2 ms·
From talking with a friend who's a retired stat arb quant, the fundamental idea is sound. My guess for why people aren't doing it yet is that first, there are
by jasonwatkinspdx 13y ago
From talking with a friend who's a retired stat arb quant, the fundamental idea is sound.
My guess for why people aren't doing it yet is that first, there are still untapped locations on land that are easier to get running (example: the Chicago to NY fiber run), and second getting the data to the ocean point isn't trivial.
I think if you want to look for early signaling of someone doing this, look for where telco folks route new fiber runs. If they take a longer path to go near one of the points in the paper I'd bet they're at least hedging for the possibility of using that fiber run for arbitrage.
Edit: a more mundane explanation may be that the arbitrage is so low margin that the payback time isn't practical vs the initial investment.
- ramchip 13y agoI don't understand why it's sound. The order still has to go all the way to the exchange. Why is it better to spend 50 ms to get the market data, 1 ms to do the decision, then 50 ms to send the order (middle of ocean to exchange), when you could spend 100 ms to get the market data, 1 ms to do the decision, then 0 ms to send the order (colocated hosting at exchange)?
- bri3d 13y agoYour algorithm is making decisions on data that's 50ms older than the data the other guy in the ocean used, I suppose. I'm not into low-latency trading enough to know whether or not that 50ms-ahead data alone could make you money, but based on how much the trading guys go nuts over it I assume it could.
- ceejayoz 13y agoBecause sometimes the order you have to place is on the other side of the ocean where the data came from, costing you an extra 100ms.