Suggestions to Improve Automated Trading...
- Thread starter greaterreturn
- Start date
Everywhere I look strong ATS traders say they use ratedata.gaincapital.com to get free tick data. I use MBTrading as my broker.
I'm told there's minor issues with the data like having quotes on some files and not others. So you need to filter it and adjust the time to your zone, etc.
Otherwise, it's clean realistic usable tick data.
Does anyone think this free tick data has some other problem? Hard to beat free if it resembles the mbtrading data feed as I'm told from reliable sources.
Sincerely,
Wayne
I'm told there's minor issues with the data like having quotes on some files and not others. So you need to filter it and adjust the time to your zone, etc.
Otherwise, it's clean realistic usable tick data.
Does anyone think this free tick data has some other problem? Hard to beat free if it resembles the mbtrading data feed as I'm told from reliable sources.
Sincerely,
Wayne
The NinjaTrader strategy analyzer is very cool. I love it in general but it isn't optimized to work on tick data.
One annoying issue is that if you want to do optimization loops, it needlessly reloads the tick data before each loop.
During the 2 minutes it takes to process a months work of back test tick data, most of the time I can see from the task manager appears to be loading all the tick data.
Also, it only works on historical data that it can load into memory. So if you want to do many months you need many Gigs of ram on your computer.
The promise that in version 7 they will optimize for ticks and add caching so you can use it on infinitely large amounts of ticks without running out of memory, etc.
I'm looking forward to version 7.
In NinjaTrader (compared to NeoTicker) I find their C# architecture dramatically simpler, easier to understand and use immediately.
So I'm rooting for NinjaTrader to get more automated trading and tick data backtesting features since it was originally built for discretionary traders.
Hope this review was useful to any other ATS traders who use tick data.
Wayne
One annoying issue is that if you want to do optimization loops, it needlessly reloads the tick data before each loop.
During the 2 minutes it takes to process a months work of back test tick data, most of the time I can see from the task manager appears to be loading all the tick data.
Also, it only works on historical data that it can load into memory. So if you want to do many months you need many Gigs of ram on your computer.
The promise that in version 7 they will optimize for ticks and add caching so you can use it on infinitely large amounts of ticks without running out of memory, etc.
I'm looking forward to version 7.
In NinjaTrader (compared to NeoTicker) I find their C# architecture dramatically simpler, easier to understand and use immediately.
So I'm rooting for NinjaTrader to get more automated trading and tick data backtesting features since it was originally built for discretionary traders.
Hope this review was useful to any other ATS traders who use tick data.
Wayne
Bespoke, I have read many of Acrary posts and shocked by the simple truth he explains about diversifying themes, time frames, AND markets rather than struggle to pick one theme on a specific market.
Wow. Genius. I have wasted years fighting the trend vs. chop vs. counter trend paradox.
He says use them all AT THE SAME TIME.
Now that's genius, pure and simple and I can see that it will work.
I'm going to keep running this simple strategy to make sure all the kinks are out of the execution system.
At the same time, I'm downloading tick data and will start working on a portfolio of simple strategies in different themes and time frames.
Wow! It feels like a whole ocean of opportunities has just opened up.
That's the real power of automated trading. You can follow more markets, time frames, and strategies than every possible as a single human being.
Sincerely,
Wayne
Wow. Genius. I have wasted years fighting the trend vs. chop vs. counter trend paradox.
He says use them all AT THE SAME TIME.
Now that's genius, pure and simple and I can see that it will work.
I'm going to keep running this simple strategy to make sure all the kinks are out of the execution system.
At the same time, I'm downloading tick data and will start working on a portfolio of simple strategies in different themes and time frames.
Wow! It feels like a whole ocean of opportunities has just opened up.
That's the real power of automated trading. You can follow more markets, time frames, and strategies than every possible as a single human being.
Sincerely,
Wayne
Another retarded thread.
How are people supposed help you if you don't provide the information required to answer them? I am not a psychic or know nothing about your system and environment. You don't expect people to be trading the same system as you, or don't think that all systems work under the same "edge".
One simple question is, why does your test results and live trading results differ?
... what the hell is wrong with you, OP?
How are people supposed help you if you don't provide the information required to answer them? I am not a psychic or know nothing about your system and environment. You don't expect people to be trading the same system as you, or don't think that all systems work under the same "edge".
One simple question is, why does your test results and live trading results differ?
... what the hell is wrong with you, OP?
Hi, Bespoke.
I have just been interested in ATS. I have an algorithm in mind and have some programming skills.
I am here because I want to discuss the way with experts
First of all, I would like to know how did u get the data for stocks?
And for the data you get, it is tick by tick or bar format (O,H,L,C,V)? So, I can play with my data also and hope I can develop the winning strategy soon.
Thanks.
greaterreturn, I wrote my own charting software. It's pretty simple actually. Just a bunch of boxes and lines. I don't use technical indicators so I don't have to draw any squiggly lines.
aonelite,
I have both types of data and use them for different strategies. But IMHO, you won't find much in OHLCV. 95% of my trading now was never backtested nor can it be (maybe some of it can be I suppose). I just test it out live in sim-mode first to make sure the concept/code is done correctly but that usually isn't enough to know the true results so I usually just go at it with 100 shares long enough to know if it actually works or doesn't. Most of the time it doesn't! Majority of my trading is just in out as quickly as possible and that is difficult if not impossible to accurately backtest with any type of data.
I got my data from opentick.com. I don't think they're accepting any new members at the moment. It's free and quality is decent for recent years though you'll have to do a lot of massaging.
aonelite,
I have both types of data and use them for different strategies. But IMHO, you won't find much in OHLCV. 95% of my trading now was never backtested nor can it be (maybe some of it can be I suppose). I just test it out live in sim-mode first to make sure the concept/code is done correctly but that usually isn't enough to know the true results so I usually just go at it with 100 shares long enough to know if it actually works or doesn't. Most of the time it doesn't! Majority of my trading is just in out as quickly as possible and that is difficult if not impossible to accurately backtest with any type of data.
I got my data from opentick.com. I don't think they're accepting any new members at the moment. It's free and quality is decent for recent years though you'll have to do a lot of massaging.
bespoke,
It would seem like if you get tick data that with every trade, bid, and ask you could test your scalping strategy, right?
At least that's my plan. I have been using the NinjaPlatform to test using tick data and have been impressed at how accurately it predicts the true behavior when running live.
Only problem is that NinjaTrader can't handle more than a month or two of tick data.
So I'm building my own back testor that uses tick data so it can test and optimize on many years of tick data across multiple markets simultaneously.
Wayne
It would seem like if you get tick data that with every trade, bid, and ask you could test your scalping strategy, right?
At least that's my plan. I have been using the NinjaPlatform to test using tick data and have been impressed at how accurately it predicts the true behavior when running live.
Only problem is that NinjaTrader can't handle more than a month or two of tick data.
So I'm building my own back testor that uses tick data so it can test and optimize on many years of tick data across multiple markets simultaneously.
Wayne
Hello greaterreturn,
among all the approches I have tried and discarded (several dozen) because non profitable, I currently believe the best approach is one purely probabilistic.
All the other approaches based on tech analysis are of no real effectiveness. They work sometimes, but just by chance, until they quit working.
Or they work very well in extreme conditions, but then the strategy is not suitable really for automation, being idle for long time intervals.
Especially in daytrading there is no point relying on backtests, even if run over 10 or more years.
I have ended up pursuing a probabilistic approach (geometric brownian motion and so on), which is, by definition itself, "robust" wrt market changes, and perhaps you also will arrive to conclude that it's the second best way of making profits, ... the first one being not to trade) )
Tommaso
>Hopefully, if you provide some useful tip or point me to a great thready I can return the favor somehow.
Sincere,
greaterreturn
among all the approches I have tried and discarded (several dozen) because non profitable, I currently believe the best approach is one purely probabilistic.
All the other approaches based on tech analysis are of no real effectiveness. They work sometimes, but just by chance, until they quit working.
Or they work very well in extreme conditions, but then the strategy is not suitable really for automation, being idle for long time intervals.
Especially in daytrading there is no point relying on backtests, even if run over 10 or more years.
I have ended up pursuing a probabilistic approach (geometric brownian motion and so on), which is, by definition itself, "robust" wrt market changes, and perhaps you also will arrive to conclude that it's the second best way of making profits, ... the first one being not to trade
) ) Tommaso
>Hopefully, if you provide some useful tip or point me to a great thready I can return the favor somehow.
Sincere,
greaterreturn