Systematic trading | Page 14 | Forex Factory

Options

Go to
Permalinks
Print Thread

Similar Threads

Trade another day through Systematic Trading 203 replies

Nonparametric Association Measures in Systematic Trading 57 replies

Anyone use Systematic from Alpari? 1 reply

A Systematic Approach to Markets 68 replies

Systematic Trading vs. Discretionary Trading 14 replies

Attachments: Systematic trading
Exit Attachments

Systematic trading

  • #261
  • Aug 15, 2009 5:44am Aug 15, 2009 5:44am
Quoting triphop
Mikkom, carrying over from another thread, you've said you found fixed %R the most profitable in your strategies. It makes market sense to me because with a fixed amount per pip you inadvertently tilt your strategy to favour days of higher volatility, which isn't necessarily a good thing.
But wouldn't weighted position sizing surely improve a strategy if you could determine which conditions had a higher probability of success than another? This is obviously very strategy dependent - personally, I've seen some weak correlations between certain...
In fact I do that. I always use risk percentage per trade but I vary the percentage based on some rules. I didn't mean fixed risk, I meant always using risk per trade (I'm not always very clear I know )

It's also important to understand that basic r% with fixed risk for every trade and volatility based stops means that you are basically using bigger position sizes with lower volatility. I'm quite sure this is one of the reasons why breakouts tend to work so well after periods of low volatility (other obvious reason is the cumulation of stops).
  • #262
  • Aug 15, 2009 5:47am Aug 15, 2009 5:47am
Quoting mikkom
Here are the results for 4/1 week donchian breakout since 2001 if anyone is interested (based on minute data) 2% risk per trade.

The framework is still quite rough as you can see but it works.
Looks a lot like my desktop...
The breaking of a wave cannot explain the whole sea.
  • #263
  • Aug 15, 2009 6:17am Aug 15, 2009 6:17am
Thanks M, obvious now you say it, not so obvious 5 minutes ago!
Isn't there a risk of over-leveraging your edge though? If you had a good breakout strat that you backtested, say you found the longer the period of low volatility, the bigger the breakout as an example.
By factoring this into position sizing, you're effectively leveraging that edge even more, meaning that any drawdowns will be proportionately much larger.

Unless your rules of position sizing are separate from the edge that defines your strategy? Does that make sense?
  • #264
  • Aug 15, 2009 6:34am Aug 15, 2009 6:34am
Quoting triphop
Thanks M, obvious now you say it, not so obvious 5 minutes ago!
Isn't there a risk of over-leveraging your edge though? If you had a good breakout strat that you backtested, say you found the longer the period of low volatility, the bigger the breakout as an example.
By factoring this into position sizing, you're effectively leveraging that edge even more, meaning that any drawdowns will be proportionately much larger.

Unless your rules of position sizing are separate from the edge that defines your strategy? Does that make sense?
Yes they are totally separate. I have a base risk for each strategy and adjust the risk with the same rules for all of my strategies.
  • #265
  • Edited 6:56am Aug 15, 2009 6:38am | Edited 6:56am
For anyone who wants to see how stops and orders cumulate, these are 2 great tools.

http://fxlabs.oanda.com/products/his...erhistory.html

http://fxlabs.oanda.com/products/his...onhistory.html
  • #266
  • Aug 15, 2009 7:29am Aug 15, 2009 7:29am
Quoting triphop
Thanks M, obvious now you say it, not so obvious 5 minutes ago!
Isn't there a risk of over-leveraging your edge though? If you had a good breakout strat that you backtested, say you found the longer the period of low volatility, the bigger the breakout as an example.
By factoring this into position sizing, you're effectively leveraging that edge even more, meaning that any drawdowns will be proportionately much larger.

Unless your rules of position sizing are separate from the edge that defines your strategy? Does that make sense?
Interestingly enough I was experimenting with this the other day, and it can work well... I have a very simple system based on a daily straddle of GBPUSD at London market open, with stop size adjusted based on previous daily ranges. I was not assigning a higher probability to any of the two sides, long or short, and just risked 1% in each trade (they are OCO orders).

At some point I decided to introduce a different weight for each one of the sides depending on the perceived probability of execution of that side, since after all we may be in the middle of a trend (I just take a look at the slope of a Hull MA computed with the previous days, but of course there is a ton of different ways of doing this). Basically, I keep the sum of the risks constant (2%), but depending on that slope an individual order can have an assigned risk of between .1% and 1.9% , and the other one the complementary risk so that the sum is 2%. This is just a somewhat arbitrary rule

With these parameters, backtesting (8 years) says the system would have made more money without increasing the drawdown. Not a brutal amount of money, but significant. The most interesting feature is the drawdown, of course... The clear daily trend greatly increases the probability of price going in your favor instead of hitting the stop. That is at least my interpretation of why I don't see a larger drawdown: I am overleveraging only trades that look good from the start based on the most recent market behavior.

I also find interesting that this weighting of the risk affects mostly trades that occur in periods of high volatility (a clear daily trend), and in that sense it goes a little against the general behavior of these systems that, as mikkom said, tend to make money after periods of very low volatility (smaller stops).
  • #267
  • Aug 15, 2009 7:49am Aug 15, 2009 7:49am
Quoting PipAdder
Interestingly enough I was experimenting with this the other day, and it can work well... I have a very simple system based on a daily straddle of GBPUSD at London market open, with stop size adjusted based on previous daily ranges. I was not assigning a higher probability to any of the two sides, long or short, and just risked 1% in each trade (they are OCO orders).
Interesting Pipadder, when I get a bit more time I'll do a similar test on my strat and see how that compares.
  • #268
  • Aug 15, 2009 2:05pm Aug 15, 2009 2:05pm
Quoting mikkom
One thing that has helped me the most is the realization that systems [at least any of that I have tested] rarely work under daily bars, on daily bars it's almost easy to find systems when you have couple of systems that work and know why they work as they do. I'm planning to test mostly daily systems with my framework too, the minute data is there mostly to make sure that stops/tp/sl's are hit in correct order.

I am coming to the same conclusion. This would then also mean it is not worth while to collect tick data as I do on the moment and just stick to 1 min bars. I still find it amazing that it is so difficult (if possible at all) to find systems in the lower time frames.
  • #269
  • Aug 16, 2009 5:56am Aug 16, 2009 5:56am
Quoting spekkiefx
I am coming to the same conclusion. This would then also mean it is not worth while to collect tick data as I do on the moment and just stick to 1 min bars. I still find it amazing that it is so difficult (if possible at all) to find systems in the lower time frames.
It is definitely possible. But IMO the lower you go the more you need to look into things like

- Market micro structure
- Trade cost analysis
- Optimizing execution models (mostly separate from the trading models)
- Accurate market simulation

Obviously the lower you go the more you need to care about market frictions.
  • #270
  • Aug 16, 2009 6:43am Aug 16, 2009 6:43am
Quoting BlackMage
It is definitely possible. But IMO the lower you go the more you need to look into things like

- Market micro structure
- Trade cost analysis
- Optimizing execution models (mostly separate from the trading models)
- Accurate market simulation

Obviously the lower you go the more you need to care about market frictions.
Black Mage, could you tell us what your approach to the market?
Mikkom, uses systems on the higher frames hence testing over many years. Craig is concentrating on statistical arbitrage. I am attempting to master the intraday lower time frame. Is there anyone amongst us that has succeeded on the lower frames?
  • #271
  • Aug 16, 2009 1:42pm Aug 16, 2009 1:42pm
Quoting spekkiefx
Craig is concentrating on statistical arbitrage.
I wouldn't say I concentrated on it, it's just one of the things I do.
In general, I think I do roughly the same thing as Mikkom, just spread over more markets and time-frames. It is not too hard to find things that work on shorter time frames, you just aren't going to find things which work for years on end, if something works for a couple of months, fine by me, I'll use it while it works, if longer great. BlackMage is correct however, in that your simulations must be more accurate, this is much easier to do in real-time, than in retrospect.
The breaking of a wave cannot explain the whole sea.
  • #272
  • Aug 17, 2009 1:02pm Aug 17, 2009 1:02pm
What are your thoughts on tick volume? I know tick=quotes <> actual trades, and all the downsides, but take a look at this. I've posted this elsewhere.

Fairly simple strategy, trading every 30 minutes, 6 months, 4000 trades ish. Using only tick volume and price. The following is without spreads, taken using the midpoint of bid/ask on an aggregate a0 esignal feed, so it's completely unrealistic return for someone on this side of the fence, but the fact the correlation is so strong makes me wonder about how market makers must price shade.

Question is, can this be translated into anything tradeable?
Attached Image
  • #273
  • Aug 17, 2009 1:57pm Aug 17, 2009 1:57pm
Quoting triphop
What are your thoughts on tick volume? I know tick=quotes <> actual trades, and all the downsides, but take a look at this. I've posted this elsewhere.

Fairly simple strategy, trading every 30 minutes, 6 months, 4000 trades ish....
It's quite hard to say anything about specific strategy without knowing more about the strategy. Overall I think tick volume tells something (frequency of trades) but not the same thing as volume (combined size of trades).

How the simulation looks when spread is added? The shorter the period/smaller the profit, more the spread matters.
  • #274
  • Aug 17, 2009 2:31pm Aug 17, 2009 2:31pm
As a strategy - it enters on high volatility, high volume, exits on low volume. My theory - and it's just that - is that given that big orders tend to be split into smaller parts (icebergs et al), that tick volume could tell something about the overall volume. I've not seen any proper research into tick vols - just plenty of assumptions from both camps.

Factor in the spread, and the strategy loses - so no, I'm under no illusions that this is tradeable at all as it stands. It trades at far too smaller TF. Correlation though is interesting to me; far more than random.
  • #275
  • Aug 17, 2009 3:32pm Aug 17, 2009 3:32pm
Quoting triphop
As a strategy - it enters on high volatility, high volume, exits on low volume. My theory - and it's just that - is that given that big orders tend to be split into smaller parts (icebergs et al), that tick volume could tell something about the overall volume. I've not seen any proper research into tick vols - just plenty of assumptions from both camps.
I haven't done any studies but wouldn't it really be quite easy to compare futures volume and tick volume and see what the correlation is?
  • #276
  • Aug 17, 2009 4:26pm Aug 17, 2009 4:26pm
It would (though I haven't), but that's resting on the assumption that futures volumes are helpful to trade full stop. Rather than going through the middle man in this, are tick volumes helpful to trade spot? The above doesn't prove it of course, but I thought it was worth sharing as it suggests they might.
I found them useful in discretionary trading, but only in market profile and then in certain conditions.
  • #277
  • Aug 17, 2009 4:38pm Aug 17, 2009 4:38pm
Quoting spekkiefx
Black Mage, could you tell us what your approach to the market?
My approach is somewhat different from most here I guess, since I am solely focusing on "high frequency" trading systems. These systems relies heavily on the particular market microstructure of spot fx and efficient execution strategies. We are working hard on developing good macro models though, and uniting these macro models with the micro structure and execution models is the ultimate goal.

"High frequency" means different to different people, but to me it is a system that processes and reacts to data on the highest possible frequencies. It does not necessarily means that the system has to execute many trades (though it often does). A trade can be open from less than 1 millisecond to several hours but is always closed at "end of day" (normally when NY session ends).
  • #278
  • Aug 17, 2009 4:55pm Aug 17, 2009 4:55pm
Quoting triphop
Rather than going through the middle man in this, are tick volumes helpful to trade spot?
Go check out GammaJammer post on this:

http://www.trade2win.com/boards/fore...tml#post345757

He's pretty clear on why using tick volume as a proxy for transaction volume is not really logical.

EDIT: Not saying that tick volume/frequency can't be useful. But I'm with GJ when it comes to using it as proxy for real volume.
  • #279
  • Aug 17, 2009 4:56pm Aug 17, 2009 4:56pm
Quoting BlackMage
My approach is somewhat different from most here I guess, since I am solely focusing on "high frequency" trading systems. These systems relies heavily on the particular market microstructure of spot fx and efficient execution strategies. We are working hard on developing good macro models though, and uniting these macro models with the micro structure and execution models is the ultimate goal.
Interestingly I plan to do the same thing in opposite order. I have some models that work on longer term data so I'm going to use them as a guide in shorter term periods - I'm especially interested in entry strategies that in itself have an edge, I assume it's quite the same thing as your high freq systems.
  • #280
  • Aug 17, 2009 5:04pm Aug 17, 2009 5:04pm
Quoting BlackMage
Go check out GammaJammer post on this:

http://www.trade2win.com/boards/fore...tml#post345757

He's pretty clear on why using tick volume as a proxy for transaction volume is not really logical.
Agh, the irony! It was my thread on the same forum (look a few threads down) with the very same man that convinced me to look at it more closely. He still isn't an advocate mind but conceded that it might, possibly have something in it.
1 trader viewing now