Here is what NinjaBee (Cloning Clyde/OKX) had to say about this method:
Estimating Live Arcade Sales from Leaderboards
I've noticed a few people making a direct correlation between the number of entries in the main leadboard of an Xbox Live Arcade game and the number of actual sales of that game.
Without revealing any specific top-secret numbers, I think I can say that my experience shows this to be a fairly accurate measure of sales. So far. However, there are some reasons why this could be less than totally accurate. I'm going to jot down some thoughts I've had on the subject.
1. You have to own the game to have an entry on the leaderboards.
This is the primary reason people consider this a valid estimate. You can't have a place on the leaderboards unless you have an official unlocked version of the game.
2. Multiple people on the purchaser's console can play the game and post to leaderboards.
This is a cool feature of Live Arcade - if you download and purchase the game, your girlfriend/boyfriend/spouse/kids/mom can play the game, too, and potentially post to leaderboards. This skews the leaderboard-to-purchase ratio way up, potentially, since you could in theory have a dozen or more players in the leaderboards from a single sale. In practice, this may not be that big a deal.
3. People with free full copies of the game can post to leaderboards.
This includes free press downloads, complimentary copies the publisher arranges to hand out, and other free cases. For instance, we arranged for a number of free copies of OKX for reviewers on serious game sites. I have no idea how high this number is for most XBLA games, but it seems like it will be fairly high. At least a few hundred, I would guess. How many people with this kind of access to the game will post to leaderboards? I don't know, but it skews the leaderboard/purchase ratio higher again.
4. Not everybody posts to leaderboards.
You might buy the game and never play it while online. People in your house may have access to the game, but may not have Live accounts and therefore not be able to post to the leaderboards. You may simply play non-default modes in the game and end up posting to some leaderboards, but not to the "main" leaderboard. This brings the leaderboard/purchase ratio back down a bit.
You'd think these things would conspire to make the leaderboard entry count a too-high indicator of sales, right? It didn't seem to for us, but we don't have a ton of data to work with, since the service hasn't been out that long. Maybe the first quarter of sales is a bad example, and we'll find out this quarter how far off it really is!
This technique for estimating sales also completely ignores additional downloadable/purchasable content, such as levels and gamer pictures. Of course, if a downloadable level has its own leaderboards, you can make new estimates based on that.
How to see the bottom of leaderboards
I don't know if any game has a feature to jump to the bottom of a leaderboard. I mostly doubt it. The way some interested parties arrange it is to score very low on the leaderboard, to be fairly sure they're at the bottom, and then look at their ranking to find a number of entries on the leaderboard. My theory is that eventually so many people will be doing this that the main leaderboard of any game will have a pile of players contending for the bottom spot... :)