So when does the poster go on sale? :)

I've never even visited metacritic, never heard about it until writers started talking about how much it influances the industry, and I've never seen any actual proof that any of it's influance is the result of anything but game developer paranoia. I can not wait till the fad involving it's existence blows over.

Rob, I know it must been a lot of work, but is there any chance you make the data set available to the public?

I agree with Toby, this would make a great poster or even t-shirt. Well done!

Thank you Rob and everyone for the compliments!

Cristian, I've included a link to download the raw data. Enjoy!

This poster doesn't discourage me. I now really want to analyze data across publications, and arrive a scoring function that is predictive of game sales. I'm surprised that gaming companies haven't done this already. Actually, why hasn't Metacritic done this? I could probably finish this analysis in a day if I had the raw data. If scoring and sales did not correlate at all, i could also solve for that. Oooh, I could even rate publications at their ability to be predictive of game sales. See, the final scoring function would have a coefficient from 0 - infinity for each publication. Higher weighted publications would be more predictive of sales. Muhaha I could probably influence billions of dollars in ad revenue. But alas I don't work for free.

Where exactly is this non-correlation demonstrated?  In fact, it shows the opposite; the Unit Sales Broken Down by Metascore graph is clearly showing higher scores are associated with higher sales!

Anyway, a 3-month data set is much too focused a sample to make generalizations like that.

You really have to look at the scatter charts in the original report -- when you run the numbers for Pearson's correlation, you don't get anything near acceptable.

Rob, I don't want to sound rude (and maybe this is a little more for the original post and not this one), but could describe a little your process? I took a quick and very dirty look at the data set and (granted, without accounting for the exponential consequences described by the VP of marketing) if you run a linear regression the scores certainly affect sales, and it is statistically significant. (Just saw your new comment, gonna check now for correlation)

Courtney, I'm pretty sure gaming companies have done this, like you say, a very quick look took me five minutes (sure, all thank to Rob who made the data set).

I looked at the r-squared values for each system's data individually. They hovered around 0.5-0.6, which is unacceptable for correlation. You'd want to have values above 0.9.

did you ran the regression with just one variable? r-squared suffer a lot from that (I'm even more courious now, I got a much worse R^2, 0.14, but that is to be expected, I don't think anyone here would argue that you can explain the sales more than that by just looking at metacritic scores). Also, sure, in class they'll tell you that you're looking for a r^2 around .9, but with real life sets that's really hard to find. But the p-values from the t-test and f-test doesn't allow you to reject the null hypotesis that metacritic scores aren't statistically significant.

@Rob, I found the same R^2 values as you did in fitting the data to an exponential curve, and I think the problem with Metacritic scoring has to do with how they weigh reviews. Giving weight to a review may actually come down to individual reviewers and publications. I wonder who the Bernie Madoff and the Warren Buffett of games journalism are?

I'm not so sure. Regardless of whether Metacritic weighs review outlets (a process that the site keeps secret), there's still this belief that reviews can drive sales, and I don't think that's the case at all. It's especially problematic when a developer's compensation is tied to these scores.

Certainly, other factors are much more influential, but my data just isn't equipped to reveal what those are.