Jump to content

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


1911 replies to this topic

#1201 Modo44

    Member

  • PipPipPipPipPipPipPipPipPip
  • Bad Company
  • 3,559 posts

Posted 10 July 2014 - 07:02 AM

View PostCimarb, on 10 July 2014 - 06:03 AM, said:

I highly, HIGHLY dislike being measured by that randomness (aka W/L the only determining factor for Elo), and the arguments that say "it will average out and give you the right number" are horrible.

You have a misconception that "it will converge over time" equals "it will average over time. It does not. In other bad news, every other system is even worse -- more random for the player -- than Elo. Way to ignore reality.

#1202 Cimarb

    Member

  • PipPipPipPipPipPipPipPipPip
  • Caladbolg
  • Caladbolg
  • 3,912 posts
  • Twitter: Link
  • Twitch: Link
  • LocationA hop, skip and jump from Terra

Posted 10 July 2014 - 07:41 AM

View PostModo44, on 10 July 2014 - 07:02 AM, said:

You have a misconception that "it will converge over time" equals "it will average over time. It does not. In other bad news, every other system is even worse -- more random for the player -- than Elo. Way to ignore reality.

Way to add an insult on the end there.

I am not ignoring reality. I am complaining about a system that rates individuals by a single factor that is dependent on group results.

A much more accurate system would take individual contributions into account, such as cap time, spotting time, damage, and other role-based factors. Elo was designed to rate players in chess, an individual game with the only outcome being a win or loss - it did not matter how WELL you played, because it is either a win or loss regardless. It was then adapted to all sorts of other sports, but every time it is adapted it loses more and more purpose, which is what brings us back to MWO and its very poor implementation of the system.

#1203 Void Angel

    Member

  • PipPipPipPipPipPipPipPipPipPip
  • The Marauder
  • The Marauder
  • 6,601 posts
  • LocationParanoiaville

Posted 10 July 2014 - 08:22 AM

View PostModo44, on 10 July 2014 - 12:31 AM, said:

Just remember that when looking at current average Elo values -- even for people with thousands of matches -- you are looking at results skewed by constantly incoming 1300 Elo rookies.

As well as the deflationary/inflationary effects of players who leave - hopefully the net effect is still positive! =P

Cimarb, the Elo system isn't rating individuals by a single factor - it accounts for multiple factors, including the Elo of other players on the team. It does use these factors to extrapolate skill based on win-loss ratios, but that's not quite the same thing. In both MWO and chess, how well you play matters - Elo focused on statistics related to win/loss/draw because he needed the simplifying assumption to make the system manageable, not because only the results mattered.

Implementing a system based on analyzing more variables (such as the factors you listed) would certainly have the potential to be more accurate and to encourage role warfare. It would also be much more prone to emergent negative behaviors if players learned to game the system, and would consequently require more effort. It might be worth the trouble, or it might not, but its potential superiority doesn't make Elo an inadequate system for determining player ranking over time. The primary cause of Elo system breakage is players who improve faster than the system can adapt - which is sadly unlikely in MWO (and would be invisible to us anyway.)

#1204 carl kerensky

    Member

  • PipPipPipPipPipPip
  • Legendary Founder
  • Legendary Founder
  • 395 posts
  • LocationMoon Base Alpha

Posted 10 July 2014 - 08:24 AM

Posted by Slide

Hi Karl I see that you are on as I type, I hope I can catch you.

Time to kill (TTK) has been a hot topic for a while now, with various solutions including increasing IS hit points and burst fire AC's

I was wondering if a simpler solution was possible, this was my thinking.

Some stats would help here but it seems to me that most of my deaths are due to engine destruction through the CT, even when running XL engines. Currently engine health I believe is 15 points. If this were to be increased to say 50 points (it stands to reason a fusion reactor would be heavily shielded) then TTK would be greatly increased without changing any other fundamentals about the mech. I think you would take more crits to other systems (HS, weapons etc) before dying.

This would increase TTK (no more one shots) and also increase the immersion factor by having your mech fall apart around you more. As an added immersion and balancing factor, as your engine take more damage you can suffer other set backs like reduced speed, higher heat etc.

Just a thought, maybe you can pass it on to Paul/design.



View PostKarl Berg, on 27 April 2014 - 04:37 PM, said:


Indeed, I can bring this up with Paul. This is 100% in his domain though, so that's the best I can do.



The is a fantastic idea. This would create more brawling and endurance fighting that MWO sorely lacks. It would up the immersion levels dramatically. It would have a more chipping away effect on the mechs themselves creating better fighting. Please look into trying this Karl. At least test it out and see how it effects gameplay.

Ck

Edited by carl kerensky, 10 July 2014 - 08:27 AM.


#1205 Heffay

    Rum Runner

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

Posted 10 July 2014 - 08:24 AM

Sorry Cimarb, but Elo based strictly on W/L really is the best method for matchmaking. It provides the best games over time, even if individual ones aren't always perfect. You can't look at a bad match (or even a series of bad matches) and use that as evidence that Elo doesn't work.

#1206 DEMAX51

    Member

  • PipPipPipPipPipPipPipPipPip
  • Elite Founder
  • 2,269 posts
  • LocationThe cockpit of my Jenner

Posted 10 July 2014 - 08:54 AM

View Postcarl kerensky, on 10 July 2014 - 08:24 AM, said:

Posted by Slide

Hi Karl I see that you are on as I type, I hope I can catch you.

Time to kill (TTK) has been a hot topic for a while now, with various solutions including increasing IS hit points and burst fire AC's

I was wondering if a simpler solution was possible, this was my thinking.

Some stats would help here but it seems to me that most of my deaths are due to engine destruction through the CT, even when running XL engines. Currently engine health I believe is 15 points. If this were to be increased to say 50 points (it stands to reason a fusion reactor would be heavily shielded) then TTK would be greatly increased without changing any other fundamentals about the mech. I think you would take more crits to other systems (HS, weapons etc) before dying.

This would increase TTK (no more one shots) and also increase the immersion factor by having your mech fall apart around you more. As an added immersion and balancing factor, as your engine take more damage you can suffer other set backs like reduced speed, higher heat etc.

Just a thought, maybe you can pass it on to Paul/design.






The is a fantastic idea. This would create more brawling and endurance fighting that MWO sorely lacks. It would up the immersion levels dramatically. It would have a more chipping away effect on the mechs themselves creating better fighting. Please look into trying this Karl. At least test it out and see how it effects gameplay.

Ck

Carl,

I think you have a misconception about engine destructions and dying...

When you are killed it is because your CT (the entire component) has been reduced to 0HP, not the engine. Mech's are not killed when the engine itself is destroyed, they are only killed when the component housing the engine is destroyed. Currently, I believe engines themselves can be critically hit (as well as arm/leg actuators and such), but your mech doesn't suffer any negative consequences when these items are "destroyed."

So, raising the HP of the engines themselves won't do anything to increase TTK, because destroying an engine doesn't kill a 'Mech.

Edited by DEMAX51, 10 July 2014 - 08:55 AM.


#1207 Cimarb

    Member

  • PipPipPipPipPipPipPipPipPip
  • Caladbolg
  • Caladbolg
  • 3,912 posts
  • Twitter: Link
  • Twitch: Link
  • LocationA hop, skip and jump from Terra

Posted 10 July 2014 - 10:49 AM

View PostVoid Angel, on 10 July 2014 - 08:22 AM, said:

Cimarb, the Elo system isn't rating individuals by a single factor - it accounts for multiple factors, including the Elo of other players on the team. It does use these factors to extrapolate skill based on win-loss ratios, but that's not quite the same thing. In both MWO and chess, how well you play matters - Elo focused on statistics related to win/loss/draw because he needed the simplifying assumption to make the system manageable, not because only the results mattered.

Implementing a system based on analyzing more variables (such as the factors you listed) would certainly have the potential to be more accurate and to encourage role warfare. It would also be much more prone to emergent negative behaviors if players learned to game the system, and would consequently require more effort. It might be worth the trouble, or it might not, but its potential superiority doesn't make Elo an inadequate system for determining player ranking over time. The primary cause of Elo system breakage is players who improve faster than the system can adapt - which is sadly unlikely in MWO (and would be invisible to us anyway.)

How well you play does NOT matter, so long as your luck is good (or bad, depending on the direction you are going). See below for more on that.

ANY system can be gamed. The current system can be gamed, especially on a new(er) account (or alt account, often), because all you have to do is lose several games in a row to tank your Elo. That is a poor excuse for having a poor ranking system. In fact, the more stats you track, the more difficult it is to game it.

View PostHeffay, on 10 July 2014 - 08:24 AM, said:

Sorry Cimarb, but Elo based strictly on W/L really is the best method for matchmaking. It provides the best games over time, even if individual ones aren't always perfect. You can't look at a bad match (or even a series of bad matches) and use that as evidence that Elo doesn't work.

Not, it is not. It does not provide the best games over time. It may provide good AVERAGES over time, but it is an average of horribly imbalanced games, which is not "best".

For instance, if my team is matched up against your team 10 times. 5 of those times we completely stomp you 12-0. 5 of them you stomp us 12-0. We now both have a perfectly equal Elo rating, yet have played 10 absolutely horrible matches.

Another example is comparing an immensely fun, balanced 12-11 match (i.e. best MM result), to a "roflstomp" 12-0 match (i.e. the worst MM result. BOTH of these matches are a "Win", and Elo is adjusted the same for both, yet the quality of those matches is as polar opposite as can be.

#1208 DEMAX51

    Member

  • PipPipPipPipPipPipPipPipPip
  • Elite Founder
  • 2,269 posts
  • LocationThe cockpit of my Jenner

Posted 10 July 2014 - 10:50 AM

Karl,

Can you shed any light on the specifics of the functionality of the Enhanced Narc module? The module's description is very vague, and even Smurfy doesn't seem to know what it actually does. Here are my specific questions, raised in another thread (to which no one has yet replied - implying that nobody on the forums actually knows...)

Quote

The description of the Enhanced NARC module is very cryptic - what exactly does it do?

It reads "Increases speed, TAG duration, and EMP"

Lets start at the beginning - "increases speed"

Increases speed of what? The NARC projectile itself? or of Missiles being fired upon the NARCed target? I would assume it's the NARC projectile, but it's not clear. Furthermore, increases speed by how much?

Second, "TAG duration"

Why would the NARC module buff TAG? Unless by "TAG duration" they actually mean the NARC beacon's duration? And again, increases it by how much?

Lastly, and most confusing, "increases EMP"

As far as I know, a NARC will cancel out an enemy's ECM if it's fired directly on the ECM 'Mech, but how exactly is this effect "increased"? Is it given an area of effect, such that it can cancel any ECM within a certain range, even if it's not the ECM 'Mech that is directly NARCed? Does it just make the EMP function last for a longer duration? And, again, to what extent?

The function of this module desperately needs to be clarified. Even Smurfy doesn't list any actual values for the module.


#1209 Heffay

    Rum Runner

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

Posted 10 July 2014 - 10:55 AM

View PostCimarb, on 10 July 2014 - 10:49 AM, said:

How well you play does NOT matter, so long as your luck is good (or bad, depending on the direction you are going). See below for more on that.

ANY system can be gamed. The current system can be gamed, especially on a new(er) account (or alt account, often), because all you have to do is lose several games in a row to tank your Elo. That is a poor excuse for having a poor ranking system. In fact, the more stats you track, the more difficult it is to game it.


Not, it is not. It does not provide the best games over time. It may provide good AVERAGES over time, but it is an average of horribly imbalanced games, which is not "best".

For instance, if my team is matched up against your team 10 times. 5 of those times we completely stomp you 12-0. 5 of them you stomp us 12-0. We now both have a perfectly equal Elo rating, yet have played 10 absolutely horrible matches.

Another example is comparing an immensely fun, balanced 12-11 match (i.e. best MM result), to a "roflstomp" 12-0 match (i.e. the worst MM result. BOTH of these matches are a "Win", and Elo is adjusted the same for both, yet the quality of those matches is as polar opposite as can be.


A 12-0 match isn't necessarily a bad match. It could mean that of the 2 equally rated teams, one of them could have made a significant tactical mistake which resulted in a quick 0-4 deficit (scout lance running into the fast attack lance), at which point the odds of a 12-0 match is significantly greater. And it still could have been a great match depending on a hundred different factors.

And the odds of 2 equally rated teams having 5 0-12 and 5 12-0 matches is incredibly low, and yet another anecdote that is very unlikely to attempt to prove that Elo doesn't work.

But it does. Winning is the only thing that matters. And that applies across the board, from the 600 Elo pilots to the pros. It's the only metric that matters at the end of the match.

Edited by Heffay, 10 July 2014 - 10:57 AM.


#1210 Void Angel

    Member

  • PipPipPipPipPipPipPipPipPipPip
  • The Marauder
  • The Marauder
  • 6,601 posts
  • LocationParanoiaville

Posted 10 July 2014 - 11:02 AM

View Postcarl kerensky, on 10 July 2014 - 08:24 AM, said:

The is a fantastic idea. This would create more brawling and endurance fighting that MWO sorely lacks. It would up the immersion levels dramatically. It would have a more chipping away effect on the mechs themselves creating better fighting. Please look into trying this Karl. At least test it out and see how it effects gameplay.

Ck

I'm certain engines are only destroyed when the component dies - You'd see Atlases dying from a PPC to yellow structure, but that never happens.

I wholeheartedly agree that time to kill is an issue, but I think it's mostly due to the effects of pinpoint fire at range. Most times that a 'Mech exposes itself to shoot, it can take immediate fire from multiple enemies - particularly with the extremely long ranges offered by Clan technology. It's not unusual for my Banshee to have an Orange arm (and this is my 3S brawler with full armor) after the first time I get shot at. This in turn causes players to be very cautious - to the point where I've seen a full, three-ring circus of Clan 'Mechs all trying to snipe around the same corner. This long-range mentality is very hard on brawlers, particularly because brawler(s) who does flank the enemy's cover is exposed to multiple long-range 'mechs who are already under cover from enemy fire - and the long-range combatants on the brawler(s') team are usually too hesitant to move in support until it is far, far too late.

#1211 Cimarb

    Member

  • PipPipPipPipPipPipPipPipPip
  • Caladbolg
  • Caladbolg
  • 3,912 posts
  • Twitter: Link
  • Twitch: Link
  • LocationA hop, skip and jump from Terra

Posted 10 July 2014 - 11:13 AM

View PostHeffay, on 10 July 2014 - 10:55 AM, said:


A 12-0 match isn't necessarily a bad match. It could mean that of the 2 equally rated teams, one of them could have made a significant tactical mistake which resulted in a quick 0-4 deficit (scout lance running into the fast attack lance), at which point the odds of a 12-0 match is significantly greater. And it still could have been a great match depending on a hundred different factors.

And the odds of 2 equally rated teams having 5 0-12 and 5 12-0 matches is incredibly low, and yet another anecdote that is very unlikely to attempt to prove that Elo doesn't work.

But it does. Winning is the only thing that matters. And that applies across the board, from the 600 Elo pilots to the pros. It's the only metric that matters at the end of the match.

A 12-0 match is a BAD match. Whether it is from a mistake or absolutely horrible MM results means little. Equally matched teams should never lose that badly unless it is on purpose, and the sheer amount of matches that end with those results shows how poor the matchmaker is "working".

While the odds that it is perfect 5 stomps on either side is absolutely very low, the odds for that are exactly the same as for any other exact number. Whether it is 12-1, 12-2, 12-3, or whatever else does not matter other than the illustration of the "best" and "worst" matches possible.

You are just stonewalling though. I used the example as an example. Whether it is 1000 matches or 10, if the results are the same, the number of matches does not matter when used for an example. The fact that you want to use that as a cop out is telling.

Just to appease you, though, say it is 1000 matches. My team wins 500 of those matches at 12-0, and your team wins the other 500 matches at 0-12. None of those are "good" matches, as every one of them is horribly imbalanced (the specific reason for each result does not matter), yet the matchmaker thinks they are the exact same thing as a 12-11 "best outcome" match.

Winning is the "only thing that matters" only because it is the only thing that Elo currently tracks. If I have an amazing match, yet my team completely derped, why am I punished by lowering my Elo? It just does not make sense to have it set up that way.

#1212 Heffay

    Rum Runner

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

Posted 10 July 2014 - 11:28 AM

View PostCimarb, on 10 July 2014 - 11:13 AM, said:

While the odds that it is perfect 5 stomps on either side is absolutely very low, the odds for that are exactly the same as for any other exact number. Whether it is 12-1, 12-2, 12-3, or whatever else does not matter other than the illustration of the "best" and "worst" matches possible.


...

I don't think you're doing that right. With a perfectly balanced match (via Elo), the *most* likely ending score is 12 - 5.5. Well... since there are no half points, that's either 12 - 6 or 12 - 5. Every other finishing score is *less* likely than that. They are not all equally likely.

The only thing that is equally likely is that over a significantly large number of matches, the W/L distribution will be 50/50.

View PostCimarb, on 10 July 2014 - 11:13 AM, said:

Winning is the "only thing that matters" only because it is the only thing that Elo currently tracks. If I have an amazing match, yet my team completely derped, why am I punished by lowering my Elo? It just does not make sense to have it set up that way.


Because it's a team game, and your individual contributions are meaningless if you didn't win. There are no Participation ribbons being awarded here. You win or you die. Your brilliant match may have cost the team the game if you acted like a lone ranger instead of a member of the team.

The individual is meaningless in the grand scheme of things.

Edited by Heffay, 10 July 2014 - 11:32 AM.


#1213 Alaskan Nobody

    Member

  • PipPipPipPipPipPipPipPipPipPipPip
  • The Determined
  • The Determined
  • 10,358 posts
  • LocationAlaska!

Posted 10 July 2014 - 11:31 AM

View PostCimarb, on 10 July 2014 - 11:13 AM, said:

A 12-0 match is a BAD match.

I stopped reading at that point.

Not to be rude Cim, but the final score is only a small portion of the story.

I have had several many 12-0 matches that were awesome matches, despite the score.

I have had several 12-0 pug matches, that everyone on both sides spreading damage well enough that we had a 0-0 until the last few minutes, at which point the snowball started rolling.

If the final score is the only measure of a good match for you, I think I am VERY gland I have not dropped with you more. :P

#1214 Void Angel

    Member

  • PipPipPipPipPipPipPipPipPipPip
  • The Marauder
  • The Marauder
  • 6,601 posts
  • LocationParanoiaville

Posted 10 July 2014 - 11:41 AM

View PostCimarb, on 10 July 2014 - 10:49 AM, said:

How well you play does NOT matter, so long as your luck is good (or bad, depending on the direction you are going). See below for more on that.

ANY system can be gamed. The current system can be gamed, especially on a new(er) account (or alt account, often), because all you have to do is lose several games in a row to tank your Elo. That is a poor excuse for having a poor ranking system. In fact, the more stats you track, the more difficult it is to game it.

[Elo ranking] does not provide the best games over time. It may provide good AVERAGES over time, but it is an average of horribly imbalanced games, which is not "best".

For instance, if my team is matched up against your team 10 times. 5 of those times we completely stomp you 12-0. 5 of them you stomp us 12-0. We now both have a perfectly equal Elo rating, yet have played 10 absolutely horrible matches.

Another example is comparing an immensely fun, balanced 12-11 match (i.e. best MM result), to a "roflstomp" 12-0 match (i.e. the worst MM result. BOTH of these matches are a "Win", and Elo is adjusted the same for both, yet the quality of those matches is as polar opposite as can be.

It's simply not true that luck determines most matches over time. If you continue to play, your Elo will adjust over time to approximate that skill. This isn't an extraordinary claim; it's just now Elo systems work and can be observed to work. If you wish to dispute an established fact like this one, you'll have to cite evidence.

Any system can be abused, true - but a more complex system offers more opportunities to do so, not fewer. You're thinking of it like it was bits of entropy in a hash function; it's actually more like having more entrances to a base to guard. Such a system may still have its advantages, but if a player is willing to accept bad games they will be able to go Elo Skydiving - a more complex system simply lets them do it in more detail, while allowing opportunities for other abuse (such as inflating scores) or unintended consequences. Such a system may still be superior - if also more expensive and risky - but that doesn't make Elo a bad system.

Similarly, you're employing some faulty reasoning when you talk about "Rofflestomp" vs. "balanced" matches. You're assuming that score is a reliable determinant of how enjoyable a game was; and you're assuming that Elo has no relationship to player skill whatsoever. The problem on the first hand is that you can easily have fun matches where everyone is beat up, yet no one dies, - a game in which both sides jockey for position until one side makes a vital play can be a fun time had by all. On the second hand, it's not reasonable to suppose that two hypothetical teams of equal Elo will take turns sweeping each other - unless they weren't playing seriously, perhaps. In any case, such a situation should result in no net change - because the playing history hasn't told us anything about relative skill.

Correlation does not indicate cause - a 12v11 score does not indicate "the best matchmaker result," for instance, because many times that score will be the result of the better team making the first mistake and then clawing back from it - to win, or to lose. Because of the cumulative effect of focus fire - a tactic that all competent MechWarriors should employ whenever possible - the first team to lose a 'Mech of significant tonnage has a great disadvantage at the end of the match. This causes two equal teams to normally have a wider disparity in scores than 12v11.

View PostHeffay, on 10 July 2014 - 11:28 AM, said:

Because it's a team game, and your individual contributions are meaningless if you didn't win. There are no Participation ribbons being awarded here. You win or you die. Your brilliant match may have cost the team the game if you acted like a lone ranger instead of a member of the team.

This is what I meant about the potential for more complex systems to produce unintended (negative) consequences.

#1215 Wintersdark

    Member

  • PipPipPipPipPipPipPipPipPipPipPip
  • 13,375 posts
  • Google+: Link
  • Twitter: Link
  • LocationCalgary, AB

Posted 10 July 2014 - 11:56 AM

View PostHeffay, on 10 July 2014 - 10:55 AM, said:


But it does. Winning is the only thing that matters. And that applies across the board, from the 600 Elo pilots to the pros. It's the only metric that matters at the end of the match.


This.

As much as I respect your opinions Cim, you're wrong on this one.

As void says, it may be possible yo build a ranking system on stats, but it'd be extraordinary difficult (and more likely impossible).

How do you weight stats? How do you account for players getting "great matches" wracking up huge damage numbers, but not actually contributing to victory? What about players who regularly bring leadership, and even though they are mediocre pilots at best the excel at herding cats? I'd put money on a moderately organized team of 11 vs. 12 disorganized pugs any day. There are _countless_ other variables that would be very difficult to quantify, and many tracked stats that are potentially highly misleading.

I could run an LBX/LRM mech and wrack up stupid damage numbers, get a really high match score but co tribute far less to victory than I do in a pinpoint mech precisely crippling enemy mechs fast.

Winning (vs. The ranking of your opposing team) is all that matters. Its the only way that intangibles can be tracked.



And remember, lowing your Elo isn't punishing you. That's a very real reason why its hidden.

Your Elo score isn't an awesomeness rating.

#1216 Cimarb

    Member

  • PipPipPipPipPipPipPipPipPip
  • Caladbolg
  • Caladbolg
  • 3,912 posts
  • Twitter: Link
  • Twitch: Link
  • LocationA hop, skip and jump from Terra

Posted 10 July 2014 - 12:14 PM

View PostShar Wolf, on 10 July 2014 - 11:31 AM, said:

I stopped reading at that point.

Not to be rude Cim, but the final score is only a small portion of the story.

I have had several many 12-0 matches that were awesome matches, despite the score.

I have had several 12-0 pug matches, that everyone on both sides spreading damage well enough that we had a 0-0 until the last few minutes, at which point the snowball started rolling.

If the final score is the only measure of a good match for you, I think I am VERY gland I have not dropped with you more. :P

I wish English was not my first (and only language). If it was, I could blame my horrible communication skills on it. Regardless, something is getting lost in the translation.

Elo, as it currently stands, is ONLY determined by a win or loss. I did not use capping in my examples, because that throws a whole bunch of other arguments in the mix without adding anything to the discussion. With capping out of the discussion, the ONLY thing that determines a win or loss is the team that gets to "12" first (I use quotes because disconnects may mean it is not actually 12, but you know what I mean). That means the most balanced match, therefore the "best" result, would be a match where it ends 12-11. That is the closest possible match, meaning that it is the ideal outcome, even though it may not be common due to the "snowball effect".

On the flip side, the absolute worst possible result is 12-0, which is the most imbalanced score possible. It means that one team killed every other opponent without losing a single mech themselves. Damage is irrelevant, since Elo does not care what happens until that omnipotent tally mark is made (i.e. a kill). Whether you are at 1% (I realize it is probably not even possible) or 99% means nothing, as long as you are alive or dead - the only thing that matters is that tally mark.

That is not what determines a GOOD match, though. The W or L does not make it a good or bad match, though it does obviously impact that decision. What determines if it is a GOOD match is in the eye of the player. If they do an outstanding job, be it tons of damage, spotting all match long to bring the rain of death upon their enemies, or being a little ninja that successfully out caps the other team, THAT is what determines a good match. A win may or may not happen, but if you had a great match, you should be rewarded for that. Instead, all that matters is whether you got a W or L at the end. That is exactly why I dislike the Elo system so much.

I am obviously not communicating that appropriately, though, so I apologize.

#1217 Heffay

    Rum Runner

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

Posted 10 July 2014 - 12:26 PM

View PostCimarb, on 10 July 2014 - 12:14 PM, said:

On the flip side, the absolute worst possible result is 12-0, which is the most imbalanced score possible. It means that one team killed every other opponent without losing a single mech themselves. Damage is irrelevant, since Elo does not care what happens until that omnipotent tally mark is made (i.e. a kill). Whether you are at 1% (I realize it is probably not even possible) or 99% means nothing, as long as you are alive or dead - the only thing that matters is that tally mark.


The reason score isn't a good metric of "good" or "bad" matches: Take that 12-0 match. The losers all were averaging 80%+ health, but all got shot in the cockpit. The winning team's healthiest mech was running at 40% health with 1 PPC left among the 12 mechs remaining on the field and 14 working legs combined among them.

Was that a "bad" match? Seems like a pretty epic "Sure wish I was recording it no matter which side I was on" kind of match.

Since there is no way to quantify a "good" match, you default to the only thing that really matters: Winning.

#1218 Alaskan Nobody

    Member

  • PipPipPipPipPipPipPipPipPipPipPip
  • The Determined
  • The Determined
  • 10,358 posts
  • LocationAlaska!

Posted 10 July 2014 - 12:27 PM

View PostCimarb, on 10 July 2014 - 12:14 PM, said:

I am obviously not communicating that appropriately, though, so I apologize.

You are communicating well (I think)

Problem is, what does constitute a "good match"?

Nothing an automated system can control.

Damage?
High damage is not a always a good thing
Low damage is not always a bad thing

Surviving?
Does not always mean anything.

Only thing an automated system can work off of to determine what skill level people have is how much they contributed to winning.

None of the stats but the W/L can measure that.

You win a lot?
You get put against others who win a lot.

You lose a lot?
You get put against those who are also not likely to win.

Is it perfect?
No.
But it is a lot closer than any of the other stats.


The fact that this is a team game, and thus the individual means less gets brought up as a counter point (CONSTANTLY)
So work with your team.

You get a lousy team?
One match (50 matches even!) will not impact your score that much.

That worried about lousy teams?
Group. Up.

Bring competent teammates.
One player cannot carry a match by themselves.
Four can.
Even if the rest of their team is lousy - use them as a meat-shield then.
Their Elo goes up a bit - but if your group is what carried the game, guess what!
They will not be going up much without you.

IE: You will continue to go up, and they will not.

#1219 Cimarb

    Member

  • PipPipPipPipPipPipPipPipPip
  • Caladbolg
  • Caladbolg
  • 3,912 posts
  • Twitter: Link
  • Twitch: Link
  • LocationA hop, skip and jump from Terra

Posted 10 July 2014 - 12:33 PM

View PostWintersdark, on 10 July 2014 - 11:56 AM, said:

This.

As much as I respect your opinions Cim, you're wrong on this one.

As void says, it may be possible yo build a ranking system on stats, but it'd be extraordinary difficult (and more likely impossible).

How do you weight stats? How do you account for players getting "great matches" wracking up huge damage numbers, but not actually contributing to victory? What about players who regularly bring leadership, and even though they are mediocre pilots at best the excel at herding cats? I'd put money on a moderately organized team of 11 vs. 12 disorganized pugs any day. There are _countless_ other variables that would be very difficult to quantify, and many tracked stats that are potentially highly misleading.

I could run an LBX/LRM mech and wrack up stupid damage numbers, get a really high match score but co tribute far less to victory than I do in a pinpoint mech precisely crippling enemy mechs fast.

Winning (vs. The ranking of your opposing team) is all that matters. Its the only way that intangibles can be tracked.

And remember, lowing your Elo isn't punishing you. That's a very real reason why its hidden.

Your Elo score isn't an awesomeness rating.

I get that making a "better" system would be difficult. Making anything better is likely more difficult than taking the easy route. In fact, PGI has done that with many of their decisions, not just Elo. Revamping the heat system would have been a lot of work, so instead we got ghost heat. Implementing convergence is difficult, which is why we have the PPFLD meta still. Redoing all of the mech weapons to actually FIT the chassis correctly is more work, which is why the Awesome has itty bitty nippies, the Catapult is suffering from infEARiority syndrome, and the Spider somehow has an ERPPC tucked in its sleeve...

I totally understand the intangibles, just like I understand that the role rewards are not in place to judge lights and even mediums correctly right now. W/L is definitely a gauge of leadership, which is why it should be included in the stats that make up your Elo score. As a Dire Wolf, my job is to deliver as much damage and kill as many enemy mechs as possible - so damage and kills should be included in my Elo score as well. As a Locust, my job is to spot and cap - so spotting and capping bonuses should be included in my Elo score. As a Cicada, I may deliver some decent damage, but my primary goal is to cap and/or provide ECM coverage, so my Elo should reflect my success in those jobs. All mechs have some balance of these jobs and the pilot's ability to perform those jobs should be reflected in their Elo score.

I am not saying W/L does not matter - it does - but it is not the ONLY factor that matters, nor is it a good representation of an individuals' skill, which is why we currently have brand new players being lumped in with players like me with over 4,500 matches (and I'm sure I am probably not up to the number of matches any of you three are).

#1220 Heffay

    Rum Runner

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

Posted 10 July 2014 - 12:45 PM

View PostCimarb, on 10 July 2014 - 12:33 PM, said:

As a Dire Wolf, my job is to deliver as much damage and kill as many enemy mechs as possible - so damage and kills should be included in my Elo score as well.


No, that is not your job as a Dire Wolf. Your job is to do what the drop commander tells you to do. If he says go sit on Sigma and defend it from lights, THAT is your job. If you win the match with zero damage doing what the DC tells you to do, you did your "job" far better than the Dire Wolf who did 1400 damage and lost the match.

And the person who does his job better more often over time has a higher Elo score to reflect that even though he is doing an average of 100 points of damage in a Dire Wolf, his 70% win percentage shows that he's doing it right.

That is a better player than a high scoring player with a 30% win rate.

Edited by Heffay, 10 July 2014 - 12:46 PM.






2 user(s) are reading this topic

0 members, 2 guests, 0 anonymous users