In futures (ES) i found it was best to "SEE" the technical setup being formed, then put in your limit orders for entry and have the stop loss order ready to go as "AS SOON" as you get filled. Then i can not blame a missed trade on a ping being slow.
For automated trading you can do this: buy a U1 rack machine, set it up and ship it to co-location service, See Paul Vixie's website http://www.vix.com/personalcolo/ if you go with a good provider, located in NY, London, or wherever, you may be abler to cut down latency, and increase reliability. I find that even with a decent connection at home, you get dropouts where the network seems to go away, then come back a few seconds later. So, if your box does not need your intervention for every trade, only supervision, i would definitely go with this setup. The coloc costs from $50 a month upward. K
This is pretty much true, if the traders who compete against each other use EXACTLY the same exact strategy. But, if you developed a successful system using dsl, dial up, smoke signal, etc....there is no cause to believe that being faster will make you more successful. In fact, it could screw it all up. There are elements of latency (relative latency) that are part of each system. Timing is so important in entering and exiting trades, and latency is an element of timing. If anyone here is a musician, I think that their knowledge of different timing and rhythm would help them understand better what I am trying to say.
Quote from pairsarb: ...there is no cause to believe that being faster will make you more successful. In fact, it could screw it all up. There are elements of latency (relative latency) that are part of each system. Timing is so important in entering and exiting trades, and latency is an element of timing. I can confirm this. As I improved my system speed, there were situations in which I performed worse, and had to re-introduce some delay to take advantage of certain market characteristics. Imagine what a reduction in bandwidth/CPU/brain requirements we would all experience if the bots stopped trying to compete in single-millisecond timeframes and paid more attention to quality and strategy instead.
Just to put a different spin on things, why not consider a faster platform, say X Trader through Velocity as opposed to J Trader which has been demonstrated to be 1/2 second slower. I don't know if this was mentioned, I didn't read the entire thread. I use X Trader and I am very happy with it and Velocity, it is lightening quick. Also FTP, dedicated fiber to the home, is available in a few markets, with more to come, Qwest has it here in Denver, in a few places. Go cable over dsl always, it is faster, unless some weird anomoly. Good luck. Additionally, if using cable modem, make sure you have a newer cable modem, I get faster speed now using a Motorola SB cable modem as compared to my outdated toshiba modem that was 4 years old. Apparently it can accept a faster speed, but don't quote me on that one. Make sure you have the fastest computer you can afford with alot of RAM. Lastly, if you are big time, consider getting dedicated T-1 or 2 bonded T-1's, 1 T-1 may be sometimes slower than cable, but it is only your bandwidth and you get better service response should there be a outage.
Probably correct. There are quit ea few times where a very brief lag would result in a better fill. ie, a flush that attracts bids. If you are too fast selling the flush, you get the worst price. With a plain old cable modem and IB , I'm already subsecond. Who is trading a system here that depends on 1/10 second latentcy?
latency is a funny thing; you need some slow down to attract bids and offers, but I would think that over all, the fastest ping times but also the best in breed speeds of some of the automated/high volume brokers would be the best combination. I think traders get caught up in single ping times, but really, trading electronically is the combination of where their server is (is it co-located?), the internal processing time of the broker (I think this is overlooked, dramatically, by the speed obsessed) and the ability of that broker to not only get those orders out to the ECN"s, but also what is done with those orders. Are they just sent, but not accepted? Have they been replicated anywhere so that the system can cancel or cancel replace? So, the latency needs to be low in sending, receiving, processing, really everywhere. It all adds up, and you really can't measure your loss to someone who may be quicker an on a more reliable system. Low latency and high reliability in tandem is probably the most important concern you can have. A simple ping test doesn't measure the efficacy of a high quality brokerage.
what about the matching engine algorithm? a FIFO (first in first out) fills on time stamp priority - so latency would be an issue - esp if scalping. a pro-rata fills on size priority - so ping time is less of an issue. you can jump the queue so to speak - if you are big enough