Develop DIY Algorithms—Safely

[MReilley]

Volpunter,

I'm assuming that when you mention "rolling window" you're referring to your 10 - 7 min bars.

In our example we demonstrated the capture of specific values based on each incoming live tick, which creates the values for our current bar (in this case we capture OHLC). This initial bar becomes our past 4 - 1 minute data points, or our "rolling window" (current and past 4 bars), with the passage of time. This time frame could be edited to your 10 - 7 minute bar setup.

As far as using this on 100-200 symbols, you just have to press play after you have created and deployed the algo...

 

[Andrew Renalds]

In terms of performance, it's not the necessarily the number of symbols as much as it is the amount of market data traffic. For the best possible performance, you should co-locate your Algo Strategy Engine on the same server as your TT gateway.

 

[volpunter]

So it means the strategy engine only works with TT gateway and hence only when subscribing to live data?

In terms of performance, it's not the necessarily the number of symbols as much as it is the amount of market data traffic. For the best possible performance, you should co-locate your Algo Strategy Engine on the same server as your TT gateway.

 

[Andrew Renalds]

No, it also utilizes the historical data server if your strategy requires it.

 

[volpunter]

Do you post costs of subscribing to such historical data server and the costs to subscribe to live data?

And what do you reckon why someone would want to consider your product over a pure oop language based solution given the code to generate a few custom bars a lot more complex and imho exhibits a steeper learning curve than doing it in a programming language of choice.

No, it also utilizes the historical data server if your strategy requires it.

 

[J.P.]

Do you post costs of subscribing to such historical data server and the costs to subscribe to live data?

They won't do it; they just won't do it. I looked into this before and I asked yet again in this thread at post #25.

They have what is touted as one of the best trading platforms ever. But, as near as I can tell, when it comes to actually selling this product, they have one of the most doornail-stupid marketing plans in the history of retail.

 

[Andrew Renalds]

Do you post costs of subscribing to such historical data server and the costs to subscribe to live data?

And what do you reckon why someone would want to consider your product over a pure oop language based solution given the code to generate a few custom bars a lot more complex and imho exhibits a steeper learning curve than doing it in a programming language of choice.

There is no cost for the historical data server software. There may, however, be a charge from the firm who hosts your TT software.

Think of ADL as a trading software library written and tested by TT. Of course you can write this library yourself. Or you even can write directly to the exhange. But why do this if the ROI isn't there.

The TT valud-add is that we provide a stable, normalized platform so that you can focus on trading instead of having to worry about exchange-specific nuances. And we offer a continuous product line of professional grade tools to meet the needs all types of traders. And we do this because not everyone is an OOP expert nor should they have to be in order to trade.

For someone with no prior knowledge of programming, I disagree with the assertion that they can learn OOP and then build a solid trading strategy faster than they can do it in ADL. And the fact that TT has hundreds of customers who have never learned to code yet use ADL successfully speaks volumes. My point is that you choose a tool based on the task at hand as well as the technical ability of the user.

If you have specific enhancements that you would like to see in the products that we offer or want to have a detailed conversation about particular topics, please send me your contact details so that I can be sure that I understand the details of what you're looking for. Have a good weekend.

Regards,
Andrew

 

[volpunter]

Thank you for the time to respond but I still respectfully disagree:

A) to consider investing time in any product I would need to see a clear and transparent cost structure. What is charged, in what way, by whom is highly opaque it seems and a big turn off to anyone seriously considering this product I believe. I only could imagine this product to be marketed as an ancillary product to those who use TT as execution vendor already.

B) we should compare apples with apples here: a fair comparison with your product would be a product along the lines of Amibroker or Right Edge. Especially RE offers a trading platform on top of which one can build C# based algorithms. Last time I checked It takes one single line of code to generate bars of custom compression from live data. The learning curve is significantly lower for someone trying to accomplish the same task than using ADL imho.I still fail to see who your target audience is. Someone not well versed in programming should never write trading algorithms regardless of language. It would be a recipe for disaster. And traders who generate ideas but do not know how to program should simply delegate such task to a capable developer. Where does your product fit in here? I fail to see it. My prediction is that your whole ADL department will be scrapped at some point and dismantled. You must be running the show at a loss to the firm. How do a few hundred customers pay the salaries of several developers, project managers , and marketers?

There is no cost for the historical data server software. There may, however, be a charge from the firm who hosts your TT software.

Think of ADL as a trading software library written and tested by TT. Of course you can write this library yourself. Or you even can write directly to the exhange. But why do this if the ROI isn't there.

The TT valud-add is that we provide a stable, normalized platform so that you can focus on trading instead of having to worry about exchange-specific nuances. And we offer a continuous product line of professional grade tools to meet the needs all types of traders. And we do this because not everyone is an OOP expert nor should they have to be in order to trade.

For someone with no prior knowledge of programming, I disagree with the assertion that they can learn OOP and then build a solid trading strategy faster than they can do it in ADL. And the fact that TT has hundreds of customers who have never learned to code yet use ADL successfully speaks volumes. My point is that you choose a tool based on the task at hand as well as the technical ability of the user.

If you have specific enhancements that you would like to see in the products that we offer or want to have a detailed conversation about particular topics, please send me your contact details so that I can be sure that I understand the details of what you're looking for. Have a good weekend.

Regards,
Andrew

 

[Gambit]

Volpunter has asked some pointed questions. I'm curious to see if Andrew and his colleague will respond.
I do hope TT will continue this discourse because it is enlightening to the ET community. Since Andrew has asked about requested features, I'll throw out some requests:
1) Stats functions such as cointegration, correlation. I understand that these features will be rolled out sometime in Q1 (?) 2016.
2) Kalman filter, realtime and integrated into the spreader. Or perhaps kalman filter as a standard block.
3) I'm not sure how TT handles its infrastructure but perhaps offering a standard configuration where the ADL algos are hosted on the same server as the TT gateway (not sure I stated this correctly). Users shouldn't have to pay a small fortune to outsource TT configuration in order to achieve competitive latencies.
4) Scanning functionality. Scanning for changes in cointegration in realtime, tick by tick so that traders can react quickly.

5) Perhaps integrated stats functions such as this one. This is a statistical calculation used to predict where the next bid ask is going which supposedly allows smarter limit order placement.

Volpunter is a skilled developer so his background definitely colors his opinion. As a "quantitatively challenged" home gamer, I'm obviously more interested in ADL. Perhaps there is a middle ground where both markets can be served.

One last question:
Since TT is obviously able to create standard code blocks and release them to their users, perhaps you could allow end users to utilize that same functionality. That way the community can grow a codebase that benefits multiple users. ISVs such as optionscity have implemented such features with some success.

I look forward to your responses Andrew and MReilly. And please don't request ET to ban members or limit members from commenting on your threads. The more you try to suppress criticism, the virulent it becomes.

 

[Gambit]

This is an algo ecosystem implemented by optionscity.
http://algostore.optionscity.com/storefront/