Jump to content

- - - - -

Matchmaking Phase 4 Follow-Up - Feedback


277 replies to this topic

#61 Kageru Ikazuchi

    Member

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

Posted 21 April 2013 - 05:19 PM

View PostHauser, on 21 April 2013 - 02:06 PM, said:

Now if you want to know my criticism on the current implementation it would be dropping new players in the middle of the elo curve. This is frustrating for everybody. Dropping them at he bottom end with accelerated elo change would be a better solution.

I agree ... looking at the graph closely, it appears that the median is somewhere around 1275. If your Elo score starts at 1300 (as per the original Elo description), that means that the newest players are dropping with other new players, and average players who might have been playing since closed beta.

In the 1200-1350 range (the biggest spike), it looks like inexperienced players account for almost 2/3 of the player base. This may account for the perception that there are a large number of ROFL stomps.

It will be interesting to see how this develops over the next few months.

#62 Texas Merc

    Member

  • PipPipPipPipPipPipPipPip
  • The Patron
  • The Patron
  • 1,237 posts

Posted 21 April 2013 - 05:40 PM

As a side note, If you see Omid on the battlefield, run! Dudes a beast.

#63 Demesio

    Rookie

  • Knight Errant
  • 6 posts
  • LocationUS

Posted 21 April 2013 - 07:28 PM

I have seen more close matches since the patch and it is making the game much more interesting for me.

#64 Kageru Ikazuchi

    Member

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

Posted 21 April 2013 - 07:43 PM

All values that follow are approximation, based on an examination of the chart provided in the original Command Chair post ...
  • The average Elo score is approximately 1262
  • The vast majority of players (approx 64%) have Elo scores between 1195 and 1395 (12%: higher, 24%: lower)
  • A simliar majority of players (approx 64%) have Elo scores less than 1295

What I think this means ...
  • New players will lose most of their first games, because they're matched with above average players.
  • Only about one third of the players have an Elo higher than the base value
  • "Elo NOOB Hell" is a big friggin place ... until you break out into the top 12%, you're going to get new players in random matches.

Edit: formatting ... and adding this bit:

I think the Elo score needs to take the number of matches played into account ... after a certain number of matches, this number no longer becomes significant ...

This would help improve the games for both new and experienced players.

Edited by Kageru Ikazuchi, 22 April 2013 - 12:40 AM.


#65 Texas Merc

    Member

  • PipPipPipPipPipPipPipPip
  • The Patron
  • The Patron
  • 1,237 posts

Posted 21 April 2013 - 08:42 PM

I agree ^^ maybe after the cadet bonus wears off +10 matches? Also Elo is class based and I'm sure new players will be trying different classes. So maybe total it out at 50-100 games played?

Or new players start with a lower base Elo, which I think would be the most elegant solution.

#66 LordDante

    Member

  • PipPipPipPipPipPipPip
  • IS Exemplar
  • IS Exemplar
  • 782 posts
  • Locationmy Wang is aiming at ur rear... torso

Posted 21 April 2013 - 09:59 PM

soo from my personal Point off view ( beeing a medium skilled mechjokey ) ELO works just fine. the pugstomping is mostly gone. and even if a 0:8 happens its not that roffelstomp that its used to be . for me sitting in gemany that happend quite often late at night when i ran every second game into an american/canadian PRO premade that focused my Wang down ( my dwang hates to get to much attraction because he loves to strike through the backdoor <_< ) tl/dr elo works better and better !

#67 PhoenixFire55

    Member

  • PipPipPipPipPipPipPipPipPipPip
  • Legendary Founder
  • Legendary Founder
  • 5,725 posts
  • LocationSt.Petersburg / Outreach

Posted 21 April 2013 - 11:33 PM

View PostMystere, on 21 April 2013 - 09:58 AM, said:

Knowing right away the enemy composition leads to pre-determined tactical decisions, which I equate to boring, especially because you know that your team is almost like the enemy team. I much prefer not knowing the enemy composition until they are seen, as it forces teams to make tactical decisions on a more dynamic basis.


Not entirely true. For example ... on Alpine most of the times assaults, heavies and mediums go to epsilon for the main event and lights run around capping. IMO it is happening exactly because you don't know what enemy has, so you send max amount of mechs for the main fights leaving 1-2 to cap. Nobody sends heavies/assault to cap 1 by 1 because there is always a chance that a bunch of lights from the other will just group up and kill them 1 by 1.

Allow me to ask you ... how many times have you seen a team cave-rushing on Frozen City or Forest Colony? I remember times in closed beta when every 2nd match was a cave rush there. Nowadays nobody does that. I sense MUCH less tactical difference in matches today then it was back there when we knew exactly how many assaults/heavies/mediums/lights each team had.

#68 PhoenixFire55

    Member

  • PipPipPipPipPipPipPipPipPipPip
  • Legendary Founder
  • Legendary Founder
  • 5,725 posts
  • LocationSt.Petersburg / Outreach

Posted 21 April 2013 - 11:59 PM

View PostHauser, on 21 April 2013 - 02:06 PM, said:

Now you're treating elo as a personal ranking system rather then a component of the match making system. The critical difference is that while the match maker does require some measure of skill to assign players to a match, it does not have to be an accurate assessment of that person. The only thing the matchmaker needs is a rating that describes how the player plays.


Sorry but I'll call this bullsh*t. Rating that describes how the player playes = personal ranking system. If you don't have an accurate assessment you don't have accurate matchmaking. You can say that there is a good chance that 4 people on the team are 'overrated' by Elo and 4 others are 'underrated' so that combined Elo score is more or less accurate, but I'll say that there is just the same chance that all 8 are either 'under' or 'over' that leades to a heavily unbalanced game we see so often now.

View PostHauser, on 21 April 2013 - 02:06 PM, said:

The situation you describe is already highly improbable in a random match making scenario. You must be the (un)luckiest player in the world to drop with such good/bad people. However if both are consistently dropping with 4 friends that are significantly better or worse, then their ratings are deserved. It accurately describes how the whole group will preform.


Uhu ... and then when that guy that usually drops with his group decides to go PUG and become an epic fail for his team because he'll never play it according to his undeserved Elo. What I meant is ... one person does like 700-800 damage in a game while his teammates do less then 100 each while team loses. Let me ask you ... does this one guy deserve and increse of his Elo score or a decrease? Remember that Elo rates HIM and not his team.

View PostHauser, on 21 April 2013 - 02:06 PM, said:

As I explained previously to Christof, given a random match maker good players will encounter more matches where their team has the advantage. The good players team will always have 1 good and 7 randoms, while the other team has 8 randoms and thus a lower chance of getting equal or more good players.


Put 4 'good' players and 4 'bad' players against 8 'average' players and 'averages' will always win. 4 can't win vs 8. If Elo is working you'll never get 1 'good' and 7 'averages' vs 8 'averages', you'll always get 1 'bad' who is usually bad enough to ruin you. And the problem is you don't see peoples Elo scores, so you can't tell if he is 'bad' or 'good' before you die and start spectating. You count on him to cover you and all he does is strip your rear armor.

View PostHauser, on 21 April 2013 - 02:06 PM, said:

Elo is an improvement on random match making because it puts the players who would lose allot w/ random match making with similar players, just as it puts people who would win allot together. The result being that for every


In other words noobs play with noobs and learn nothing. Noob friendly game, you are gonna win 50% of your games even if you are a complete *****? Is that what we need? Don't know about you but I'd rather see a game that motivates people to improve their skill to win more games. Right now there is more reason to 'noobify' yourself to play in easy mode vs non-so-good players.

Also... you are saying that now pretty much everybody wins 50% of their games. If everyone does then everyones Elo is just there around that middle 50% win zone (just as graph shows). It is just as I was saying, has nothing to do with judging your skill, just your win/loss ratio.

View PostHauser, on 21 April 2013 - 02:06 PM, said:

I'm not defending the Elo system. I'm explaining how it works. I'm trying to explain that in the long run, if you play enough games, the goods and the bads will level out against each other. The only constant contributing factor that can steer your elo up or down is you.


It works for solo-play olny like in chess for which it was designed. You can use Elo to rate teams with permanent roster. You can't use Elo to rate idividuals from the team or teams with changing roster. I'm yet to hear from you why you do not agree with Elo that is based off the game score.

View PostHauser, on 21 April 2013 - 02:06 PM, said:

Now if you want to know my criticism on the current implementation it would be dropping new players in the middle of the elo curve. This is frustrating for everybody. Dropping them at he bottom end with accelerated elo change would be a better solution.


Nice to see we agree on smth at least <_< PGI will surely need an interactive tutorial at some points because their current 'tips' are kinda useless and sometimes hard even for veterans to figure out. Like 'If you want to disappear from enemy radar press 'P' to power down'. I already see newbs going into water on Forest and powering down in the open when they get 'incoming missiles' and a gazillion of PPC coming at them.

#69 PhoenixFire55

    Member

  • PipPipPipPipPipPipPipPipPipPip
  • Legendary Founder
  • Legendary Founder
  • 5,725 posts
  • LocationSt.Petersburg / Outreach

Posted 22 April 2013 - 12:04 AM

View PostKageru Ikazuchi, on 21 April 2013 - 07:43 PM, said:

- "Elo NOOB Hell" is a big friggin place ... until you break out into the top 12%, you're going to get new players in random matches.


Not a problem when there is an equal amount of them in each team. A big problem if its not. We surely need a 'number of games' stat to separate new players with preset average score from veteran players who just have average game score.

#70 Kageru Ikazuchi

    Member

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

Posted 22 April 2013 - 12:54 AM

A bit of clarification from my previous post ...

The graph depicts "more than 50 matches" as the darker part of the curve ... in order to improve the new player experience, my recommendation would be that Elo start at about 1000 for new players, with each match played (up to 50 matches, since that's the metric they're using) automatically increasing the base Elo by 6 points, in addition to the win/loss adjustment.

This should spread the new player distribution from the 1095-1395 range to 795-1395, flattening the light blue spike quite a bit, and more often matching new players with others of less experience and skill until they have had a chance to learn the game.

Edit ... maths is hard ... :o

Edited by Kageru Ikazuchi, 22 April 2013 - 05:42 AM.


#71 hammerreborn

    Member

  • PipPipPipPipPipPipPipPipPip
  • Legendary Founder
  • Legendary Founder
  • 3,063 posts
  • LocationAlexandria, VA

Posted 22 April 2013 - 03:26 AM

View PostPhoenixFire55, on 21 April 2013 - 11:59 PM, said:


It works for solo-play olny like in chess for which it was designed. You can use Elo to rate teams with permanent roster. You can't use Elo to rate idividuals from the team or teams with changing roster. I'm yet to hear from you why you do not agree with Elo that is based off the game score.


Because game score is an utterly worthless stat that is almost entirely influenced by damage, a worthless number to go by in terms of skill. LRM boats would be considered gods amoung men, poptards that know what they are doing would be below average, and anyone who plays a light properly would be near the bottom of the barrel.

#72 PhoenixFire55

    Member

  • PipPipPipPipPipPipPipPipPipPip
  • Legendary Founder
  • Legendary Founder
  • 5,725 posts
  • LocationSt.Petersburg / Outreach

Posted 22 April 2013 - 07:16 AM

View Posthammerreborn, on 22 April 2013 - 03:26 AM, said:

Because game score is an utterly worthless stat that is almost entirely influenced by damage, a worthless number to go by in terms of skill. LRM boats would be considered gods amoung men, poptards that know what they are doing would be below average, and anyone who plays a light properly would be near the bottom of the barrel.


I've seen a jenner to 800+ damage in the age of ECM and streaks. Most assault pilots can't even come close to that. LRMs are doing what they should now, not *redacted* damage they were doing before. I never said that game score should stay what it is, its up to PGI to improve it. But as I said, 99% of games I've seen people who actually play well end up high in game score, lights, mediums, heavies and assaults alike. Making Elo based off game score makes much more sense then wins/losses.

#73 hammerreborn

    Member

  • PipPipPipPipPipPipPipPipPip
  • Legendary Founder
  • Legendary Founder
  • 3,063 posts
  • LocationAlexandria, VA

Posted 22 April 2013 - 08:26 AM

View PostPhoenixFire55, on 22 April 2013 - 07:16 AM, said:


I've seen a jenner to 800+ damage in the age of ECM and streaks. Most assault pilots can't even come close to that. LRMs are doing what they should now, not *redacted* damage they were doing before. I never said that game score should stay what it is, its up to PGI to improve it. But as I said, 99% of games I've seen people who actually play well end up high in game score, lights, mediums, heavies and assaults alike. Making Elo based off game score makes much more sense then wins/losses.


Funny, because I win every match with a game score of 10 because I do my job as a light and capture the base. Are you saying the 7 other people who died on my team but did 200 damage are better than me because their match score thats entirely based on damage says so?

#74 Tickdoff Tank

    Member

  • PipPipPipPipPipPipPipPipPip
  • Bad Company
  • Bad Company
  • 2,647 posts
  • LocationCharlotte NC

Posted 22 April 2013 - 08:48 AM

View PostRedoxin, on 20 April 2013 - 01:39 PM, said:

Win/loss is the result of personal skill. These 2 are strongly related. You wont win EVERY game, but you can strongly increase your teams chances to win. Over a larger number of games, this higher win chance will accurately be reflected by a higher number of wins which will again result in higher Elo.
Of course people always fault others and not themselves for any loss. Thats human nature, but irrational at the same time.



I would consider Elo to be much more viable if we actually had true 8v8 fights. I kept track of the last 40 matches I did and in 29 of them there was at least 1 person either DCd or "AFK". (Meaning listed as disconnected or they never moved from the starting zone and did not do anything the entire match).

Once they can get the DC/AFK issues under control (note that I did not say "fixed", we will always have afks/dcs) the Elo may be more effective.

In 10 matches one team was down by 3 or more people. In 1 of those 10 the team I was on had 2 DCs and 3 people did not move or defend themselves, so it was 3v8. We got slaughtered (as expected). It is hard to believe that Elo can give an accurate rating with this sort of team imbalance.

Note: I am sure that I forgot a few matches (both good and bad) but I was trying to track my win/loss across different weight classes along with damage done. I was curious as to how my performance was impacting my teams. I did not keep track of which drops were in a premade and which were solo, I jump around a lot.

#75 Chaos7

    Member

  • PipPipPipPipPip
  • FP Veteran - Beta 1
  • 133 posts
  • LocationUK

Posted 22 April 2013 - 09:11 AM

Why don't you take into account the amount games played when you calculate the Elo rating ? Obviously mid-range Elo with 0 games is potentially far worst than mid-range Elo with 1500 games, no ?

#76 PhoenixFire55

    Member

  • PipPipPipPipPipPipPipPipPipPip
  • Legendary Founder
  • Legendary Founder
  • 5,725 posts
  • LocationSt.Petersburg / Outreach

Posted 22 April 2013 - 10:34 AM

View Posthammerreborn, on 22 April 2013 - 08:26 AM, said:

Funny, because I win every match with a game score of 10 because I do my job as a light and capture the base. Are you saying the 7 other people who died on my team but did 200 damage are better than me because their match score thats entirely based on damage says so?


It is a problem of a current game score that it gives crap for base capping, but it is always your choice to go basecap instead of going and shooting stuff. Thing in question here aren't the details of the game score system tho, but rather Elo matchmaking. Thing you can't deny tho is that if you just go and basecap you didn't really do anything, its other team fault to just let you come and cap. If smb comes to defend base and you kill him and then cap you'll be 1st in game score.

#77 hammerreborn

    Member

  • PipPipPipPipPipPipPipPipPip
  • Legendary Founder
  • Legendary Founder
  • 3,063 posts
  • LocationAlexandria, VA

Posted 22 April 2013 - 10:41 AM

View PostPhoenixFire55, on 22 April 2013 - 10:34 AM, said:


It is a problem of a current game score that it gives crap for base capping, but it is always your choice to go basecap instead of going and shooting stuff. Thing in question here aren't the details of the game score system tho, but rather Elo matchmaking. Thing you can't deny tho is that if you just go and basecap you didn't really do anything, its other team fault to just let you come and cap. If smb comes to defend base and you kill him and then cap you'll be 1st in game score.


Apparently winning the game is doing nothing. Get a load of this guy.

And if you kill a single guy before capping you still end up last, but with a 20-30 game score instead of 10. Meanwhile poptard mcpopperson doing 500 damage lookin like a fool jumpin up and down gets a 50.

#78 Hauser

    Member

  • PipPipPipPipPipPipPip
  • Legendary Founder
  • Legendary Founder
  • 976 posts

Posted 22 April 2013 - 10:52 AM

View PostPhoenixFire55, on 21 April 2013 - 11:59 PM, said:

Sorry but I'll call this bullsh*t. Rating that describes how the player playes = personal ranking system. If you don't have an accurate assessment you don't have accurate matchmaking. You can say that there is a good chance that 4 people on the team are 'overrated' by Elo and 4 others are 'underrated' so that combined Elo score is more or less accurate, but I'll say that there is just the same chance that all 8 are either 'under' or 'over' that leades to a heavily unbalanced game we see so often now.

Uhu ... and then when that guy that usually drops with his group decides to go PUG and become an epic fail for his team because he'll never play it according to his undeserved Elo. What I meant is ... one person does like 700-800 damage in a game while his teammates do less then 100 each while team loses. Let me ask you ... does this one guy deserve and increse of his Elo score or a decrease? Remember that Elo rates HIM and not his team.

Put 4 'good' players and 4 'bad' players against 8 'average' players and 'averages' will always win. 4 can't win vs 8. If Elo is working you'll never get 1 'good' and 7 'averages' vs 8 'averages', you'll always get 1 'bad' who is usually bad enough to ruin you. And the problem is you don't see peoples Elo scores, so you can't tell if he is 'bad' or 'good' before you die and start spectating. You count on him to cover you and all he does is strip your rear armor.

In other words noobs play with noobs and learn nothing. Noob friendly game, you are gonna win 50% of your games even if you are a complete *****? Is that what we need? Don't know about you but I'd rather see a game that motivates people to improve their skill to win more games. Right now there is more reason to 'noobify' yourself to play in easy mode vs non-so-good players.

Also... you are saying that now pretty much everybody wins 50% of their games. If everyone does then everyones Elo is just there around that middle 50% win zone (just as graph shows). It is just as I was saying, has nothing to do with judging your skill, just your win/loss ratio.

It works for solo-play olny like in chess for which it was designed. You can use Elo to rate teams with permanent roster. You can't use Elo to rate idividuals from the team or teams with changing roster. I'm yet to hear from you why you do not agree with Elo that is based off the game score.


Let me put it another way then. You're treating elo as a digital e-peen rather then a part of the match maker. The only thing you should measure the effectiveness of elo by is the quality of the matches it helps the matchmaker produce. If a rating is deserved or not is utterly irrelevant.

You misunderstand how the match maker works. The match maker starts with a specific elo level in mind. Then it starts looking for players who are at this level. It does not try to average out the elo level for the whole team.
I think you might be confused by my use of a random match maker. This is a hypothetical match maker used to show that it is possible to get an indication of how much a player contributes to winning the game even though he is part of a team. The basic idea is that because the player is good, he will win more on average in the long run. But please read my original post to Krystof about that and perhaps the original posts about Elo.
The same argument can be used for an elo match maker as it selects random people near your rating. If you're still winning it's because you're better then them and should go up. If you're leveling out you have the correct rating. I do however prefer tot use the random match maker as an example because it cuts out most of the moving parts that make it hard to understand.

Unfortunatly you also seem to misunderstand how elo works. If you win a close game you go up a little. If you win a game that was stacked against you, you go up allot. If you lose nothing may even change. As such your winloss ratio is not an indication of your elo rating, nor is it used to determine your elo rating. At best it can be used as weak indication of how your elo rating is changing over time.

The end result is indeed that, noobs will play with noobs if they manage to hang on long enough. And this is what we need because you dont want to play with them (on your team) and they don't want to play with you (on the other team).

The end result the system is aiming for is that indeed you will win 50% of your games. More if you're improving, less if you're getting worse.

It's called a fair fight.

Now I regret to see that you think this is a bad environment for new players. I regret too see that you think that a noob friendly environment is a bad thing. I regret to see that you suggest trashing your elo and stomping on noobs so you can have an easy experience. An attitude like this prevents the game from growing.

Also do me a favour, if you are going to type this much please make a coherent argument. I'm not interested in replying to every mismatched paragraph.

Edited by Hauser, 22 April 2013 - 10:53 AM.


#79 Symbiodinium

    Member

  • PipPipPipPipPip
  • Veteran Founder
  • Veteran Founder
  • 162 posts
  • LocationNew Zealand

Posted 22 April 2013 - 10:53 AM

Until the ubiquitous disconnects are solved, I don't see Elo meaning much. There's usually at least one disconnected player in each game, which often hands the game to the other team. In my opinion if a team has a player DC'd before the game starts then the Elo result should be voided for that game, as it's no longer statistically valid.

#80 PhoenixFire55

    Member

  • PipPipPipPipPipPipPipPipPipPip
  • Legendary Founder
  • Legendary Founder
  • 5,725 posts
  • LocationSt.Petersburg / Outreach

Posted 22 April 2013 - 10:55 AM

View Posthammerreborn, on 22 April 2013 - 10:41 AM, said:

Apparently winning the game is doing nothing. Get a load of this guy.

And if you kill a single guy before capping you still end up last, but with a 20-30 game score instead of 10. Meanwhile poptard mcpopperson doing 500 damage lookin like a fool jumpin up and down gets a 50.


Lets put it this way ... McPopperson (and others) kept their entire team busy long enough for you mr 3L-son to go and cap.





1 user(s) are reading this topic

0 members, 1 guests, 0 anonymous users