Let me just first say I did not write the paper, do not understand half the paper (but I understand the other half), and do not even know enough to properly defend the paper.
My personal view though is that team WL ratio for solo QP is not a good indication of individual performance. It might be the best in MWO's case though, but I make a distinction between what's good and what's available.
Zergling, on 25 March 2018 - 12:52 AM, said:
Then that research doesn't apply to games like MWO, because everyone in a battle has an influence, and bad players absolutely result in losses where there would otherwise be wins.
Unless you want to say that a player that runs into the enemy team and suicides at the start of every battle, or just sits AFK at spawn doesn't influence the odds of their team losing by creating a 11 vs 12 situation in every battle they are in?
Everyone should have an influence even an AFKer. Even an AFKer might have a positive influence on a team, might be surprising but it's possible, eg that AFKer is an intentional TKer or likes to yell in VOIP, or could have led the team down the wrong corridor, etc. It's not just that it's 12 vs 11 so automatically it's a negative influence. However the paper suggests that the strongest influencers are the top 2 players in the team. From my reading it's saying there's a stronger correlation of the top 2 players influencing a team's outcome (win/draw/loss), than the other players.
Put it into MWO terms. Have you had games where even with TWO afk team mates, your team still pulled a win? I know I have, and I've seen Baradul win despite being down 12v10. Sure maybe your team lost, but regardless of win or loss, we can agree that that AFK event influenced the team? Sure it did. But we cannot immediately conclude it's positive or negative. What was the biggest influence of the outcome, win or loss? Without examining every match in fine detail from every player's perspective, the paper is suggesting that the top 2 players in the team had the biggest influence, while everyone else did have some influence, with a mathematical correlation between the performance of these 2 top players and the actual final match outcome, win, draw or loss.
Zergling, on 25 March 2018 - 03:23 AM, said:
If bad players did not have a negative influence on their battles in the solo queue, then they would have W/L within a few standard deviations of 1.00 W/L; statistically normal results.
Because if they do not have any influence on their battles, then it is effectively a 50/50 coin toss to determine if they win or lose a battle.
But many players have low W/L with many thousands of battles; it is mathematically implausible, to the point of being impossible, for players to produce W/L so far outside standard deviation.
Go look at Jarls list and sort by Games Played; there's one player with 21,410 battles and a 0.79 W/L ratio in there (approximately 9,449 wins to 11,961 losses).
The odds of that occurring (if MWO battle outcome was a 50/50 coin toss) is a staggering 1 in 2,556,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000
For a less extreme example, there's a player further down with 13,683 battles and a 0.90 W/L ratio (approximately 6,481 wins to 7,201 losses).
The odds of that occuring with a 50/50 coin toss is still a ridiculous 1 in 24.94 billion.
Given the majority of those players are solo queue, it is a fact that those players are having a negative influence on the their team win chances.
My personal take is that if I follow with the paper's arguments, then I will stop coming back to WL to argue why the paper's conclusions (or my interpretation of its conclusions) are wrong. Otherwise it's just going in circles.
A: WL is not relevant.
B: WL must be relevant, this guy has so many games with a <1 WL ratio, he's a bad player.
A: WL is not revelant.
B: WL must be relevant, this other guy has >1 WL ratio, he's a good player.
The paper alleges that the biggest influencers are the top 2 players in a team. If you're not the top 2, don't look at WL as a determinant for individual skill. If that statement cannot be accepted cuz whatever eg you don't believe the maths, don't agree with the paper, don't understand the paper, don't believe me, want to believe something else, want to hold onto long-held understandings, etc, it's then literally pointless to even debate its merits, because the standing counter-argument is WL
is a determinant for individual skill.
Sjorpha, on 25 March 2018 - 03:12 AM, said:
I don't really think this paper supports the conclusion that only top 2 players matter, at the very least the writers doesn't seem to defend this specific claim/conclusion anywhere. It also doesn't seem to be the question they are trying to answer. I'd be very sceptical of far reaching conclusions going sideways from their chosen data set, I think you need to work specifically on that question itself if claims are to be made about it.
There is also a number of pretty significant differences between analyzing professional soccer and a game like MWO, the two most important being that 1; soccer tournaments play with the same teams throughout the tournament (while MWO SQ is random teams) and 2; the skill disparity in professional soccer is much smaller, there are no actual bad players in professional soccer for starters. They are all pretty damn good at playing soccer, they all train as a team and play as a team, this may create a stable background against which the best players performance can be the most trackable variable between matches, but we don't know what happens if you'd throw in some complete amateurs in some of the teams or play the tournament with random teams. Now that said I don't agree this article supports your general claim that well even in case of professional soccer.
This problem also: If your statement were true only the best players in the game, those who qualify for consistently being one of the top 2 players, should have a positive W/L while everyone else should trend towards a 1/1 W/L over time.
But that isn't the case at all! Instead there is a granular scale where all sorts of high and low W/L ratios are represented, and very bad players have very bad W/L just like good players have good W/L. Why do bad players predictably trend towards bad W/L if they don't affect match outcomes?
Using myself: I'm not a top tier player, I don't consistently fall in the top 2 player category but my above average skills still net me a fairly stable W/L record of around 4/3. How is that possible if your claim is true?
I think you misunderstood what it meant by "top 2 players". It's the top 2 players in the team, in a match. In MWO it's impossible to know, outside of established names in the community, who are presently the top 2 players in the team. Also, the paper is saying that the top 2 players have the biggest influence in a match. It is not saying they are the ONLY determinant. Every player has influence in the outcome of a match, positive or negative, but the ones with the biggest, and according to the paper by a wide margin, are the top 2, maybe 4 players if stretched as the paper caveats.
If considering match outcomes, these top players in the match are more statistically likely to affect a match's outcome, that means there's a stronger cause and effect between the match's outcome with the performance of the top 2-4 players. This means both ways, negative or positive. That means if that top player is on a bad day, his negative influence on the match has a stronger impact to the outcome of a match than say the negative influence of an average or genuinely bad player.
And finally, we're circle-jerking abit. The paper is talking about ignoring WL if you're not the top 2 players in the match. Yet the counter-arguments thus far are to still regard WL. My personal position, after reading the paper, is that if I'm not 1 of the top 2 players of the team in that match, granted I will have some influence to the match's outcome, I'm just not the biggest influence. I may have edged it abit, but I'm not the fulcrum I think I am. The problem is that we have no way to tell at any given match who's these 2 pilots short of celebrity pilots. Hence the extrapolation by the top 17% (2 divided by 12) of the MWO community. Since I'm not in the top 17% of the MWO community (hmm am I?), I would disregard my WL ratio as a determinant of my individual skill, and I will rather look for other indicators. Using WL ratio to decide yes or no to being in this top 17% will just be more circle-jerking.
For some of you who are indeed in that top 17%, by all means do use your WL ratio to examine your own individual skill, as that has meaning, more meaning than those outside that 17%.
The missing gap is how to tell if one is a top player or not. The paper explains in detail how they measured player performance in team dynamics to score a player, eg # of accurate passes, etc so I can only assume their assessments of who are the top 2 players are justified. There is existing data on MWO pilot activities eg kills, score, assists, etc etc, but whether they are good assessments or not to finally arrive at a correctly adjusted leaderboard, is well beyond me.
Edited by arcana75, 26 March 2018 - 01:32 AM.