monpere currently all the conditions must line up on on bar close of of trigger bar. When all conditions are met on that last tick (bar close) of trigger bar order will enter on first tick (open) of new bar.

When I manually traded my strategy I would put a stop limit in market ahead of time as you mention. I have never coded my auto strategy to put order in ahead of time. The strategy would essentially have to forecast what would happen on that last tick. This can be done but would require a lot more coding. I did like getting those orders in ahead when I manually executed my strategy. However, with the kind of fills I am getting I don't miss having to get them in ahead and just enter with a regular limit.

How many orders did you take during those 3 months with all getting filled?

monpere around 50. My strategy is a day trade strategy but takes a fairly low number of trades compared to some but has fairly large targets. I am not sure if would get same results from something like what you do with 10 or 12 trades a day. However, I can say my limits on CL would get jumped on a weekly basis and this no longer happens at least over last 3 months. I would get jumped estimated around 1 out of every 4 or 5 trades depending on market. So my improvement has been significant even over a low sample of trades. Over the last few weeks with market moving so fast I expected to get my limits jumped but was filled every time. My guess is would have a very positive effect on a scalping strategy although might not eliminate all jumps. Will be interesting to see if I get any jumps over a 12 month period. Would also be interesting to see how would do on a scalping strategy.

I get around 10-15% no fills with automated entries on bar close. It has become an acceptable occurrence for me now. I get fewer trading the signals manually, because I have my stop/limit waiting in the market. I can pretty much guarantee that it would improve my automated strategy, because a good 90% of the trades that don't get filled go to target within seconds. The reason they don't get filled is because the price is moving so fast towards the target and I'm to slow to get picked up on the way.

monpere around 50. My strategy is a day trade strategy but takes a fairly low number of trades compared to some but has fairly large targets. I am not sure if would get same results from something like what you do with 10 or 12 trades a day. However, I can say my limits on CL would get jumped on a weekly basis and this no longer happens at least over last 3 months. I would get jumped estimated around 1 out of every 4 or 5 trades depending on market. So my improvement has been significant even over a low sample of trades. Over the last few weeks with market moving so fast I expected to get my limits jumped but was filled every time. My guess is would have a very positive effect on a scalping strategy although might not eliminate all jumps. Will be interesting to see if I get any jumps over a 12 month period. Would also be interesting to see how would do on a scalping strategy.

After much deliberation I just set up a dedicated server for my auto trading strategy via Ninjatrader. I am writing this post to provide information to others who might be considering the same path. I debated doing this for the last year.

I would not recommend going down this route if you manually make trades as probably will not help you. If you trade manually may even slow you down because placing orders through remote server will eat up any latency advantage you may have gained. Also reliability while important is not as critical on a manual machine because you are always watching it and can call your broker and get out of trades if your machine goes down or disconnects. But anyone who has an auto trade system should seriously consider the benefits of using a server closer to the exchanges and the reliability a good data center can provide. Being on auto pilot has certain risks that can be lowered although not eliminated with a dedicated machine in a data center.

The system I was using at home was very fast and very powerful with decent latency (for home system) but with my hard earned capital at risk everyday I wanted something better. I will continue to use my home system as a backup and for backtesting. So my investment there will not go to waste

My main two reasons for doing this were latency and reliability.

I am paranoid about my auto trading system. There are just so many things that can go wrong and I want to eliminate as many problems as possible. They all can't be eliminated but where I can I want to protect myself. I try to watch my trades once my system alerts me that one has been taken. But there are times I can't watch which has it's risk. So system risk mitigation is paramount especially when I can't monitor a trade directly.

Latency

After switching to dedicated server in Chicago I now have latency in low single digits which makes me sleep better trading CL crude futures. Latency at home was 43 to 53 ms to Chicago. There was no way to overcome the distance from my home to exchanges in Chicago. So solution was to get a dedicated server in Chicago. Wow what a difference a closer server with direct connections into exchanges makes on a ping. I cant even get single digit latency by pinging a server in my own town on my home system much less to Chicago. Going from 50ms to single digit mm is quite a difference and worth the cost of a dedicated server close to the exchange even without the reliability advantages.

Reliability

While in some ways I liked the fact my home trading machine was sitting right beside me I could not reproduce the reliability of an outside dedicated server. Redundant power back up, internet connections etc can be reproduced at home to a degree. But not at a level that I was ever comfortable with or found affordable. Lets face it I am not going to hook a generator to my house in case power goes down. I also found dual wan router situations to work in a limited way but was not satisfied the way NT handled disconnects. The seamless way a data center handles redundant internet connections should work much better. I also really like the fact if my outside server goes down I will be notified via email because I have it ping monitored. If I was out and my home trading machine went down could not be notified because machine down sitting behind router could not notify me since it was down after all. Maybe third party services that could overcome that but I think would have problem accessing machine disconnect behind router.

I decided to go with steadfast networks Steadfast Networks :: Chicago Web Hosting, Dedicated Servers, and Colocation. They seem to get great reviews and I like the fact they are largely geared toward traders. I also like the fact their servers sitting at 350 E Cermak Rd, Chicago, IL. It is my understanding they are directly connected to the exchanges thus the incredibly low latency. I can't comment on whether they are a good company or bad other than reviews I have read and my account setup experience. But after a few months of using will update based on my experience.

My experience from signing up for a windows 2008 machine to deploying my NT setup on the machine took maybe 90 minutes or so. So was easy and smooth. Email tech support was fast and responsive to my questions during setup.

The dedicated system I chose was the

Atom D510 1.66Ghz 1MB Cache
500 GB SATA II Hard Drive
2 GB DDR2 SDRAM


The above cost $99.95 plus $20 for windows 2008 per month. I was a little hesitant about the Atom D510 having enough power to run NT. But seems to be running it fine so far. I may be tempted to bump my specs latter but the cost almost doubles so we will see how the Atom does. I will not use it for backtesting so I think will be fine.

I chose dedicated over a VPN or cloud solution. Using a VPN or cloud would lower costs but risk reliability. Most VPN's are oversold and I did not want other users on the same VPN to potentially bring my system down. As far as cloud goes I think you run into same problem but even on a larger scale. So the risk to reliability just not worth the savings imo. One bad incident could cost exponentially more than the difference between Dedicated and VPN. If my goal is to eliminate as many risks as possible I determined dedicated was only way to go.

If anyone has any questions I would be glad to answer about my experience. My intent with this post was not to plug steadfast networks but to try and relay my experience thus far as I know many traders debate going down this path.

Update 6/8/2011:

I have been running NT7 with couple of live charts on dedicated server for a few days. I mentioned in my original post that the configuration I went with was the Atom D510. I also mentioned my reservations about the Atom D510 having enough power to handle NT.

I have been running performance monitor in windows 2008 and for the most part the processor is only taxing system around 5%. So to live trade via NT the Atom 510 seems to handle it with no problem. I have a much more powerful system to do my backtesting on so will not use the Atom in that regard. Only thing that might give a performance bump is to upgrade system with a raptor or ssd drive as well as more ram. Although I am not sure that will cause trades to be executed faster.


Update: Latency

I have been running ping tests to 64.202.118.5 which from what I understand is a close as you can get to zenfire servers that allow a ping. I don't think this is their actual order server.

I am getting sub 1ms ping which is fantastic. Windows only measures in 1ms increments so don't know how far below 1ms. It is so low I wonder if those servers are not in same building. That is what you call blazing fast.



Update 7/6

Update 7/6

Okay here is the low down with Steadfast Networks.

I ended up moving from the Atom. Even though it may run NT okay most of the time I think it is under powered. So I bumped up to a Core duo machine that should run NT7 with ease. Unless back testing I do not think NT really uses more than 1 or 2 cores anyway. I wanted to get away with using the Atom even though I knew better. I liked the cheaper price but if going to pay money for server really is better to pay extra and get what you need. I am a penny pincher but in the trading world there areas you should not pinch. Actually the cost is quite small compared to trading profits and a year can be paid for in a few trades so why pinch.

My initial exchange with Steadfast on the matter was not so good but was a misunderstanding. Steadfast bent over backwards to keep me as a customer so I am happy. I do like the fact they concentrate on traders. They do not focus on leasing servers for web hosting. I think they did at one time but now concentrate more on the financial markets.

Their prices are not the highest out there but not the lowest either. But I think you get a much better network than most. I have spoken with others about other Chicago based server providers that even have very low latency and offer some better deals hardware wise but have nightmare stories about servers going down, disconnections etc. However, Steadfast Networks has a very good reputation among users which is why I am willing to pay more. There are times you get what you pay for especially in server market.

I will update again in the future any network or hardware issues that arise or don't arise and how Steadfast responds.

Update 8/23/2011

After 3 months of using a dedicated server in Chicago with sub 1ms ping I can say it makes a significant improvemnet to my fills. I typically use limit orders on my bot. When I was using my own machine at my house even with decent residential latency it just does not compare to being in Chicago.

I use range bars and if a limit gets jumped on bar open and next bar prints my entry orders will cancel by design. So I would quite often miss entries that never would get filled. To many jumped orders creates problems with my expectancy. I am happy to report after switching to a server in Chicago in 3 months I have not missed one entry. I get filled every time like clockwork.

hi, thanks for the info. I am thinking of getting the same myself. D you recommend windows 2008?

hi, thanks for the info. I am thinking of getting the same myself. D you recommend windows 2008?

Thanks Liquidcci for this important information. Below 1ms is indeed blazing fast, I'm getting an average of 111ms with 20 pings to this address (The Netherlands is located 6700km, or 4100 miles, from Chicago).

For the ping, the signal has to move 13,400 km to Chicago and back to your place. Light travels at the speed of 300,000 km per second in vacuum. For the speed of the signal in a medium, you need to divide that value by the refractionary index. For optical fibre this index is 1.52, so you can assume that your signal travels at a speed about 200,000 km per second.

That means that the best - theoretically obtainable - ping value from the Netherlands to Chicago would be

13,400 km / (200,000 km / sec) = 0.067 seconds = 67 milliseconds

You can't get below that. Add something like 1 millisecond per hop, and you know where you can get at best.