Let's compare the two services.
To release a game on XBLA you'd need a development kit, which aren't terribly cheap - last I heard they were many thousands of pounds. To develop an XBLIG title you need an Xbox 360, a PC, a live subscription, an XNA Creator's Club subscription, and some bits of hardware for testing (memory cards and the like), totalling a few hundred quid.
XBLA games need to pass Technical Certification Requirements, of which there are around 130, all of which are explicitly stated. If you can justify a good enough reason to break one, you may get away with it. XBLIG titles need to pass peer review that, whilst sometimes whimsical and overrun with pedants, has a core set of around 20 tests.
To illustrate how much overhead XBLA certification may add to a project, consider the number of DBP winners who have yet to see their game on XBLA, or have had their games released but took a great deal longer in doing so.
XBLIG allows any developer with an XNA CC account and the skill to get a game through peer review to release a title. Some would argue this is the service's biggest weakness from a consumer's perspective, as it means there's an unending torrent of dross and 'Hello World' style games. XBLA games are published by publishers (duh!) with an established relationship with Microsoft. They get a certain number of slots to release things in the ever-more-crowded calendar, so need to be reasonably picky about what they release. Particularly outstanding games may be published by Microsoft directly.
The value of publishers can't really be understated. Much like record labels in the music industry they work as quality filters, finding things that are likely to be well received and promoting them. Hence anything that a publisher (or indeed record label) promotes is likely to be received with less contempt by the press than a complete unknown.
With a publisher in the loop, revenues will be split between three parties. If an advance was paid, this will need to be recouped before royalties reach the developer. Without a publisher in the XBLIG model, revenue is split between Microsoft and developer in a respective 30/70 split.
XBLA games are available to buy in 26 countries; XBLIG in 10. There aren't any hard figures on the matter, but small-scale research by the likes of myself and general common sense shows that there are a huge swathe of Xbox LIVE users who don't touch the XBLIG service. Visibility of the service itself and finding games within it remain problems.
XBLA games get a lot of mainstream gaming media attention, and due to the fact all XBLA games are published they are likely to get a better reception from the press at initial outset. XBLIG titles are often featured in brief 'ace/arse'-type articles separating wheat from chaff. The XBLIG ecosystem does benefit from a dedicated and enthusiastic fanzine-esque community.
Here we are with the real kicker. Official sales data for XBLA are not released, and XBLIG sales data are provided at the will of the developer, and can't be externally verified. Hence we should take all figures and comparisons with a pinch of salt.Pricing
Digging through VGChartz' sparse data isn't going to yield anything that enables me to make any confident and accurate statements, so I'm not going to get bogged down in the numbers. Suffice to say that the biggest sellers on XBLA have generated a lot of revenue - 7 (probably 8 figures by now) for games like Castle Crashers. The poor-performers have sold as badly as the best-sellers have done well, with some titles I'm aware of selling less than 5,000 copies over a lifetime, despite advertisting campaigns, 7-figure development budgets and positive reviews. These titles are typically 800MSP or more though.
XBLIG's best candidate for top seller has sold about 200,000 copies at a price point of 80MSP with Avatar Drop coming in a probable second with 120,000 copies. Considering the budget for each, I expect these have made an utterly fantastic return on investment. One only has to read the Share Your Sales Data thread on the XNA CC forums though to see the overwhelming majority of cases where games are selling tens or hundreds of copies, if that.
The poor ability of XBLIG to make a return isn't limited to amateur or low-skilled developers with a sub-standard product. I know of one developer in particular who has released a series of acclaimed and popular XBLIG titles, but is bearly breaking even due to a combination of their business decisions, the cost of creating a competitive title, and the low retail value of the average XBLIG game.
XBLA titles retail at 800, 1200, 1600 and 2000MSP, but how often do we see an 800 point title these days?Technical Constraints
XBLIG titles can sell at 80, 240 or 400MSP if they're 50Mb or less; 240 or 400MSP otherwise.
It's telling that experienced marketers are driving up the price of XBLA titles making the 800MSP price point a rarity, whilst those in the most crowded marketplace are racing to the lowest possible retail value.
The cheapest XBLIG title would have to sell ten times more copies in order to generate the same amount of revenue as the cheapest XBLA title. So whilst XBLA games still have the ability to tank, a low (relative to the platform) number of sales still brings in lot more cash.
If Microsoft like you enough, there's pretty much no limit on file size these days for XBLA. Well, I tell a lie - apparently the infrastructure doesn't like downloadables over 2Gb, but no doubt as soon as Bungie/Activision need to do something bigger than this, the constraint will be worked around. XBLA games can include up to 200 GamerScore points' worth of achievements, as well as leaderboards and downloadable add-ons.
XBLIG downloads are capped at 150Mb (with fairly good reason, given the volume of titles, and the number of intermediate steps that binaries have to take through Microsoft's servers). No achievements (again with good reason), no leaderboards, no DLC and also no auto-updating.
I'd argue that it's not a clear-cut case of which service is the best choice if you're fortunate enough to be in a position to choose between the two. Once again I'll refrain from naming names, but there's more than one XBLA developer looking at testing the XBLIG market.