Wednesday 8 September 2010

Arbitrage finale

The Australian election result is finally out - Gillard takes it.

I know nothing about Australian politics. But I do know a bit about arbitrage. I trade pretty heavily on iPredict. Throughout the period, Labor traded at a premium (Coalition at a discount) on iPredict relative to BetFair and CentreBet. So I started by buying the undervalued position at iPredict, reckoning that NZ prices would move to match Oz prices over time. When that didn't happen terribly quickly, I started up accounts with BetFair and CentreBet to arbitrage.

In the end, my $927 on Liberals at iPredict turned to $0 while my AUD$925 at BetFair turned to AUD$1650 after commissions; $1160 at CentreBet will turn into $1718 when that market closes. I'm up around NZD$400 on arbitrage, about $100 on range trading at iPredict, another $50 from a last minute dumping of Liberal stocks on iPredict during the press conference.

It's more than possible to make a bit arbitraging across the markets. But unless you have fun doing it, it's certainly not worth the opportunity costs of your time. There's probably an impossibility theorem in there - if the opportunity costs of your time are low enough to make it worthwhile, you're likely not able to do it successfully. Unless you're a student, I suppose.

iPredict is relatively thin, so you have to accumulate a bit of position before taking the hedge if you want it to be worthwhile; otherwise, you're making tons of small bets at the other markets. If I knew how to program, setting up a bot that could automatically make trades across the three markets would be an awesome way of dribbling in nickles and dimes. Since it hasn't been done yet, it's probably not worth the opportunity cost of the programmer's time. And, even the seemingly most perfectly contrived arbitrage isn't always without risk.

As a final note, iPredict sensibly thins out the market maker in the last hour or so to limit its losses as we move towards certainty. It does this, for technical reasons I think, by moving to single unit MM price steps rather than by, say, dropping to every tenth step. On a busy market, that's really irritating - it makes it hard to see where the thickness is in the book. You have to go to the API to get the full order book, and that's a bit cumbersome when you're also trying to watch prices on another market at the same time.

Blogging will be light over the next few days as daycare is closed. But follow me at @EricCrampton for shorter updates.

No comments:

Post a Comment