You are using an out of date browser. It may not display this or other websites correctly.
You should upgrade or use an alternative browser.
You should upgrade or use an alternative browser.
Deciding on a Backtesting and Trading Platform
- Thread starter Gyles
- Start date
I've sent an email to TBlox and TStudio for more info on a few points and await replies.
Another question on TradersStudio - one which i don't think has been covered on this or any thread. How good is TS's strategy building wizard? Perhaps not quite as easy to use as TBlox's?
I've assumed both can be used for datamining. That is right isn't it?
Syn
Another question on TradersStudio - one which i don't think has been covered on this or any thread. How good is TS's strategy building wizard? Perhaps not quite as easy to use as TBlox's?
I've assumed both can be used for datamining. That is right isn't it?
Syn
Quote from fundjunkie:
Multicharts has this problem but chose not to address it. I think you're right to look at redesign. Without turning this into a workshop on TS planned RT design can I ask how you propose to address the twin requirements of a tick handling application:
- Ability to handle and store symbols in bulk (sequential write heavy operations)
- Ability to respond quickly for specific sybols actually being traded
A database will always slow the second part down. So I hope you're looking at doing somethig like cacheing tickfiles on disk (for trading) and perhaps later storing in a DB for storage and simulations etc. I struggle to see how performance can scale if you try to do everything from a relational DB (too much overhead). I expect every company out there struggles with this question.
Thx
D
Really interesting points, if a little above my head technically speaking. Can we please have an answer Mr Murray Ruggiero.
Thanks
Syn
Quote from Synonym:
I've sent an email to TBlox and TStudio for more info on a few points and await replies.
Another question on TradersStudio - one which i don't think has been covered on this or any thread. How good is TS's strategy building wizard? Perhaps not quite as easy to use as TBlox's?
I've assumed both can be used for datamining. That is right isn't it?
Syn
I haven't used the TS strategy building wizard, but from memory it is not as good as TB - with the understanding that the basic TB (not Builder) is really a giant strategy building wizard.
I actually think TS is better at datamining because of the ability to customize reporting. Not sure if TB can do this. TS can basically create an Excel sheet of just about anything - it is very handy.
Quote from Synonym:
I've sent an email to TBlox and TStudio for more info on a few points and await replies.
Another question on TradersStudio - one which i don't think has been covered on this or any thread. How good is TS's strategy building wizard? Perhaps not quite as easy to use as TBlox's?
I've assumed both can be used for datamining. That is right isn't it?
Syn
You're labouring under false pretences if you think that either Tradersstudio or Trading Blox may support any kind of data mining functions.
They're productivity and research tools only. Stretching the definition to an extreme one might say that the productivity boost might enable data mining by shortening the research cycle. But in reality there is no connection at all.
Thx
D
Quote from droskill:
I haven't used the TS strategy building wizard, but from memory it is not as good as TB - with the understanding that the basic TB (not Builder) is really a giant strategy building wizard.
I actually think TS is better at datamining because of the ability to customize reporting. Not sure if TB can do this. TS can basically create an Excel sheet of just about anything - it is very handy.
Thanks for that. So yet again, (from my prespective) they both have something over the other! Both are appealing, but the flexibility of TS might be starting to won me over.
Syn
Quote from fundjunkie:
You're labouring under false pretences if you think that either Tradersstudio or Trading Blox may support any kind of data mining functions.
They're productivity and research tools only. Stretching the definition to an extreme one might say that the productivity boost might enable data mining by shortening the research cycle. But in reality there is no connection at all.
Thx
D
Perhaps i've misused the term datamining. Initially, without (or should i say before) actually developing a strategy, i want to be able to see what evidence there is to support a hunch i might have about a market's characteristics/behaviour. If something is encouraging, i then want to refine my investigations with the final outcome being the creation of a strategy/system.
Does datamining mean having a program trawl data first to see what characteristics/behaviours exist, which can then be used as a base for further work? If so, then this is something i'd like to be able to do...but maybe in the longer term. I think it's less essential for me at the moment.
Am i understanding what you mean correctly?
Cheers
S
fundjunkie is right on in his previous note. Now, to answer your question:
1. What you first described is general systems development - that is, test out a simple idea, see if it has merit, then develop into full system.
2. Scanning for specific behaviors is not something any of these programs can offer - you have to do the work. Neural Net programs claim to be able to do this, but I've never been convinced that they do.
1. What you first described is general systems development - that is, test out a simple idea, see if it has merit, then develop into full system.
2. Scanning for specific behaviors is not something any of these programs can offer - you have to do the work. Neural Net programs claim to be able to do this, but I've never been convinced that they do.
Quote from Synonym:
Does datamining mean having a program trawl data first to see what characteristics/behaviours exist, which can then be used as a base for further work? If so, then this is something i'd like to be able to do...but maybe in the longer term. I think it's less essential for me at the moment.
Data mining is a means of searching for patterns and relationships in data. A fitness function is used to highlight those parts of the data that are fit (of interest). Te term is most often used in conjunction with wide ranging searches numbering in the tens of thousands.
Neural networks are a type of data mining algorithm (regression) amongst many. But be clear data mining isn't just a case of setting a program loose on some data, sitting back and letting the good times roll with the results. it's way, way harder than that and the results and benefit when appled to the markets are debatable and far more etheral. Indeed, the consesus view is that neural networks, for example, are of no value in trying to model markets.
Without wishing to discourage you i think you'll be wasting your time by even considering that as an avenue to be looking down when you're only thinking about automation and have little to no coding background. But be mindful of it for different reasons. Data mining is a process. Data mining apps indistrialize and scale up that process. System oriented research is a process of mining on a smaller scale which brings it's own dangers of curve fitting and data mining bias (being fooled by the data). I can imagine you might produce dozens of versions of a trading concept before finding something that looks good. that is data minging...
So what you need to concern yourself with more (if results look good) is mitigating any data mining bias and avoiding curve fitting (methodological rigour).
Thx
D
p.s. I am not an expert either on data mining or data mining apps.
Quote from droskill:
fundjunkie is right on in his previous note. Now, to answer your question:
1. What you first described is general systems development - that is, test out a simple idea, see if it has merit, then develop into full system.
2. Scanning for specific behaviors is not something any of these programs can offer - you have to do the work. Neural Net programs claim to be able to do this, but I've never been convinced that they do.
Firstly let me say, that i wanted to reply to you and fundjunkie at simultaneously, but decided to do so seperately for the sake of clarifty.
Very sorry to both you and fundjunkie for any confusion i may have caused!
Okay. So, on point 1, I am right in assuming that both TB and TS support this kind of analysis arent i?
On point 2. Now we have this point clear between us (i was tired and obviously a little fuzzy in my thinking at the time i wrote the problem-causing post!), i can clarify my position; neural nets, etc are not something i'm interested in at this point in time. Like you i seriously doubt they are of value in trading. What i really mean to say was, longer term, if I have everything else in place and all is working well for me, then i'd be open to further investigation of that kind of thing, simply in an interest to see if there is anything worthwhile in it.
Cheers
Syn