posted 24 January 2014 - 12:46 PM

Even my likes to post ratio is 50/50. Forums ELO is DOOOOOMED!!! Matchmaker needs a nerf with all these high rollin posters! :unsure:

Edited by lockwoodx, 24 January 2014 - 12:49 PM.


#335 Roadkill

    Member

  • PipPipPipPipPipPipPipPipPip
  • 3,610 posts

Posted 24 January 2014 - 12:54 PM

View PostRussianWolf, on 24 January 2014 - 12:21 PM, said:

Math works. Always. BUT math can give flaw results when placed in the wrong environment.

No. The math still gives the correct results. The results are just being used wrong.

Math is truth. Whether or not you can handle the truth; whether or not you can make proper use of the truth; those are different questions.

Elo works in this environment. This isn't the optimal environment, but it still works. It just takes longer to stabilize and has greater variability. But it still works.

If you use it wrong, though, or don't understand the conditions in which it has been used, then you can improperly interpret it and get bad results. Arguably, that is what's happening.

View Postlockwoodx, on 24 January 2014 - 12:46 PM, said:

Even my likes to post ratio is 50/50. Forums ELO is DOOOOOMED!!! Matchmaker needs a nerf with all these high rollin posters! :unsure:

BEST POST IN THE THREAD.

#336 RussianWolf

    Member

  • PipPipPipPipPipPipPipPipPip
  • Legendary Founder
  • Legendary Founder
  • 2,097 posts
  • LocationWV

Posted 24 January 2014 - 01:01 PM

View PostRoadkill, on 24 January 2014 - 12:54 PM, said:

No. The math still gives the correct results. The results are just being used wrong.

Math is truth. Whether or not you can handle the truth; whether or not you can make proper use of the truth; those are different questions.

Elo works in this environment. This isn't the optimal environment, but it still works. It just takes longer to stabilize and has greater variability. But it still works.

If you use it wrong, though, or don't understand the conditions in which it has been used, then you can improperly interpret it and get bad results. Arguably, that is what's happening.


BEST POST IN THE THREAD.

Thank you, you just agreed with my whole premise.

Change the semantics all you want, but you just said the same thing.

This is not the optimal environment for ELO to be used.

Thanks again.

#337 Roadkill

    Member

  • PipPipPipPipPipPipPipPipPip
  • 3,610 posts

Posted 24 January 2014 - 01:08 PM

View PostRussianWolf, on 24 January 2014 - 01:01 PM, said:

Thank you, you just agreed with my whole premise.

Change the semantics all you want, but you just said the same thing.

This is not the optimal environment for ELO to be used.

Thanks again.

No, I didn't. We differ in the conclusion. Your conclusion is that Elo doesn't work. My conclusion is that it does, in fact, still work, just not optimally. It takes longer to stabilize. There's greater fluctuation in your rating. But it still works.

#338 Abivard

    Member

  • PipPipPipPipPipPipPipPip
  • Shredder
  • 1,935 posts
  • LocationFree Rasalhague Republic

Posted 24 January 2014 - 01:08 PM

View PostRoadkill, on 24 January 2014 - 10:52 AM, said:

Been thinking about this.

At least until CW comes out, MWO is just a casual game. There's no point to winning or losing because there's nothing to win or lose.

Back in 8 vs 8 when it was relatively easy for 3rd party organizations to hold unsanctioned tournaments, people played hard in order to practice for those competitions. But it's really tough to organize those tournaments now so most of them have stopped. Which means there's no reason for serious players to practice. Which in turn means that most of the matches played are just people goofing off or grinding c-bills. (Or stroking their e-peens, but we'll ignore that for now.)

I think a lot of the frustration comes from the fact that there really is no point. We want there to be a point. We want MWO to be more than just a casual game used only to goof off. But right now that just isn't the case.

Hopefully CW will arrive in time to give the serious players a reason to come back and play again.


Which doesn't change the fact that it works.



You can not run a diesel engine underwater, so no it doesnt work, you need an electrical engine when submerged.

Just as you seem incapable of grasping that concept, you seem incapable of understanding why Elo that rates 1 on 1 competition does not work in a random team environment.

Not a single person says the equations are wrong, they say they do not apply to this situation.

Internal combustion engines work great, on the surface, but will kill every person on the sub if you try to run them under water(with out a snorkel)

Straw arguments are easy to win, maybe you should address the points people make instead of twisting them into what was not said.

#339 Roadkill

    Member

  • PipPipPipPipPipPipPipPipPip
  • 3,610 posts

Posted 24 January 2014 - 01:16 PM

View PostAbivard, on 24 January 2014 - 01:08 PM, said:

You can not run a diesel engine underwater, so no it doesnt work, you need an electrical engine when submerged.

Yeah, actually, you can. You just need to properly vent the exhaust.

It's actually the perfect analogy. Running a diesel engine underwater isn't optimal, but it can be done and works just fine if you take the proper precautions.

Using Elo in MWO's current environment isn't optimal, but it works just fine provided you understand the constraints.

Abivard said:

Straw arguments are easy to win, maybe you should address the points people make instead of twisting them into what was not said.

Pot, kettle, black.

#340 RichAC

    Member

  • PipPipPipPipPipPipPip
  • 661 posts

Posted 24 January 2014 - 01:17 PM

View PostRoadkill, on 24 January 2014 - 12:44 PM, said:

Ah, I misunderstood you. I thought you were saying that if someone disconnects they should get no reward at all. No win/loss bonus I can agree with. That seems totally reasonable to me.


Yes they still get the stats and points they accumulated, they just dont' make the finish line. Who disconnects is listed right there on the scoreboard, I don't know why they would rather punish the whole community.

Quote

You don't need to have "faith" in math. Math is fact. Math is science. Math is not a religion where you just need to belieeeeeeeeve hard enough and it will be true. Math is true whether or not you like it, or believe it.

If someone gets lucky and gets a win they don't deserve, their Elo will go up briefly. Then they'll play another game and it will go back down. It is self correcting.

The same thing can happen with match score. I played in a game last night where - god knows why - the other team completely ignored me. I could have stood still in the middle of the fight and they'd have just kept going as if I wasn't there. I ended up with almost 800 damage in a goofball Shadowhawk build that should have sucked. I only made it to test maneuverability with an XL325, but I wrecked face through no skill of my own. A 3-yr old could have wrecked face like that while being ignored.

Luck exists. It can briefly affect an Elo rating, but it can also briefly affect whatever other stat you propose as well.


The thing is, how long does this take before it eventually gets accurate? Apparently even you don't know.... We have to have faith its going to happen in a timely manner, and I believe it probably won't. Having to wait longer for a proper avg, because a system not designed for team games ranks people wrong for some matches is not acceptable. I'm sorry. Waiting for the system to "correct" a mistake that shouldn't have been made in the first place is unacceptable.

Quote

You rate your skill based on an inaccurate measurement because it makes you feel good. Fine. This is a game, after all. The goal is to have fun.


again i say its more accurate then rating my skill based on nothing but the random teams i've played with. Why this is even disputed, could only be because you don't see pc games the same way many see athletic sports. You don't realize the factors of teamates should have no bearing.

Quote

But it doesn't mean that you belieeeeeeeving really hard makes match score an accurate gauge of skill. Match score doesn't rate skill - it combines a combination of stats (that we don't fully understand) and relies heavily on damage scored. But that's not skill. And it's easy to game - much easier than it is to game an Elo rating.


Well what do you consider skill is the question maybe? I consider skill a combination of AIM, situational awareness, tactics, strategy, and teamwork. There is no way you are calculating any of these things with an ELO based on wins and losses. Especially in a random team game. I don't care how many team based pc games in this dying industry try the same thing.

Quote

Huh? You lost me again. WTF are you talking about?


I'm saying taking away incentives to win matches ruins the spirit of competition. It turns off most players with a sense of sportsmanship. It takes respect away from the game. It also makes a system rating players based on said wins contradicting and worthless.

Quote

No. I mean what I said, not what you want to twist in into.



lol...what you mean is what you say again on the first line of your next statement...>>

Quote

No one cares about match score. Match score is an actual thing you know. It's on the end-of-game screen. No one cares about match score because it isn't tracked. Match score isn't, as you say, "kills, assists, or damage done." Those are different stats, and yes people do care about those. Match score probably even takes those into consideration - along with other things of which we have no idea - but match score is NOT the same kills, assists, and damage done.

People congratulate me after a match for doing lots of damage. They congratulate me after a match for getting lots of kills. They even congratulate me after a match for getting double-digit assists.

No one has ever said a word about match score. Never. I've never seen it mentioned after a game at all. It just isn't anything that anyone cares about. I couldn't even tell you what my best ever match score was... but I can tell you my best kills, my best damage, and my best assists

I realize that you love match score and think it's awesomesauce. That's great. But no one else thinks about it because it isn't tracked.

If PGI were to start tracking it and record your average match score or something, then I agree with you that people would care about it. But right now they don't.



haha wow, did I touch a nerve or something with match score? But ya thats what I thought you said. My highest match score is 149 I believe, with 8 kills and 1050 dmg, not sure how many assists.

To me match score is just a summary of all the other stats combined. Kills, damage, assists etc... Not probaby, but most definitely they all contribute to your match score. People will say nice dmg, but sometimes not nice kills because maybe you didn't get any, or vice versa...assists etc...

Saying nice score just isn't specific enough, or memorable to an experience on the battlefield, but I for one am glad its there and I will personally use that to judge who had the best performance on a team.

What most people can agree on, is that people should not be ranked solely by one stat over the other, hence we have a match score. Its the reason it exists. Unfortunately, noone is actually ranked by that with the matchmaker.

They are only ranked/rated by how their random team does in general, regardless of their performance in the match, which is nonsense to anyone with any sports sense.

"Just because something is simple doesn't mean its wrong, and just because something is extremely complicated doesn't mean its practical."

Edited by RichAC, 24 January 2014 - 01:32 PM.






14 user(s) are reading this topic

0 members, 14 guests, 0 anonymous users