Algo's reaction time in very short time-frame

[pstrusi]

Hi Ninjas,

I'm working in a code which might work in a very short time-frame environment. Let's say few ticks, or few points in Bar range chart for any popular liquid instrument. My question is simple:

In your experience, and having in mind that you're working with a very simple script:

- How fast could the link NT-Interactivebrokers work ?
- Could it that link, process a couple of trades in a second for example, if Algo required to get those trades done? I'm not talking at all about High Frecuency trading Algo, just an end-user's typical Algo.

Thanks for your attention

[NinjaTrader_PatrickH]

Hello pstrusi,

Thank you for your note.

Standard latency will apply to all orders sent to the broker via NinjaTrader or TWS. A few trades within a second should in general be perfectly fine without any abnormal delay.

Please let me know if I may be of further assistance.

[pstrusi]

Hi Patrick, thanks for your response.

- How much time does standard latency last?
- Can you elaborate it a little bit more when you have time, this is not an issue

[koganam]

Hi Patrick, thanks for your response.

- How much time does standard latency last?
- Can you elaborate it a little bit more when you have time, this is not an issue

There is no thing as "standard latency". Latency will always depend on the hardware and software and environment conditions prevailing at any given time. These factors include the current executing speed of your computer; the current speed and state of EVERY router between any 2 data points (ie., the state of the actual internet routing between the exchange, the broker, the charting package, the data vendor and your computer).

ref: http://en.wikipedia.org/wiki/Latency_%28engineering%29

[pstrusi]

Thanks Koganan for sharing your knowledge helping us. All what you say makes sense. I assume that you're a systems engineer or similar.

I'd like to ask you if you have experience in this world of producing and selling automated trading systems. Imagine for a second that you could have a very interesting system, and besides putting to work for you, what steps will be necessary in order to sell it to a bank investment, broker, mutual funds...etc I imagine that you should have kind of proven track of records at least.

Thanks again

[NinjaTrader_PatrickH]

Hello pstrusi,

Thank you for your response.

Standard latency would be the distance between your PC and the broker's server, as well as the distance between the broker's server and the exchange. As the order is submitted to the broker's server from your PC and then to the exchange.

[koganam]

Thanks Koganan for sharing your knowledge helping us. All what you say makes sense. I assume that you're a systems engineer or similar.

I'd like to ask you if you have experience in this world of producing and selling automated trading systems. Imagine for a second that you could have a very interesting system, and besides putting to work for you, what steps will be necessary in order to sell it to a bank investment, broker, mutual funds...etc I imagine that you should have kind of proven track of records at least.

Thanks again

Before a bank will even talk to you, you will need to be introduced by someone that they already know, or else you would yourself have to have so much money with them that they are scared to not kiss your feet when you call. Pretty much the same for a mutual fund. Both of these entities, if they trade, would have their own internal teams to take care of quant trading. These teams are usually very well funded, and have a high concentration of computational brain power, which is why banks and mutual funds are unlikely to be interested in anything you develop, until it has already, provably, made a boatload of money while trading real money.

Now with a broker, you might have an in. The first thing would be to open an account at that broker, then trade the system exclusively, until it has a proven track record. Typically, if the system is making consistent large profits, THEY will contact you. In any case, the reason why you must trade the account with them, is because they have all the records and because of new rules, machine generated orders carry a marker, so they can verify that all those orders were actually machine generated, not you manually clicking a button.

That having been said, it will probably take quite a while. You can of course approach the broker yourself if you think that you have generated a sufficient record. That will at least put you on their radar. Needless to say, you must choose a broker who has a facility for autotrading by machine algorithm.

Latency is not an issue for the systems that I create and trade because they do not require that kind of precision (deliberately). After all, as is typical for a retail trader, as opposed to a HFT scalper, I am going for large targets, and entering with Limit Orders. Either they get filled or they do not.

If I were going for small targets like 2 to 5 ticks, that would be a different matter, as that would require quick entries and exits. I had no success scalping manually, and even though I tested successful autotrading scalping systems, the commission levels were unacceptable. Given the smaller commission levels that a large institution with seats on the exchange might have, such a system may be viable, but I am not in a position to live test such.

[pstrusi]

Thanks Koganan for sharing you experience about it. I've had some time collecting this kind of experiences for an eventual probable future on myself. I hope to begin shortly a decent track record.
By the way, another important point: security of your script. I've read a lot and how to protect your codes from being copied or another similar unauthorized use.

There are obvious steps:
1. Having a physical place with good security
2. Compile your code with some encryption program
3. A reliable IP provider, with router configuration that doesn't allow external connection

In your experience, having a good code, which steps will you take in order to protect to be stolen, copied...etc ?

I visited your site, very interesting approach with that kind of indicators, I hope you're doing just great with them.

Thanks again and all the best

[koganam]

Thanks Koganan for sharing you experience about it. I've had some time collecting this kind of experiences for an eventual probable future on myself. I hope to begin shortly a decent track record.
By the way, another important point: security of your script. I've read a lot and how to protect your codes from being copied or another similar unauthorized use.

There are obvious steps:
1. Having a physical place with good security
2. Compile your code with some encryption program
3. A reliable IP provider, with router configuration that doesn't allow external connection

In your experience, having a good code, which steps will you take in order to protect to be stolen, copied...etc ?

I visited your site, very interesting approach with that kind of indicators, I hope you're doing just great with them.

Thanks again and all the best

Frankly if you are providing the script/dll to a third party, physical security, and man-in-the-middle attack considerations become rather pointless.

Unfortunately, protection of .NET executables is really almost a misnomer. Just about all half-way successful schemes to protect .NET executables are really security-by-obscurity: there is just too much information, (in the form mostly of metadata), that is leaked by .NET executables, to have a really robust protection solution.

[pstrusi]

Imagine for a moment that you have an interesting script to work with in behalf of group of partners, only you have and run this code; so despite there's no such thing as robust protection, what steps would you take, at least, in order to protect your code from reverse engineering ? I'm aware of some suggestions like the Protect compiled assemblies from SecureTeam's ClicSecure, Themida...etc but I've read too that so far, there's not any .NET file that can't be decompiled, so a real difficult task.

Best

[koganam]

Imagine for a moment that you have an interesting script to work with in behalf of group of partners, only you have and run this code; so despite there's no such thing as robust protection, what steps would you take, at least, in order to protect your code from reverse engineering ? I'm aware of some suggestions like the Protect compiled assemblies from SecureTeam's ClicSecure, Themida...etc but I've read too that so far, there's not any .NET file that can't be decompiled, so a real difficult task.

Best

My answer really does not change. Securing .NET code is just not a robust option. You can use any of the available obfuscation tools. Unfortunately, they will not provide the kind of robust encryption that can be applied to native code. You also need to test each of them against the available decompilers. A whole lot of the obfuscation tools that are available are one step worse than useless.