Don Marti
Tue 11 Apr 2006 09:08:45 PM PDT
Software Deficiency Market
(This is a post to the fsb mailing list, expanded and with some links to similar ideas.)
What if a cross-project bug tracking system got a full-blown prediction market instead of just a bounty mechanism? Users could trade futures on a change in status of a bug, and anyone could invest on either side.
For example, if the program "exampleview" did not support PNG images, a user could file a wishlist bug "please support PNG" and then place an offer to buy a contract on "Not closed successful by 30 September". If the user had to open and view a PNG on 1 October, and if exampleview didn't do it by then he would have to license a copy of ProprietaryImageViewer Pro, then the user could use the market to hedge his exposure to the software deficiency.
The obvious counterparty in the transaction would be an exampleview developer, who would take the other side of the contract, check in the feature, change the bug status to "done" and wait for an impartial party such as a project leader, distributor or aggregator to close out the bug, upon which she would get paid.
So far, what's so much better about this than a bounty system? Nothing if there are only two participants, but opening it up lets anyone play the market, and allows for diffusion of information across projects by arbitrage. A user could place a complicated request that would be best filled by cooperative action across projects, and an arbitrageur could transfer the incentives to the places in which it would make sense.
There is the problem of gaming the system: if you're a speculator and you know that joe@example.org gets up at 10:00 and wouldn't let a certain bug go unfixed, you could get up at 9:00 and snatch away what would have been his rightful bounty under a simple bounty system. But the user's software deficiency exposure is already hedged, so the user is taken care of. The open questions are whether the information and liquidity added by a prediction market offers more value than what gets siphoned off by the speculators, whether it's easier to game a prediction market than a straight bounty system. Most markets like having speculators.
A common problem with bounty systems is matching many small unclaimed bounties to developers. Adding the speculators and the arbitrageurs to the mix would give the system some more liquidity, and give participants who are not in a position to close out a bounty a chance to feed information into the system. Since the simplest set of transactions in a prediction market is exactly equal to the bounty system, a prediction market can't do worse than the bounty system.
A BTS/prediction market would also provide for realistic predictions of software feature completion dates.
In the long run, this kind of thing could hedge exposures on larger scales, too. The "late release that killed Christmas" problem would be gone, since people with knowledge of the development process would already have been trading on it.
Notes
Chris Rasch: Wall Street Performer Protocol
Paul Harrison: Rational Street Performer Protocol
Fundable.org lets you place conditional pledges and pay only if your outcome happens.
The Launchpad bounty system
Project-based markets: guru.com and rentacoder.com.