Custom Strategy Code (IP) - which platform(s) to trust?

[TradersRUs]

It's amazing that you are worried about alpha dilution even before you know that you have alpha.

What's amazing is that you would even consider trading in a live account (with presumably your own hard earned money) without having backtested thoroughly first. That's the very reason why backtesting software even exists in the first place! To give you an indication of alpha.

Calls into question your actual trading experience and your judgment skills... but we've seen enough of the latter already.

Go troll somewhere else please. This discussion with you has been a complete waste of time. You can have the last word (as no doubt you always do).

 

[userque]

I'm currently evaluating various platforms that allow custom development of strategies and indicators - Ninjatrader, Metastock, Tradestation etc.

It seems when you install the software (on your computer), there is no way to protect your custom strategy code from getting uploaded to their servers.

Are there any (installed) platforms that can be trusted to protect your IP, even though it is obviously visible to their software? Are other traders concerned about this and what is your preferred platform/solution to safeguard your IP (and thereby your "edge")?

Imagine a rogue employee looking to steal a great strategy. How would he go about it? He would likely need to be alerted when a great strategy has been discovered. He'd do this rather than vacuum every strategy of every customer, daily/weekly, imo. That's more work, before and after the download. And it would create a bigger data flow footprint.

So, I imagine a rogue employee would write code that would scan for strategies with great backtests or with great simulated account results, or with great real account results. Maybe his code only aggregates these results, and he manually examines them periodically. Maybe it's more sophisticated.

In your case, I would avoid running backtests with the platform. You can code your own backtests.

I suspect using alerts is safer than having the strategy make trades, real or sim. The thief would have to work harder to convert those alerts into a metric he could analyze.

I would also add fake, random alerts for further obfuscation. You, of course, would have some way to identify the real alerts.

Nothing is foolproof, but something is better than nothing.

Good luck and keep us posted!

 

[qaz]

Well, I do come here because I am bored so by that definition I am a troll. However, while I might be ignorant of many things in the world, as an institutional PM with decades of experience, I am pretty aware of what works and what does not. Seems like the members here feel that I got a lot to contribute.

In any case, I was merely pointing out that if you have not established that your strategy actually works in live trading (most things do not work, FWIW) it's not worth worrying about protecting your IP. The only exception to that would be if you are running some sort of an arbitrage strategy and literally don't want others to know that this arb exists. However, if you are REALLY paranoid, you can take a public domain backtester like backtrader or zipline, deploy it on your own machine and test the strategy without ever "giving the code" to anyone. If you truly think you've invented something that is unique (as I said, very unlikely in my experience), this approach is more scaleable and customizable. Pretty much every quantitative tool you might need has been implemented in whatever language you decide to use (C++, Java or Python) and you will not need to reinvent the wheel. Also, you can take the machine off-line and put it in titanium-lined safe for additional IP protection at the backtesting stage. When it comes the time to deploy it, the very same codebase can be connected to your broker API without every giving away your code.

I believe it is good practice for any trader with a viable systematic trading strategy to first look at securing his intellectual property, irregardless if he is a beginner or not. It is a one off value premuim to pay for protection against potential rapid alpha degradation.

 

[fractalize]

Not to sound dismissive, but I really doubt anyone trading through Ninjatrader would have come up with anything worth copying. Despite everyone claiming to make 100% per year, I have not seen that many people here that have a clue and have not seen anyone doing something completely unique.

Very much agree here. These retail platforms (Ninjatrader, Tradestation. etc...) lack the capability to perform the broad spectrum analysis needed to recognize alpha factors, as they are one-dimensional. No professional with a clue, would copy, or steal any strategies designed with them, and the probability of someone quantitatively maintaining consistent alpha with them, is pretty much nil. Most pros who build their own custom analysis tools also have integrated or companion execution apps. It's a completely different level.

 

[userque]

Very much agree here. These retail platforms (Ninjatrader, Tradestation. etc...) lack the capability to perform the broad spectrum analysis needed to recognize alpha factors, as they are one-dimensional. No professional with a clue, would copy, or steal any strategies designed with them, and the probability of someone quantitatively maintaining consistent alpha with them, is pretty much nil. Most pros who build their own custom analysis tools also have integrated or companion execution apps. It's a completely different level.

Ninjatrader fully uses C# and even links to Visual Studio.

So, is C# unable to "perform the broad spectrum analysis needed to recognize alpha factors."

Are you saying C# is "one dimensional?"

How much do you really know about coding strategies with Ninjatrader/C#?

 

[fractalize]

Ninjatrader fully uses C# and even links to Visual Studio

Sure, I get all that, as all of them have some type of linking capability. What I was trying to emphasize, is that most (profitable) professionals won’t deal with that headache, or dependency. They’ll customize for specifics, and to have full control.

 

[userque]

Sure, I get all that, as all of them have some type of linking capability. What I was trying to emphasize, is that most (profitable) professionals won’t deal with that headache, or dependency. They’ll customize for specifics, and to have full control.

I know of no other trading platform that links to a major, general language IDE. Correct me if I'm wrong.

But I do agree with your point: that's why I'm coding in pure C#, rather than NT. But you seem to suggest that NT isn't capable of doing what 'the big boys'/'pros' (I forgot your term for them) are doing. Capability and preference are not equivalent.

 

[Same Lazy Element]

But you seem to suggest that NT isn't capable of doing what 'the big boys'/'pros' (I forgot your term for them) are doing. Capability and preference are not equivalent.

I think it's less about the ability to do some specific computation but more about flexibility and scale. For example, can my platform take pre-computed PC factors and do run-time matrix calculations on 1k x 1k thousand matrix of stocks? Can I break calibration away from the runtime backtest? Can the platform import external fundamental data or alternative data, align it with the historical bars and use it in a backtest? Can it provide custom backtest analysis stack like volume participation, ToD trade bucketing etc?

It's not even about NT, I think I have said elsewhere that I have not seen a single pre-canned backtesting platform that can do everything needed for professional quant trading. I mean, it might be possible with some hacking, but if I have to do work every time I want to add a necessary feature, I might as well build my own.

 

[userque]

I think it's less about the ability to do some specific computation but more about flexibility and scale. For example, can my platform take pre-computed PC factors and do run-time matrix calculations on 1k x 1k thousand matrix of stocks? Can I break calibration away from the runtime backtest? Can the platform import external fundamental data or alternative data, align it with the historical bars and use it in a backtest? Can it provide custom backtest analysis stack like volume participation, ToD trade bucketing etc?

To the extent I think I understand this question. Yes. Again, can't all of this be done in C#?

It's not even about NT,

Well, for me it is. I know of no other platform that implements a mainstream general purpose language, and that links to the IDE utilizing the bells and whistles of coding assistance (intellisense) and debugging. Had you not included NT, I would not have initially commented, my greater knowledge in this regard is specific to NT and C#.

I think I have said elsewhere that I have not seen a single pre-canned backtesting platform that can do everything needed for professional quant trading.

If pre-canned backtesting platforms aren't marketed/designed for professionals, then this makes sense.

Was some professional in this thread asking for recommendations to professional software, and someone said NT? I don't recall and have been too lazy to look.

I mean, it might be possible with some hacking, but if I have to do work every time I want to add a necessary feature, I might as well build my own.

Not hacking: coding. All platforms come with precoded features. Coding additional capabilities is not hacking the platform, as that's how the platform is designed to be used. All platforms have feature requests. You can code your own; or wait months for the next upgrade. No platform has everything that everyone wants.

Yes, if there is a platform better suited to one's needs, it only makes sense to use that.

Again, I agree with your premise of using the right tool for the job. But I don't agree with saying that a tool that can implement the C# language deserves to be characterized as you did earlier. IMO.

 

[fractalize]

Was some professional in this thread asking for recommendations to professional software, and someone said NT? I don't recall and have been too lazy to look.

No, some non-professional who started the thread is struggling to comprehend why the chances of some NT "insider" stealing his golden goose are zero, since the vast majority of serious operators don't fool around with retail platforms. They build their own, or have someone build it for them. Not easy, but well worth it if you have clear cut specifications.