Inflation Information

[sully800]

As you may know, I've been recording data from the scoreboard to show how the the scores rise as time progresses (and more players join the site). I kept track of about 3 days a week starting in October, then I missed several weeks and in the last 2 weeks I've been able to get data every day. I will continue to record info (not necessarily daily, but probably whenever I get on CC) and I'll post updates each month.

I'm writing down the scores of 1-10 and taking the average because the score of an individual at the very top is highly variable compared to further down the scoreboard. I've also been recording the scores of #100, #250, #500 and #2500. All data sets have shown a marked increase in points since I started logging them.

[sully800]

There is a decent amount of scatter in this data set, but the upward trend is clear. The R^2 value in the upper left is a measure of how well the data fits the straight line plotted on the graph. An R^2 value of 1.00 would be a perfect line, and an R^2 value of 0.00 would show zero correlation.

Also in the upper left is the equation of the best fit straight line. The slope shows how much the points have increased each day (x).

For the top 10 average, their scores have increased 2.447 points per day since I began recording.

[sully800]

The score of 100th place shows better linear correlation (.7545) but a much lower score increase: .72 points per day.

[Blitzaholic]

very nice, well done sully

[lord voldemort]

have you thought maybe the growth is expenitial....
you can still work out an r^2 value for that...
my graphics calculator used to find me best fit i.e highest correlation
would be intresting to see if it was an expenital growth

[sully800]

From the score of #250 and beyond the linear correlation is excellent (all values are well above 0.90) which means the trend should be expected to continue steadily into the future.

Additionally, the slope of each graph decreases slightly which is expected. That means that the players lower on the scoreboard increase by less points per day (on average) than the players at the top of the scoreboard. Note: I'm referring to increases due to score inflation, not an individuals increase in score from winning a game

[sully800]

Same trends described before apply to the score of #500.

[InkL0sed]

Interesting, but at least in my opinion, not very clear. You keep saying the top 10. What do you mean? The top ten scores? As in the people with the ten highest scores? Also, what are the blue dots?

Maybe I didn't read carefully enough, but I do think you do need to clarify. I do appreciate this though, good work.

EDIT: OK, upon reading again, I got it. I'm just slow.

[sully800]

And finally the score of #2500 shows the best linear correlation of all so far. A very steady and notable increase in scores over the last 2 months.

[Twill]

It fits perfectly well with a general trend of inflation in ranked online gaming.

Raph Koster says it better than I ever could (I think I've posted this before)

Skill is something that is a rare commodity, and it takes time and patience and a willingness to fail to develop it. A lot of people simply do not have the time and patience. Right off the bat, for any given type of game, there’s going to be a lot of folks who simply will not get in the door because of the skill threshold demanded.

This gets worse when you’re dealing with a multi-player scenario. Picture a group of six people in a multiplayer game. One of them is 10% better than the others. He therefore wins. His win record is now 1-0, and everyone else’s is 0-1. He’ll continue to win most of the time — though not all — and his win-loss record will be tilted towards the wins side — say, 8-2. But most people in the group will have 0-10 records, and a couple might have 1-9. A small margin of skill is enough to make a cumulative record look devastating. In competitive arenas like this, most people lose most of the time.

I’ve described this before as “the average user is below average” — meaning, the median user lies below the mean on the win-loss curve, because the win-loss curve turns out to be a power-law distribution. And what happens to people whose average experience is humiliation, frustration, and defeat? Well, they quit.

This would explain why inflation is faster at the top than in the middle, I'd imagine you've noticed a drop in the low end (even slower than the mid range inflation) as they donate to the top players.

(http://www.raphkoster.com/2007/04/23/th ... readmills/ if you want to read the original article or ever want to design an MMO-game)

Great stats Sully!
Twill

[sully800]

have you thought maybe the growth is expenitial....
you can still work out an r^2 value for that...
my graphics calculator used to find me best fit i.e highest correlation
would be intresting to see if it was an expenital growth

Yes, I still think that the point growth may be exponential since its likely that the growth of members to the site is exponential. It might make sense for both trends to follow each other but I don't think theres enough data yet to support that conclusion (especially since the data for #250, #500 and #2500 appears so linear). When the data set is larger we should be able to tell if the line stays straight or not.

Interesting, but at least in my opinion, not very clear. You keep saying the top 10. What do you mean? The top ten scores? As in the people with the ten highest scores? Also, what are the blue dots?

Maybe I didn't read carefully enough, but I do think you do need to clarify. I do appreciate this though, good work.

Sorry, let me try to clarify. Each day I look at the scoreboard and write down the scores of the top 10 members (its not necessarily the same people each day) and then take the average of their scores. The first point on the first graph is the average score of the top 10 players on October 8th. The second point on the first graph is the average score of the top 10 players on October 16th. By repeating that process over almost 3 months you can see that the scores have increased over time.

It's important to realize I'm not tracking the score of an individual player. That's probably easier to understand if you think of the score of #100. If the person currently sitting in 100th place wins a game, they move up, and the person who used to be in 99th place is now #100 (often the score will be a point or two higher). As new members join the site, more points are added into the community and everyone at the top eventually has a small increase in score. Or at least, that is the inflation which I'm trying to show here

[lord voldemort]

have you thought maybe the growth is expenitial....
you can still work out an r^2 value for that...
my graphics calculator used to find me best fit i.e highest correlation
would be intresting to see if it was an expenital growth

Yes, I still think that the point growth may be exponential since its likely that the growth of members to the site is exponential. It might make sense for both trends to follow each other but I don't think theres enough data yet to support that conclusion (especially since the data for #250, #500 and #2500 appears so linear). When the data set is larger we should be able to tell if the line stays straight or not.

yer that would make sense. nice stuff there mate

[friendjonny]

its nice to see a graph like that showing about how fast the scores are really climbing upwards

[Heimdall]

Time to increase the lending rate to curb inflation

[hecter]

Hey sully, with the current information that you have, do you think you could show us a graph of what future points will be like in say. Lets say, a month or even a year from now.

[MeDeFe]

Hey sully, with the current information that you have, do you think you could show us a graph of what future points will be like in say. Lets say, a month or even a year from now.

You've got the average point increase per day, just do the math.

[Itrade]

MeDeFe, that would work, except the point growth is exponential.

Anyway, Sully, are you implying that my score has been cheapened?

Also, is there a way to know what the average score for players with above 1000 points is?

[DiM]

Anyway, Scully, are you implying that my score has been cheapened?

yes that's exactly what he's implying.

let's say i have a skill rating of 8/10 and at today my score is 2600.
assuming i will not improve my score should be roughly the same if we have no inflation but since we have inflation i'll probably reach 3000 points next year even if i don't improve my skills.

generally speaking the top scores will improve faster not only because of inflation but also because of skill. as the quote twill provided says a marginal skill advantage can turn into a huge point difference. also the top players are always improving their skills and thus keeping that advantage over low rankers.

so it's safe to say that 5000 10000 or why not even scores of 100000 are reachable given enough time and enough new players.

that's why a solution for the inflation must be found as soon as possible. i'm very confident that this year we'll reach 5000. i'm willing to bet on it.

the best way to look at inflation is to check the captains.
4-5 months ago when i reached captain i was ranked 200. now the last captain is ranked 400+ on the scoreboard. so in just 4-5 months we've seen the number of captains double.

also look at older people that last year were in top 20 with 2200 points. now they wouldn't even get in top 200.

in another 4-5 months major will be what captain was half a year ago and we'll have 50-60 colonels.

[MeDeFe]

MeDeFe, that would work, except the point growth is exponential.

"Is exponential"? So far we're no further than "might be", frankly, I'm rather sceptical about it, especially since no numbers on membership increase have been presented yet.
And to calculate an accurate representation several figures that might not be readily available are necessary, total number of players that joined during whatever interval is set, the number of drop-outs during this interval, the number of active players at the beginning of any given interval, total number of players at the beginning of any interval in case the drop-outs are still telling people that there's a cool site but they don't have the time for it anymore.
Word of mouth propaganda isn't something straightforward, and just looking at the graph describing how many people are on the scoreboard at a given day will not even get you halfway there.

[Knight2254]

It fits perfectly well with a general trend of inflation in ranked online gaming.

Raph Koster says it better than I ever could (I think I've posted this before)

Skill is something that is a rare commodity, and it takes time and patience and a willingness to fail to develop it. A lot of people simply do not have the time and patience. Right off the bat, for any given type of game, there’s going to be a lot of folks who simply will not get in the door because of the skill threshold demanded.

This gets worse when you’re dealing with a multi-player scenario. Picture a group of six people in a multiplayer game. One of them is 10% better than the others. He therefore wins. His win record is now 1-0, and everyone else’s is 0-1. He’ll continue to win most of the time — though not all — and his win-loss record will be tilted towards the wins side — say, 8-2. But most people in the group will have 0-10 records, and a couple might have 1-9. A small margin of skill is enough to make a cumulative record look devastating. In competitive arenas like this, most people lose most of the time.

I’ve described this before as “the average user is below average” — meaning, the median user lies below the mean on the win-loss curve, because the win-loss curve turns out to be a power-law distribution. And what happens to people whose average experience is humiliation, frustration, and defeat? Well, they quit.

This would explain why inflation is faster at the top than in the middle, I'd imagine you've noticed a drop in the low end (even slower than the mid range inflation) as they donate to the top players.

(http://www.raphkoster.com/2007/04/23/th ... readmills/ if you want to read the original article or ever want to design an MMO-game)

Great stats Sully!
Twill

Perhaps it has something to do with all of the multi player games where all of the Colonel/Majors/Captains all band together to play a mixed team of cooks and Cadets?

I have serious difficulty even finding a game to play where the odds going in are not already heavily stacked against me.

[hecter]

Hey sully, with the current information that you have, do you think you could show us a graph of what future points will be like in say. Lets say, a month or even a year from now.

You've got the average point increase per day, just do the math.

But that's not in a handy little graph now, is it?

[jimboston]

Could we curb inflation by deleting un-used accounts... and also then cutting a corresponding number of points from all players?

i.e. If CC cut out say people who didn't play for 6 months... and their
total points was say 1% of total CC points... then you could drop all players totals by 1% at the time you make that cut. Thereby reducing the inflation.

If this was done every six months or so... you'd slow down the rate of inflation... or at least the part contributed by people who only play a few times then split.
===

I also understand the idea that 'avg.' player doesn't win 50% of the time... because the multi-player factor of the game.

That's why points are helpful... because you don't have to be 50% or better to increase your point count.

Assuming you play all players of comparable skill or point range... you can easily get points by winning only 26% or 27% of the time...

If you play all 4-player and win more than 25% of the time... your points will increase. If you play 6 player all the time and win more than 17% of the time... your points will increase.

(Again... assuming players of same skill range... and you don't lose to cooks....)

[Blitzaholic]

sully------------------------------------------------the graph master

[Itrade]

MeDeFe, that would work, except the point growth is exponential.

"Is exponential"? So far we're no further than "might be", frankly, I'm rather sceptical about it, especially since no numbers on membership increase have been presented yet.
And to calculate an accurate representation several figures that might not be readily available are necessary, total number of players that joined during whatever interval is set, the number of drop-outs during this interval, the number of active players at the beginning of any given interval, total number of players at the beginning of any interval in case the drop-outs are still telling people that there's a cool site but they don't have the time for it anymore.
Word of mouth propaganda isn't something straightforward, and just looking at the graph describing how many people are on the scoreboard at a given day will not even get you halfway there.

Erm, I don't actually know exactly what exponential means, so, erm, I just used it to make what I said look smart. I'd be totally interested in figuring out what the inflation rates will be in the future.

Also, jimboston's suggestion, sadly, is absolutely useless. The accounts are un-used, and thus they aren't adding to the inflation from the points up top. The inflation, I think, comes from new members who join a few games, lose or deadbeat, and never come back.

Also, if there's an inflation, my score should be increasing. However, I went down from being a captain with more than 2000 points to a measely Corpral 1st Class (I've got exactly 1200 points at the time of posting). I think this is unrefutable proof that we are not experiencing inflation at all! Like DiM said, the points should be increasing if my skill level, uh, remains high. Oh.

[DiM]

Also, if there's an inflation, my score should be increasing. However, I went down from being a captain with more than 2000 points to a measely Corpral 1st Class (I've got exactly 1200 points at the time of posting). I think this is unrefutable proof that we are not experiencing inflation at all! Like DiM said, the points should be increasing if my skill level, uh, remains high. Oh.

there are several explanations for your point loss:

1. bad streak of luck/inspiration. you'll soon get back
2. the average skill just increased and you went bellow that line even if it's just by a bit. as twill pointed a person with 10% more skill can get 100% more wins than the other guy. a small skill difference means a large point difference.