83 replies to this topic

[Hotthedd]

Heffay, on 26 January 2015 - 06:54 AM, said:

Yes, it should be just W/L. Winning isn't everything. It's the *only* thing that matters.

The Direwolf who does 0 damage sitting on a cap point during conquest to secure it and help his team win is more valuable than the one who ran off, did 1600 points of damage with 8 kills while ignoring the objectives.

Damage doesn't mean squat if you lose the war.

I would think that a Direwolf capping in conquest, and not doing any damage, did nothing to help the team win.
W/L needs to be the biggest component, but not the ONLY component, due to the many good reasons already given. In chess (Elo given), a player rated 775 would not lose rating points by losing to a Grand Master, and the Grand Master would not GAIN points from beating a beginner.
I think your match score, weighted against the combined Elo of the enemy team, with a bonus for a win would be the best metric.

[Davers]

Hotthedd, on 26 January 2015 - 08:29 AM, said:

I would think that a Direwolf capping in conquest, and not doing any damage, did nothing to help the team win.
W/L needs to be the biggest component, but not the ONLY component, due to the many good reasons already given. In chess (Elo given), a player rated 775 would not lose rating points by losing to a Grand Master, and the Grand Master would not GAIN points from beating a beginner.
I think your match score, weighted against the combined Elo of the enemy team, with a bonus for a win would be the best metric.

Unless the team won on caps that is...

[Hotthedd]

Davers, on 26 January 2015 - 08:36 AM, said:

Unless the team won on caps that is...

No.
MW:O is a TEAM game. Players have roles. The team may win on caps in that scenario, but that would have very little to do with the DW in question. Had the DW done ITS job, a faster mech may have survived to cap, or at least been freed to cap.
Tom Brady is a great quarterback, but I would never use him to return punts.

[Davers]

Hotthedd, on 26 January 2015 - 08:41 AM, said:

No.
MW:O is a TEAM game. Players have roles. The team may win on caps in that scenario, but that would have very little to do with the DW in question. Had the DW done ITS job, a faster mech may have survived to cap, or at least been freed to cap.
Tom Brady is a great quarterback, but I would never use him to return punts.

It's funny that you picked quarterback, as it is a very versatile role. He has to be proficient at both throwing and running plays himself.

[Hotthedd]

Davers, on 26 January 2015 - 08:49 AM, said:

It's funny that you picked quarterback, as it is a very versatile role. He has to be proficient at both throwing and running plays himself.

Yes, because he probably COULD return punts. However, returning punts would be a risk at best, and an utter waste at worst. The QB has a job to do in order to have the best chance to win the game, and returning punts, while important, ain't it.
The same thing can be said about the DW in question.

[Heffay]

What if that Direwolf was also the drop caller, picked an overwatch spot where he could defend 2 cap points, and therefore funneled the enemy into predetermined kill zones to be picked apart by his team?

There are FAR too many factors that go into winning a match other than herpaderpa damage. That's why you base it on whether you win or lose. The most valuable unit on the field isn't necessarily the one that has the highest score in anything on the scoreboard.

Winning isn't everything. It's the *only* thing.

[Joseph Mallan]

Heffay, on 26 January 2015 - 07:40 AM, said:

*sigh*

It does Mr Mallan. Your individual stats mean nothing if you lose the battle. And if you consistently have phenomenal stats, odds are it will *influence* the battle and help you win, which will lead to your Elo rising. If you have a 2200 Elo score, odds are you're going to be a pretty hard hitter in the game, because that does help you win. But you could have the same stats as someone with an 1800 Elo score. The difference? The 2200 player pays attention to objects a *lot* more than the 1800 rated person.

My individual stats mean everything to how good a Player I am. They show if I can bring the heat when needed, or to make the sacrifice when it called for.

Billy Sims one of football's best running backs could not carry the Lions to the superbowl. What could he have done were he on the Cowboys or Any other top rated team of teh time?

[Mercules]

Heffay, on 26 January 2015 - 07:35 AM, said:

Elo does represent the skill of the player. It takes *all* factors into account, such as mech, ability, computer, experience, yadda yadda yadda. It normalizes everything based on your ability to use all those factors into accomplishing the main objective: Winning the match.

Over thousands of matches, yes it would. However you are not playing thousands of matches with ideal conditions for measuring such a thing. Remember that is thousands of matches for each weight class as each has a different Elo. So if you played the same mech with the same loadout for thousands of matches you would get an accurate Elo score for you in that mech, but then if you switched from say, a Trebuchet to a Hunchback properly using it's quirks the Elo score no longer correctly reflects your Elo. Maybe you go from using a Firestarter to kick some butt then switch to an unmastered Raven 2X with a poor loadout. Your Elo will need time to adjust to the change in mechs.


Now, yes, it will eventually adjust for all the different mechs you use but now you have gone from it needing thousands to get properly adjusted to needing significantly more, maybe millions of matches.

Edited by Mercules, 26 January 2015 - 09:13 AM.

[Hotthedd]

Heffay, on 26 January 2015 - 09:04 AM, said:

What if that Direwolf was also the drop caller, picked an overwatch spot where he could defend 2 cap points, and therefore funneled the enemy into predetermined kill zones to be picked apart by his team?

There are FAR too many factors that go into winning a match other than herpaderpa damage. That's why you base it on whether you win or lose. The most valuable unit on the field isn't necessarily the one that has the highest score in anything on the scoreboard.

Winning isn't everything. It's the *only* thing.

So you are saying that the entire enemy team said to themselves: "Oh look! a lone Dire Wolf! Everybody run away!!!"
Meh, I'm not buying it.

That Dire Wolf got carried, and did not help his team win. He was just the last one standing when the fruits of his team's labors came in.

[Joseph Mallan]

Davers, on 26 January 2015 - 08:49 AM, said:

It's funny that you picked quarterback, as it is a very versatile role. He has to be proficient at both throwing and running plays himself.

But not every player is versatile like that.

[STEF_]

Joseph Mallan, on 26 January 2015 - 03:44 AM, said:

I've always questioned Using W/L as teh metric for a player's Elo in a team based game.

For instance I can have 4 kills and 6 assists, The Team loses. I get points taken from my Elo cause of teh team's failure to win.

I can have no kills 1 assist the team wins and I get point Added to my Elo ...for being carried.

Our Elo system does not judge MY performance properly this way.

That's right considering those 2 matches.
But with great numbers, if a player is good his team tend to win more, thanks for his contribution.
This for Puglandia.

With team is all other business, imo

[Bobzilla]

The biggest problem is that you can have less than a 1 win/loss and still gain elo. You can also have higher than a 1 win/loss and still loose elo. Because it's the difference of elo of the teams averages that effects the amount gained/lost.

So elo and win/loss are completly random and its basically pointless and has no effect on making close matches.

They've only said that their prediction of match outcomes works, not that the match was even at all (12 elo of 100 vs 12 elo of 2000, we predict the 2000's are going to win, they do, working as intended).

They gotta stop making one team then the other, they gotta do both teams at the same time and balance as it goes, other wise you have good+bad on one side with average on the other. Seeing as a few mech not contributing hurts a team more than a few mechs doing really well helps, the overall average team will do better.

[BourbonFaucet]

Instead of elo, why not use the current "Match Score" statistic instead? We know that people who are complete potatoes don't get very many match points, although it is fairly easy to get, say, 30 points. Have every player have an average of all their match scores (start new players at say, 60), and use those averages to pick teams from the MM instead.

[Joseph Mallan]

Stefka Kerensky, on 26 January 2015 - 09:18 AM, said:

That's right considering those 2 matches.
But with great numbers, if a player is good his team tend to win more, thanks for his contribution.
This for Puglandia.

With team is all other business, imo

I have hundreds of those 2 matchs!

[Koniks]

Joseph Mallan, on 26 January 2015 - 03:44 AM, said:

I've always questioned Using W/L as teh metric for a player's Elo in a team based game.

For instance I can have 4 kills and 6 assists, The Team loses. I get points taken from my Elo cause of teh team's failure to win.

I can have no kills 1 assist the team wins and I get point Added to my Elo ...for being carried.

Our Elo system does not judge MY performance properly this way.

Elo doesn't care how you win, just whether you win. And with a sufficient number of matches to work with, it's very good at figuring out which players are correlated with winning, even in team games. It actually works better with figuring out an individual player's Elo with random teams than it does if a player only dropped in a static group. That's because you're the only common variable in your matches.

ETA: the problem isn't Elo ratings it's how matchmaker puts together teams. It emphasizes time to drop over match quality. PGI repeatedly cites the difference of average Elo rating between teams but doesn't talk about the average Elo within a team or how that spread compares for both teams. The former lets you predict which team is most likely to win or lose. The latter should give a better idea of how close in ability all the players in the drop are.

It's worth pointing out that making good drops isn't an easy challenge for PGI. We have a fairly small game population split across 4 modes. That's either going to lead to some long wait times or some bad drops.

My preference would be for lobby systems with or without a ladder system. That's worked for everything from online card games to first person shooters for a lot longer than dev-matchmaking with hidden Elo. Though obviously they'd have to find a way to deal with preventing players from exploiting the economy.

Edited by Mizeur, 26 January 2015 - 10:03 AM.

[Tombstoner]

Mizeur, on 26 January 2015 - 09:44 AM, said:

Elo doesn't care how you win, just whether you win. And with a sufficient number of matches to work with, it's very good at figuring out which players are correlated with winning, even in team games. It actually works better with figuring out an individual player's Elo with random teams than it does if a player only dropped in a static group. That's because you're the only common variable in your matches.

Unfortunately things are confounded by the not so random selection of mech type.

[Joseph Mallan]

Mizeur, on 26 January 2015 - 09:44 AM, said:

Elo doesn't care how you win, just whether you win. And with a sufficient number of matches to work with, it's very good at figuring out which players are correlated with winning, even in team games. It actually works better with figuring out an individual player's Elo with random teams than it does if a player only dropped in a static group. That's because you're the only common variable in your matches.

Archived Totals
Kills / Death 2,732 / 2,226
C-Bills 1,886,322
Experience Points 2,494,658
Wins / Losses 2,577 / 1,806
Kill / Death Ratio 1.23
Accumulative C-Bills Per Match 120,511.81
Avg. XP Per Match 569.17

Current:

Kills / Death 286 / 533
C-Bills 1,886,322
Experience Points 372,854
Wins / Losses 271 / 370
Kill / Death Ratio 0.54
Accumulative C-Bills Per Match 84,063.19
Avg. XP Per Match 581.68

I apparently Used to be better than I am now, or I was carried much better by a dedicated team.

[Anton Shiningstar]

Archived Stats of my Solo PUG Anton Shiningstar
Kills / Death 353 / 305
C-Bills 2,040,913
Experience Points 359,210
Wins / Losses 255 / 205
Kill / Death Ratio 1.16
Accumulative C-Bills Per Match 145,284.32
Avg. XP Per Match 780.89

Now Current

Kills / Death 18 / 52
C-Bills 2,040,913
Experience Points 28,366
Wins / Losses 26 / 31
Kill / Death Ratio 0.35
Accumulative C-Bills Per Match 64,774.42
Avg. XP Per Match 497.65
Signed Joe Mallan

Edited by Anton Shiningstar, 26 January 2015 - 09:55 AM.

[Joseph Mallan]

Game got harder for me for some reason but I have not complained. At least I have proof I have not been been a hypocrite.



I just noticed... My C-Bills have not changed??? What the Heck?

Edited by Joseph Mallan, 26 January 2015 - 10:00 AM.

[Koniks]

Tombstoner, on 26 January 2015 - 09:47 AM, said:

Unfortunately things are confounded by the not so random selection of mech type.

Nope, that's all factored into the outcome. You are either able to choose mechs and play them in a way that is correlated with winning or you aren't.

Elo is just a complicated statistical algorithm for a simple concept. Players are either more correlated with winning than the average player, less correlated with winning than the average player, or equally correlated with winning as the average player.

The only things that throws off an individual player's Elo are sample size and things like whether they almost exclusively play with exactly the same teammates. In the latter case, the only difference in Elo between them and their teammates would be whether their teammates ever played with other players.

Joseph Mallan, on 26 January 2015 - 09:52 AM, said:

Archived Totals
Kills / Death 2,732 / 2,226
C-Bills 1,886,322
Experience Points 2,494,658
Wins / Losses 2,577 / 1,806
Kill / Death Ratio 1.23
Accumulative C-Bills Per Match 120,511.81
Avg. XP Per Match 569.17

Current:

Kills / Death 286 / 533
C-Bills 1,886,322
Experience Points 372,854
Wins / Losses 271 / 370
Kill / Death Ratio 0.54
Accumulative C-Bills Per Match 84,063.19
Avg. XP Per Match 581.68

I apparently Used to be better than I am now, or I was carried much better by a dedicated team.

You're obviously playing a lot less. So it's possible that you haven't played frequently enough to retain your edge.

And the game has changed a lot since the stat reset. AC nerfs. The SRM hit reg fix. Jump jet heat. Jump jet nerfs. Clan introduction, nerfs, and buffs. Pulse laser buffs. Quirks. 2nd quirks.

So perhaps you haven't fully adapted.

Very little of that would have to do with Elo. And since they haven't reset Elo, it could be a painful learning curve while you figure it out if you were at a point in the Elo distribution where you were grouped with other players who adapt faster. And with all the games you've racked up, it could be awhile before your rating will regress enough.