I may have to put on my flame suit, but my inner statistician has wondered about this. The correlation isn't awesome (R squared is 0.17), but it is definitely not a meatball, either. I hope no one takes this too seriously - I don't.
I used the year-to-date average transaction price from the latest bostonpinball.biz Ebay prices and the IPDB rank as of yesterday. There are some problems that I noted. Some problems include:
1) Differences in ranking between LE and standard edition machines. In some cases, the LE ranked worse than the standard edition. Meanwhile, the Boston Pinball survey in some cases did not differentiate between LE and standard.
2) Some machines had a price sample size of one or no data was available. In cases where no data was available, I left the data out - the difference between 2013 prices and the long term (before "pinflation") data is just too large in most cases.
This kind of data obviously does not take into account many factors. Some things I can think of that make this type of correlation suspect:
1) The prices have not been adjusted for condition. Highly coveted titles with very good IPDB rankings are probably in better condition than less coveted titles, on average.
2) Hard-to-find machines (such as Safe Cracker or Cactus Canyon), which necessarily have higher prices, ranked consistently low on IPDB. Both the higher prices and low rankings are probably just artifacts of low quantities and would tend to throw off this correlation.
3) The data do not reflect each individual transaction. Therefore, a single data point may be 1 machine or 20 machines. That is a problem since, ideally, there would be 20 data points if a title actually had 20 transactions. This results in machines will fewer transactions having undue influence on the correlation compared to titles with more transactions.
4) The fit is definitely much higher than reality. The few titles fetching extraordinarily high prices move the best fit line up badly. (I would love to be able to sell my Judge Dredd for $5000, according to the best fit line.) As I noted, please don't take this too seriously.
Here is what I found interesting. I would have thought that prices would start flattening at around rank 50 or so, but the prices seem to continue declining within the top 100. Obviously it has to flatten out at some point, but it is not before 100.
UPDATE 1/10/14: I redid the correlation but against Pinside rankings. The some caveats apply, but the correlation is much better. R squared is 0.45 which most statisticians would accept as a good correlation.
One thing I noticed is that Pinside rankings are more up to date with the latest Stern machines and that IPDB tends to have somewhat older reviews. That may have something to do with IPDB rankings being somewhat better for older machines than Pinside rankings that seem to heavily favor DMD machines.
The distinction between Stern and Williams machines appears to be noteworthy. The Stern machines seem to have little correlation - the prices are all within a band regardless of rank. Also, if only Williams machines are considered (remove Stern data points), prices start to level out at approximately rank #25.