Hacker News Reader

Comment by joosters | Hacker News Reader

Same here. If they were really cunning, they could fabricate their own signal with a different timestamp. You could shut down rivals by making their computers believe that the trading day was over, or that their market feed was too slow to trust...

View on HN

Most people are using GPS-disciplined oscillators, not raw GPS clocks, so don't need GPS coverage except for very long-term stability. The crystal in the oven stays plenty stable over the timeframe of hours and days, and takes priority over any shady-seeming GPS updates. The worst you can do is cause a loss of a few ppm clock stability. (These are only $200 on eBay and I have one at home. From time to time I have a bad GPS signal and my consumer-grade ntpd marks the GPS clock as a "false ticker" and falls back to the PPS signal only.)

You could do that over very small timeframes (say, a millisecond earlier or later) but most systems don't automatically adjust times if the delta is very large

A millisecond is a long time.

You can make a million dollars in a millisecond.

I wasn't being sarcastic.

For regulatory purposes, milliseconds are negligible. FINRA allows for up to 3 seconds of divergence from standard time sources.

For deep alpha signals, a millisecond is pretty negligible (especially when you are calculating minute or 10 minute bars).

For microstructure trading strategies, since the timing doesn't manipulate the book (which is where most microstructure trades derive alpha from), the effect is similarly negligible.

And for cross-exchange and cross-asset arbitrage, where microseconds may seem matter (regnms), slight drifts are permissible insofar as you can plausibly argue that even if the clock is slightly off, your view of the market is internally consistent

Shift of the clock signal by milliseconds may well not matter, for sure, assuming all relevant clocks are affected.

Reply on news.ycombinator.com