Let's say I'm developing an HFT application and seeking arbitrage in futures markets between MAY contract(M) and JUNE contract(J).
In this strategy, my spread is
J-M. I did not check with real data but I think this is a mean-reverting series... This spread should be around the interest rate, right? In some short periods of time, anomalies occur and the spread gets greater than the interest rate.
My question is, how can I detect these anomalies. Does it work:
|spread(t) - mean(spreads in 1 minute)| > stddev(spreads in 1 minute) ?
Is there a more robust solution?