[MUD-Dev] D&D vs. MMORPG "complexity"

Daniel.Harman at barclayscapital.com Daniel.Harman at barclayscapital.com
Wed May 21 10:35:45 New Zealand Standard Time 2003

From: Dave Rickey [mailto:mahrinskel at brokentoys.org] 
> From: <Daniel.Harman at barclayscapital.com>
>> So your are saying that people who have high level chars are more
>> likely to renew their account? Surely you could draw the opposite
>> conclusion that people who renew their account are more likely to
>> have high level chars?

> How are those two statements different?

Well given that achievment is largely predicated on time invested,
it seems implicit that those who tend to renew more will have higher
level characters. If you are inclined to cancel, then its less
likely that you will have high level characters, there are a lot of
opportunities to quit before then!

i.e. it could mean either :

   A) They won't quit because they have high level chars.
   B) They don't have high level chars because they quit.

So the cause and effect are transposed depending.

If you interpret it one way, you might erroneously assume that since
those with high level chars are less likely to quit, then you will
have higher retention by handing out high level chars to people or
lowering the xp curve. That may be true, but you can't reliably read
that from the data.

You might be able to extract further meaning by looking at hours
played as well, from a players perspective it is probably more
telling. As a company it is easy to get fixated on billing periods,
but when I play games I don't worry about that I worry about if I'm
bored yet.

> What I found was that if I divided the accounts by the level of
> the highest character on that account, the month-to-month
> cancellation percentage rates were invariant by any other
> subdivision (within statistical variation). The age of the
> accounts was irrelevant, if that is what you were getting at, the
> cancellation rate of accounts 3 months old with 50th level
> characters was statistically identical to that of accounts that
> were over a year old, and for all other levels.

That definitely supports the idea that looking at hours played might
yield more info.

> Oddly enough, so was the cancellation rates of 5th level mains in
> the same time periods (in other words, there were an appreciable
> number of year-old accounts with their highest level character at
> 5th level, and they cancelled at exactly the same rate as accounts
> that were 3 months or less in age).

In cases like this, it would be interesting to see when they had
last logged in & how often they died each level. Were these
forgotten accounts, accounts regularly played by socialisers, people
who made lots of characters but never leveled that far, or
incompetent players who just couldn't advance etc.

> I had truly reliable data for only the first quarter of 2003 and
> June of 2002 when I ran my analysis, what was interesting was that
> the June data *did* show some discrepencies when analyzed by class
> and level, specifically a few classes with higher cancellation
> rates and a big bulge in the cancellations of 41st level mains
> (there was a "hell level" at 41 at that time).  After the changes
> made to address those initial findings (smoothing the levelling
> curve to eliminate the hell level and buffing up a few classes),
> all that was left was that invariant linear relationship.  In
> other words, it was exactly as if the linear relationship between
> cancellation and level was some kind of universal constant, that
> would pop out when the game was "close enough".

Your June conclusions sound fairly safe, and they no doubt supported
suspicians you held anyway?

Whilst I did jokingly suggest that you might hand out high level
characters, it is quite possible for DAoC that it might work. After
all there is a belief that the game is most fun when you get to the
high level RvR content which obviously requires a high level
char. Of course it would probably break other things :)

If you had wanted to know more, then an exit pole when people
cancelled might have helped too.

> It's only data from one game, and from a fairly short period. But
> it was so unexpected and so strong of a correlation, it has to
> mean *something*.

What that something is, however, is non trivial to work out!

MUD-Dev mailing list
MUD-Dev at kanga.nu

More information about the MUD-Dev mailing list