what's the average rating

[elxir]

Ok so like, once we let all the flying corpses lose their 1000 points to float down to the bottom of the cess pool, what constitutes the "average" players rating?

My theory is that like, the average mediocre (tautology sup!) player is rated like, 2000-2500. I don't count anyone who can't even get to 2000 as a real player because like, let's be serious...I can stay above 2000 playing rev ta...

So, math or otherwise, what is the "average" ladder player rated at? At what line do you normally think, yes, this player is competent?

Or are you like me, who thinks the rating system is not very indicative of absolute skill(since I have eyes and can see who is good and who is bad), but rather only a measure of relative skill (congrats, you're better than these players, but you're still pretty bad)?

[leggomyeggo]

Originally Posted by elxir

Or are you like me, who thinks the rating system is not very indicative of absolute skill(since I have eyes and can see who is good and who is bad), but rather only a measure of relative skill (congrats, you're better than these players, but you're still pretty bad)?

i think like this. everyone is bad until proven otherwise.

[wolf'j'max]

let's take someone as an example. Not entirely random because i was looking for a certain record: Tekn0 we all know he turned into a pretty good ball player, he doesn't have the biggest game sense around but is a decent/good player. Yet he has 8 more losses than wins. He still manages to get to a rating of almost 3k. Is that really the way ladder should be? Play a lot and you gain points? i myself don't play a whole lot and have 1/4 of tekn0's ball games played as a total. And i can't get higher then 2k at this point. Maybe if i'd play more, but i don't think that should be the way the rating works.

tl;dr the rating system seems like it's based on whoever plays the most.

edit: of course there's a lot of players who DO win more than average, but this is just an example of the possibilities of playing A LOT.

[elxir]

I'm not here to insult nobo's math skills, I'm here to ask what you think the "average player" is rated at.

[wolf'j'max]

well let's keep it at 2k-2.5k as you said. But that's just because the 'average ladder player' at this point has over 200-300 (in both modes) games played a season.

[soccernamlak]

To note, I am only looking at Ladder: Ball stats here.

Season 4 of Ladder started December 2, 2011.

Since then, 124 days has elapsed. I used 200 games as the cut-off for "normal" ladder player, or about 1.61 games/day.

I took the ratings of all those who have played at least 200 games as of 1630 EST today.

The data set included 131 individuals out of the 556 recorded in the ladder database, or 23.56% of the player set.

Results

The average number of matches played is 414.687 +/- 216 games per player

The average ranking of these players is 2153.38 +/- 578.8.

I plotted the rankings against number of games played.

As you can see from the graph, there is a very weak correlation here that an increase in the number of games results in an increase in score. Of course, plotting all 556 users in this graph might produce more support for this hypothesis that number of games played is directly correlated with ranking.

[Clapon]

ive never broken 2000, what mean?

[Slender]

ive never been above 1500, what mean?

[soccernamlak]

Originally Posted by Clapon

ive never broken 2000, what mean?

I'm assuming you go by [FRO]Clapattack?

If so, this means that your current score places you in the bottom 25% of active (>200 games played) Ladder players, but at the top end of that 25%. Essentially, 75% of ladder players have higher rankings than you.

If we base it on your highest score (still sub-2000), it means you are in roughly the bottom 40%, or about 60% of the active ladder players have higher scores than you.

That's just based on statistics alone based on today only. This is only a snapshot, so this can't compare you to the average ladder score over time. As such, I can't tell you if you've improved, remained the same, or dropped compared to the average of active players over the course of the season (which off topic would be a brilliant piece to add to the website: average rating of players over X games which increases as ladder goes on...shows you how you compare to your average peer).

Ultimately know that this is only statistics speaking here. You could take lix's view of "the rating system is not very indicative of absolute skill but rather only a measure of relative skill" as your increase/decrease of rating is only dependent of those who you play, not the rest of the ladder server.

[soccernamlak]

Originally Posted by SlenderMan

ive never been above 1500, what mean?

See post above for Clapon for meaning, but

Your current score puts you in bottom 4% of active ladder players; 96% have higher rankings than you.

Your highest ever score puts you in bottom 10% of active ladder players.


Note here that only 131 data points were used in this analysis, as again, this was the number of players (out of 556) who have played greater than 200 games. So these bottom/better than percents are only compared to people who have played more than 200 games as of today. As such you could be higher up in comparison if I threw in >100 games or >50 games even as requirements for "active players."

[nobodyhome]

Originally Posted by wolf'j'max

Yet he has 8 more losses than wins.

I've explained multiple times how win/loss records or ratios mean absolutely nothing in terms of determining your skill. Because of the balancer, ladder actively works to make your games harder and harder for you to win the higher you're ranked (by giving you worse teammates and harder opponents). Thus tekno is 1000 points higher than you because he still maintains a 50% win rate despite his games being much harder to win than yours.

[leggomyeggo]

what the math

[Solqom]

Originally Posted by soccernamlak

Did anyone else take a few minutes to work out what dot they were?

[soccernamlak]

As a further note to wolf'j'max's hypothesis and nobo's rebuttal that score is directly correlated to number of games played, when ladder season ends, I'll see if I can produce some more fun charts and statistical findings for all 556+ players to see what kind of correlation is there. As I stated earlier, the above charts are only regarding 200+ game players, which means we miss out on the newer (and probably lower-ranking) players, meaning the correlation was going to be weak at best (in which case it was, with such a low r-value)

[some poor soul]

it is my impression, based on your scatterplot, that about eight or ten players who have played way more games than other people are entirely driving the existence of any positive correlation between games and score. so even if causation runs from the number of games played to a vaguely higher score, that's only true for about ten players, and is not a general pattern. (also, that very, very low r2 suggests that there is not much of a correlation.)

and it's worth noting that causation may run the other way: players who are good may play more.

[soccernamlak]

Originally Posted by some poor soul

it is my impression, based on your scatterplot, that about eight or ten players who have played way more games than other people are entirely driving the existence of any positive correlation between games and score. so even if causation runs from the number of games played to a vaguely higher score, that's only true for about ten players, and is not a general pattern. (also, that very, very low r2 suggests that there is not much of a correlation.)

and it's worth noting that causation may run the other way: players who are good may play more.

Oh absolutely. In no way am I saying there is definitely a correlation here; only a very weak one at best. But because I've already limited my data to those playing >200 games, I'm already excluding a good portion of the population (76%) that would either help or hurt any sort of correlation here. <200 games are the "new(er)" people who, if wolf's hypothesis holds true, would have overall lower rankings as they haven't played enough games yet. That in turn would lead to a either stronger or more positively correlated data set. Or, of course, it could lead to data that doesn't hold any sort of meaning; resulting in showing support for the null hypothesis: that number of games played is not directly correlated to ranking.

I might also do a data set with >10 games compared to >0 games as to exclude anomalies or "luck" in artificially high or low scores to start, as the goal of the ladder system is for you to hover around your true ranking as quickly as possible.

EDIT: When ladder season ends, I'll also take a look at excluding outliers on the right side (those who have played way too much) to see if that provides any difference in results.

[[hlx] FatTristan]

Originally Posted by wolf'j'max

let's take someone as an example. Not entirely random because i was looking for a certain record: Tekn0 we all know he turned into a pretty good ball player, he doesn't have the biggest game sense around but is a decent/good player. Yet he has 8 more losses than wins. He still manages to get to a rating of almost 3k. Is that really the way ladder should be? Play a lot and you gain points?

This kind of thinking is how I thought for the longest time, but is actually falacious.

I'll first respond to the idea that win percentage determines how good a player is. The ranking system tries to put you at a place where you win 50% of the time. If you are approacing a 50% win ratio from above 50%, you are overrated. If you are approacing a 50% win ratio from under 50%, you are underrated. Someone with a 50% win ratio at 2000 is not a good as someone with a 50% win ratio at 3000. It took me a long time to realize win percentage matters a lot less than rating and only begins to matter when you examine people with similar number of games played.

I'll address the idea that you gain more points just by playing more games so people who are less active are just screwed. Ladder gives you more points for your wins the longer you've been away. Nobo has very smartly designed a ladder that gets you ranked correctly as fast as possible. He has even put in factors like gaining more points the more you win in a row and vice versa. If shmo played 10 games, he'd be higher than a lot of people who've played a hundred games.

I think people who play a lot simply get better and that explains why they get higher with such a low win percentage. Dozer from a couple seasons ago is a good example. He played a ****load of games and got to the top 25. I very much agree this is because he played a lot of games. His plethora of games played led to his success just because he got a lot of practice. You'd used to groan back then that he was on your team, but now you think "yay, Moser's on my side!" I think Tek is a similar example even though the 11 win streak means he's probably on a bit of a hot streak at the moment. He has simply gotten better and his ladder rating has started to reflect that. But if the season wasn't ending, I'd bet he'd probably have gone a lot down though.

[zz-]

Lol i just cringe and sigh at people's failure to understand much about ladder it's really so simple.


your # of games played has absolutely no inflationary effect on your ranking -- yes, ladder inflates, but when your rating goes up, your teams get more difficult, and you must play at that rating to maintain it. no stat has any indication about skill whatsoever except for player rating, which is like 99.99% to be accurate within +/- a few hundred. All of this assuming you have played a few dozen games and are playing to win and blah blah.

edit: i didn't see tristans post, i wrote mine because I read wolf'j'max post and have heard many people express the same mistaken line of thinking. Also, that graph is just plain silly because it neglects hundreds of variables, the most obvious of which is the more you play, the more practice you get. There's about 7 people that are "outliers" in # of games played, and of course theyre average or above average. If you eliminated these pieces of data you have a completely random scattering of points with no correlation.

In tekno's case, he has more losses than wins and a high rating because his wins have been worth more than his losses. That's caused by either constantly playing as an underdog and winning large amounts while losing small amounts, or by streaking hardcore and reaching that rating quickly. Note that in both of these cases, the rating is still accurate, and if you were to look at tek's win% (low) or discount tekno for having a high # of games played to determine how well tekno has played, you would be misled.

[zz-]

Re drunk-edited the above post and to answer elixir's original question : it changes as the season goes on and all ratings climb relatively. But I think if ladder were still zero-sum (no inflation) it would be 1500 as every action that contributes to your team winning takes points away from the other players, and vice versa, so a player that completely and totally played perfectly average would have the mathematical expectatiion of permanently remaining 1500. So, my reasonably informed guess is that its the sum of all players' ratings divided the number of players.

Also, there is kinda no such thing as an average tier player, as altitude is a game where there are countless opportunities for contributing for your team, and each keystroke that helps your team at ALL usually ends up contributing to your overall likelihood of winning. Dodge one loopy missile, and you slow the loopy down in his energy usage and position and you've made it that much more likely for your team to win. So i think just about every individual player represents a unique overall average contribution. Of course there are "echelons" and players can be somewhat generally grouped together, but given the nature of the game, its a pretty evenly spread continuum IMO. Especially as the ladder season goes on - the higher the spread of ratings, the more accurately players are can be divided into true echelons, like the top 4 being several hundred points higher than the rest of the top 10, which is accurate IMO. For me it's much more interesting when there are like 20 or 30 tiers of players, instead of about 10 like there are for the first month or two of a new season. Another reason (besides months and months of fun balanced games) that we should have an extremely long ladder season with inflation in place, and a higher starting rating and larger base change for a win/loss (i assume its still 25).

Apologies if this post sucks (it doesnt) but i just got done masting to a ton of sabermetrics and im in a mathy mood

[zz-]

LOL triple post but there's one more thing to add: what an "average" player is, and what a player's rating is, depends completely on who is in the games they play. Imagine a hypothetical player that got on a ladder server at 4 AM with 12 accounts, and started games, and only played as one player, with the other 11 not even spawning, winning each game 6-0. His rating would go up forever, but would have absolutely nothing to do with the other 600 players in ladder, but the 11 with which he'd played. i think shmo's tendency to be ultra-selective about which players he played with/against made this element of ladder very apparent. For a player's rating to be truly indicative of their value relative to the entire pool of ladder players, you have to not only be completely indiscriminant about when you play, but you have to have basically no tendencies at all and play a lot of games against all players. For the vast majority of players, the effect of this phenomenon is very very minor, as most people play when they feel like playing, and the quality of games is more or less fairly consistent, but it's just another one of the tons of "contexts" that must be kept in mind when evaluating the information the ladder standings give us.

[elxir]

i'm completely indiscriminate about what i play does that count

[soccernamlak]

Originally Posted by elxir

i'm completely indiscriminate about what i play does that count

I would think that depends. Are you playing a plane/setup that is not your best setup? There are, of course, plenty of reasons to do this: for example, if a team needs a whale, you might be better at loopy, but will whale to help your team. Regardless, it would depend on whether your setup is costing or winning you more games compared to your predicted outcome.

For instance, you might be on a team that has a higher chance of winning the match (resulting in a lower increase in ranking if you should win), but if you play a setup that isn't as good as another, are you winning these games more often than not or no? Because if you're losing these games, your ranking drops a bit faster to try and balance you out at a lower, average ranking.


What it boils down to is this: playing a wide variety of planes because you can / because it's helping balance planes on your side / because you are actually helping your teammates will probably result in a lower average score than playing your best plane/perk setup. This assumes that you could be good at all planes, but still have one or two setups in which you dominate. This is where there is a fundamental difference between one idea of the spirit of ladder (doing what needs to be done to win, even if it means whaling) versus how ladder organizes teams (assuming that higher rated players can be paired with worse teammates against tougher opponents as they increase in rank, meaning you are effectively required to carry/play optimal plane setup to continue winning and increasing rank).

[Ingbo]

how does ladder inflate? isnt there an equal point lost for every point gained?
sry ive probably missed a lot but there are a bit of tl:drs here so im kinda just asking

[Ribilla]

Originally Posted by Ingbo

how does ladder inflate? isnt there an equal point lost for every point gained?
sry ive probably missed a lot but there are a bit of tl:drs here so im kinda just asking

It used to be zero sum, but not anymore, so that new players who appear late in the season start below average. Basically extra points are injected in form of bonus points for everyone (say +2) each game. So if everyone in ladder had played an average of 25 games, the average score would be 1550 (1500 + 2*25). I think it's a little more complex than this in reality, but that's the gist.

[Ingbo]

ah ok ty <3

[Urpee]

You would hope to see some correlation between games played and average score, simply because those that play more have more practice and more opportunity to learn and improve.

Of course if ladder was full of people who have no ability to learn anything then yeah we'd see nothing.

Incidentally data from repetition learning shows that people do improve over their life-cycle even if they are already well versed.

[elixirwithani]

Originally Posted by Urpee

You would hope to see some correlation between games played and average score, simply because those that play more have more practice and more opportunity to learn and improve.

Of course if ladder was full of people who have no ability to learn anything then yeah we'd see nothing.

Incidentally data from repetition learning shows that people do improve over their life-cycle even if they are already well versed.

Yea, but this game makes no sense and there are those players who have been playing ladder for over a year who are still terrible.

[leggomyeggo]

Originally Posted by elixirwithani

Yea, but this game makes no sense and there are those players who have been playing ladder for over a year who are still terrible.

shout out to panic the pig.