83 replies to this topic

[Errinovar]

Lets assume a solo drop player:

How high would the win loss ratio need to be over how many matches in a specific chassis to be considered significant in that players contribution to the overall performance in pug matches. Considering all the variables that come into play concerning match makeup which is pretty much impossible to account for (How many of your team mates were solo drops on how many matches, how did the other players perform in comparison on any given match,etc) at what point does a chassis win loss ratio become useful information?

I ask because I have run 790 matches (all solo drops) in my YLW with a win loss of 1.3 and I question whether that value has any meaning at all as to my performance with that mech in solo queues or is it still just a function of random luck on what kind of a team you draw.

[Vandul]

You didnt choose the PUG life...

W/L in solo drops is primarily based on the PUG lottery. Unless you were a very active leader in those 790 matches.

Group play, W/L has more significance, IMO.

[Joseph Mallan]

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.

[Lily from animove]

well depends, I think you need a few hundred matches to get known to the game and its mechanics, then you also needs few maybe 50-10 matches (if you already know how MWO works) to get known to your mech. e.g.: where can it fit through, from what position on what map you can you fire without hitting the ground, where cna it jump to and where not, etc. The issue with those stas are: we can not archive them actively and make a "cut" to see our current performance. you will always have the messed up stats of your "l2p" matches, or your testbuilds. and so you can not differ your true performance vs your.

further the stats are messed up because they put pug and group queue games together. One being good in the groupqueue may suck in the pug queue, because pugging works different.

And if you pug only your stats of W/L will be influenced by both, luck and skill. if you are a player that cna constantly make good matches and deliver good peformance, then the chance of your team to win raises definately. yet you can hardly decide theoutcome of a match alone. so a part is also luckbased. But I guess luck equals out, becuase your opponents are affected by the same randomness (in pugland).

Bets proof simply is my statistics withing clanmechs, clanmechs were implemented after my lp2 phase, and the w/l of tthe clanemchs I have good performance in is very much in relation to the w/l i achieve. simply because your performance in those mechs affects the outcome of the match. surely also infected on how muc easymode a specific mech is *caughTBRcaugh*

Edited by Lily from animove, 26 January 2015 - 03:50 AM.

[Errinovar]

Actually the PUG life is more a matter of living in Japan with a full time job and 2 small children that kind of make it difficult to make time commitments for playing games. So rather than annoying team members by constantly having to back out of engagements because of R/L, I have been left with the PUG life. ~3k drops and every last one of them solo.

[Joseph Mallan]

Errinovar, on 26 January 2015 - 03:46 AM, said:

Actually the PUG life is more a matter of living in Japan with a full time job and 2 small children that kind of make it difficult to make time commitments for playing games. So rather than annoying team members by constantly having to back out of engagements because of R/L, I have been left with the PUG life. ~3k drops and every last one of them solo.

If a team gets annoyed by R/L they obviously need to get one.

[Errinovar]

Actually with my YLW, it was the first 250 matches before I really figured out how to play fairly decently. At that point I had about a .8 w/l ratio and a .5 k/d ratio (I went through the rambo phase, the face tanking phase, the not so stealthy ambusher phase, the LRM bait phase, the circle duel phase, the lone push phase,etc..) and then one day I just kind of woke up and started dropping 400-700 damage games with multiple kills. The following 443 games managed to bring the numbers back up to 1.3 w/l and 1.58 k/d ratio. I have 97 drops since release and in the current stats I still have a 1.32 w/l and a 1.58 k/d with my YLW.

Lily from animove, on 26 January 2015 - 03:45 AM, said:

well depends, I think you need a few hundred matches to get known to the game and its mechanics, then you also needs few maybe 50-10 matches (if you already know how MWO works) to get known to your mech. e.g.: where can it fit through, from what position on what map you can you fire without hitting the ground, where cna it jump to and where not, etc. The issue with those stas are: we can not archive them actively and make a "cut" to see our current performance. you will always have the messed up stats of your "l2p" matches, or your testbuilds. and so you can not differ your true performance vs your.

further the stats are messed up because they put pug and group queue games together. One being good in the groupqueue may suck in the pug queue, because pugging works different.

And if you pug only your stats of W/L will be influenced by both, luck and skill. if you are a player that cna constantly make good matches and deliver good peformance, then the chance of your team to win raises definately. yet you can hardly decide theoutcome of a match alone. so a part is also luckbased. But I guess luck equals out, becuase your opponents are affected by the same randomness (in pugland).

Bets proof simply is my statistics withing clanmechs, clanmechs were implemented after my lp2 phase, and the w/l of tthe clanemchs I have good performance in is very much in relation to the w/l i achieve. simply because your performance in those mechs affects the outcome of the match. surely also infected on how muc easymode a specific mech is *caughTBRcaugh*

Edited by Errinovar, 26 January 2015 - 04:02 AM.

[LordKnightFandragon]

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.

Its why WL alone should not be the determining factor. It should be all the player's stats in everything combined and averaged.

It should be you
WL
KD
Avg XP
Avg dmg across all mechs
Avg capture
Avg defense
Avg Conquest points.....I suppose capture only plays a part if you play those game modes.
Accuracy
Shot efficiency

But it should be your overall in game performance rating. Not any 1 single stat, no 1 stat tells the whole story. An LRM Boat will likely have low accuracy, low damage and low kills, but hes actually really good at covering his team. And leads to their winning alot. Meanwhile, a guy who steals kills with 2 hitpoints will have alot of kills, but will prolly have low overall damage, a high accuracy rate but likely a bad WR.

[Mercules]

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.

Well.... supposedly even though you performed well your team may have been "intended" to lose. So the impact on your Elo was 0. In games where you did nothing but your teammates won the game you may have been slated to win anyway so there was hardly any impact on your Elo.

Now that in itself is frustrating but if Elo is working correctly then you are not losing rank because you have 11 derps and you, because Elo stopped and went, "Wait, he has 11 derps and so should lose to this team."

[RedDragon]

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.

Who except PGI hasn't?
Seriously, I still have no idea how they came up with this in the first place.

[627]

pug life...

Let's take your wang for example.


And with that the term and concept of carrying a match:

That YLW is a mech for skirmish and flanking, fast in, fast out. You can't "carry" in this mech. Carrying means you lead and dictate the match, you form and design it. Big words for actually gettin sh!t done.

If you try to lead in a wang, you get your big gun blown off and that's it. If you wait for your big guys to start fighting you can hop in and get some kills and heavy damage, but you need those assaults who are in rpg-speak your tank while you are the DD.

So in the end, you can't influence the outcome that much. The more weight you bring, the more you can do, just with your presence and actions. But the lighter you get the more you are forced into another role.

[Alistair Winter]

Good question. I think you're going to get a lot of replies from people who don't have a good grasp of statistical significance. Myself included.

[Impossible Wasabi]

RedDragon, on 26 January 2015 - 05:43 AM, said:

Who except PGI hasn't?
Seriously, I still have no idea how they came up with this in the first place.

Someone watched chess or more probably WoT too much and thought "This will work great in a team based game with a variety of different weapons and chassis!!!"

It's really why the Space Pope questions ELO, it works for one in Chess because it is a game with very little starting variation (in setup, players can of course decide upon a wide variety of moves once the game starts) and one where both players play the game with the exact same pieces.

WoT manages with ELO because the population in the game is large enough to avoid many of the problems of making ELO a centerpiece of a team based game. However, players still on occasion complain about how the matchmaker organizes teams.

In MWO, ELO works as a sadistic training tool, in that the better you do, the more likely you are to be tasked with carrying your team (which sort of works in contrast to Chess and WoT, where higher ELO means better opponents and teammates).

Edited by The True Space Pope, 26 January 2015 - 06:26 AM.

[generalazure]

The probability of winning 447 or more out of 790 for evenly matched games (50% chance to win) would be about 0.012%. Dunno if thats what you wanted to know.

[levitas]

Not sure if Elo prevents this from being a useful analysis, but assuming a known mean for wins of half the number of games played (12 people win, 12 lose in each game and ties are rare so this should be good) and a 50% win probability given the null hypothesis that you are not impacting win rate, you find a standard deviation of 14 games after 790 played.

With an estimated 447 wins vs an expected rate of 395, you are sitting on 3.7 standard deviations better than expected, with a p value of ~.0001

So, yes, you are making a statistically significant contribution to your teams and we can discard the null hypothesis that you are not.

Keep it up! For reference, I'm sitting at a 1.26 w/l after 1458 games, putting me at 4.4 standard deviations away from it being the mean coin flip result. You are definitely making a measurable and significant impact beyond doubt.

Edited by solar levitas, 26 January 2015 - 06:47 AM.

[Lily from animove]

RedDragon, on 26 January 2015 - 05:43 AM, said:

Who except PGI hasn't?
Seriously, I still have no idea how they came up with this in the first place.

becaue its easy to implement. making some more complicated algorithms makes stuff way harder, especially since this is not CSS where everye starts with the same equipment. we have mechs configured for supports, what does damage and kills mean on them? hardly anything.

Not even matchscore currently is a good value to judge.

[Heffay]

LordKnightFandragon, on 26 January 2015 - 05:10 AM, said:

Its why WL alone should not be the determining factor.

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.

[Lily from animove]

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.

but then we know its not about winning anymore or maybe it is?, becuse the direwolf making those 8 kills may have prevented 8 opponents from capping, or killing that single direwaolf camping the spot. So we can not entirely say how much everyone distributed to the kill within this entire scenario.

Actually a good distributed score to the matchscore depending on what one does would be the best way, but this now includes so many different variables, that it is very hard to judge how much truly distributed how much to the win scenario.

[Tombstoner]

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.

but one cant cap if the other doesn't carry..... This is clearly team work between a vet and there nube friend..."i'll kill them you go cap" i cant see how its cut and dry.

If your winning more often the loosing. then your basically above average relative to the others playing thus ELO should increase. Everyone gets pulled from the same que. equal chance to be on a exceptionally good team. meaning some pugs just gell as a team because some puggers know its utility and try to work as a team. thus W/L is a good measure to use.
the better players in a pug que will over time have a higher ELO due to spontaneous non verbal team work.......

[RedDragon]

Lily from animove, on 26 January 2015 - 06:54 AM, said:

Not even matchscore currently is a good value to judge.

Why? It doesn't give a perfect view on how you perform in matches, but it is way way way better than W/L. Because it at least measures your personal skill, either in capturing points or in dealing damage/helping your team. Basing it off W/L just doesn't make any sense (except maybe for teams that play a lot together in the same formation). You can be the best player on your team and it won't matter. Likewise you can be the worst player and it also won't matter.

You basically can't be a bad player and reach a high match score, and if you always have low match scores, you probably are not a good pilot. On the other side, it is perfectly possible that the best pilot here on the forums loses every single game because of bad team mates.

Edit: The trick would be not to base Elo on the absolute match score but on the relative one in regards to your team. If you are constantly doing better than your team mates, your Elo should rise and vice versa.

Edited by RedDragon, 26 January 2015 - 07:06 AM.