So how is CoH REALLY doing? (An answer)

Tony, thanks for doing the normalization to US dollars.

What I find fascinating is that if you divide number of subscribers by dollar revenue (for the period they give subscriber numbers) it shows that the average subscriber spends $45 per month on the game. How is that possible? And yet, the numbers are consistent at that level.

Whether something is off or not, the average ratio of revenue to users can be used to calculate an estimate of subscribers for the last few quarters where date was not provided. It works out to:

2009 Q1 - 107,311
2009 Q2 - 115,534
2009 Q3 - 98,142

Eyeballing your graph, it is apparent that the biggest post-CoV gains were from Issues 9 (Inventions) and 12 (Cimmeroa/Midnighters/proliferation). The biggest losses coincide with PvP and AE rebalancing, respectively.

Interesting stuff.

think he means subscriber numbers.

Take the Magic: The Gathering 'What Color Are You?' Quiz.

well af, if gr were only what we saw at herocon, i'd be inclined to agree, 2 powersets, a side switching feature, a graphics feature that may not affect a number of us(though leaving it optional was a good thing to do, kicking the low end systems off forceably would be nasty) feature and a zone that s i seemingly a mix of 1984 and minority report only goes so far. but lets wait till we know more than the first murmurings of content. I would be surprised if the large staffing up and relative lack of storyline content or new art for over a year only lead up to that. that would actually get me dooming if that was all. I suspect a lot of older players have burned out a bit on the current stuff and are waiting in the wings for gr, and looking at the spike villains gave coh, even with a poorly received feature like ED which hopefully GR will lack, lets say im reasonably optimistic that it will draw back some people for at least the short term, from then on the ball is in the devs court to keep them.

Take the Magic: The Gathering 'What Color Are You?' Quiz.

OK, I took Johnny's concept of averaging the $/access to extrapolate the Access and concurrents for the missing values going forward.

The yellow and pink lines are the extrapolated numbers at the end of 08 going forward. The scale for those two are on the left side vertical axis.

The blue line are published numbers and the scale is on the right.

Obviously this is rough since the $/access did vary a bit. The average value was taken from the published access figures after the CoV launch ($42.03). Concurrents were calculated from a similar timeframe and worked out to 12.5%.



I also entered in the various events for boxed sets and booster packs etc. and most of them did not make an obvious difference to the income stream (not shown in the graphic). However, 06 Q3 still looks funny compared to the rest of the income chart. The only thing I can see is that the income could have come from preorders for the boxed edition of GvE exclusive to Walmart, which came out the following quarter.

Also according to MMOChart.com ((link)):

By the way, a commentary on these numbers. Both the access numbers and the revenue numbers are good for long-term trending to some degree, but they have issues when you attempt to extrapolate short-term changes in them.

The access numbers are not subscription numbers. Lets get that out of the way now. Everyone who says that NCSoft "used to provide" the subscription numbers and "now they hide them" is simply wrong. Its extremely unlikely that is true, for two reasons. One: the devs have stated in the past that they are capable of, and generally track, access. So the notion that they can't do that and are substituting the subscription numbers as a best guess is false. Two: the numbers oscillate too wildly relative to prior assertions that the subscription numbers themselves do not usually change dramatically from one month to the next.

The access numbers are a reasonable proxy for subscription numbers most of the time, at least for relative comparison, provided that during the comparison period the average player logged in about as often, at least to within one month intervals. However, the one time that assertion cannot be assumed is the case where someone attempts to assert "flagging interest." If someone claims that people are getting bored with the game and quitting out of boredom, they *cannot* assert equal login frequency, and the access numbers become worthless to that assertion in the short term. Access numbers *amplify* interest and disinterest in the game within a certain set of margins. In other words, if access numbers drop significantly in a single short term period, the one thing I'm basically absolutely certain of is that the subscription level dropped significantly less than the access numbers in that short term period.

Conversely, the revenue numbers have their own problem. They almost certainly reflect the subscription revenue as its collected, even though subscription revenue is not necessarily for a single quarter. In other words, the long term subscribers are probably adding 6-mo and 12-mo period fluctuations in the data which are very difficult to factor out: I don't think there are enough data points to do a very good time-domain analysis of subscription numbers, especially with all the other factors superimposed. Its clearly apparent, though, that for some reason or another there are definite oscillations in the data. There are several dips in the revenue numbers for which the revenue bounces back by most if not all of the dip. While the trend is generally lower, the trend line is dropping far slower than any obvious single quarter revenue dip. So extrapolating from those is less than worthless, because historically speaking they've never been consistent.


I would say that unless you want to do very sophisticated analysis of the numbers, I'd stick to looking at annualized trends of the data only. The data clearly shows there has not been a "catastrophic" or dramatic drop in CoX subscriptions or activities. But the actual magnitude of the general trend in specific terms has a lot of margin for error.