I need a hug

[rnicoll]

My auto-trading strategy is losing money all over the place. It'll pick itself up, laugh off the loss and make it all back again as soon as the markets stabilise, I know, but yikes it's hard to not get depressed.

Sorry, felt a need to rant, figured this was the best place to do it

 

[Crap Buddist]

*Hug* yes, but also recall that you are a part of that strategy that thinks. So why not examine the losing trades to see if you can evolve the strat to be more reactive to market actions.

Or What about running a counter trend strat alongside a trend catching strat, rather than a catch all in one strat, that would smoothen the curve ?

 

[Mr.J-Arthur]

Why auto-trade?
Sorry, always against that, unless it's truely working well for you. Why autotrade if you're still going to feel the pain? Seems contradicting.

Here's your cyber *hug* anyways.

 

[Masquerade]

Sorry to hear - i'd been following your thread on it. I guess you'll just have to put it on hold as not many automated trading systems are going to be able to deal with this volatility. Just put it on hold and wait for the market to stabilise and get some predictability to it. In the mean time, why not trade "properly"

 

[rnicoll]

*Hug* yes, but also recall that you are a part of that strategy that thinks. So why not examine the losing trades to see if you can evolve the strat to be more reactive to market actions.

Or What about running a counter trend strat alongside a trend catching strat, rather than a catch all in one strat, that would smoothen the curve ?

Mostly I'm just finding it quite hard to find exclusion criteria. My best idea so far was to have it paper trade after a losing trade, but generally the worst of the damage is done before the first losing trade is finished. I might have it block entering more than 2 or so trades if both are already losing, keep a brake on things.

I'm currently soothing myself by backtesting on NinjaTrader, which shows some comfortingly positive numbers for most of the year, and some nasty red figures for this week. Suggests strongly it's just bad luck so far.

Counter-strat; definitely, although trying to pick out when to switch is also tricky. Once I've got more stats, possibly looking for patterns in how it goes wrong (it seems to go in long strings of profit/loss when working, and flick between the two when broken). Realistically, I probably should also take a step back, and start again from a less weird strategy (I'm basically trading statistical artefacts of the data, rather than anything more normal such as fundamentals hinted at by the technicals); even if it makes 1/10th the profit, that's still a good start.

 

[rnicoll]

Why auto-trade?
Sorry, always against that, unless it's truely working well for you. Why autotrade if you're still going to feel the pain? Seems contradicting.

Here's your cyber *hug* anyways.

Thanks for the hug

Auto-trading because basically it plays to my strengths. Generally hate being this arrogant, but essentially I'm an excellent programmer, excelling in a variety of areas well suited to stock trading (distributed systems are my speciality). On the other hand, I'm also a nervous twitchy wreck, which makes me rather unsuited to trading by hand. Also... essentially, if I can see something to trade, I can quantify it and code it, my problem is in finding a reliable trading system.

My attempts to trade this system by hand... well, lets just say the computer is considerably ahead, shall we?

 

[rnicoll]

Sorry to hear - i'd been following your thread on it. I guess you'll just have to put it on hold as not many automated trading systems are going to be able to deal with this volatility. Just put it on hold and wait for the market to stabilise and get some predictability to it. In the mean time, why not trade "properly"

Well, I'll keep it paper-trading at least, might as well collect the data for analysis. Proper trading... because the computer's going to do as good a job as I can, essentially. Well. Improving on bid/ask is proving tricky, and is exceptionally hard to backtest, but mostly...

Good to know I'm being interesting, though

 

[Mr.J-Arthur]

Haha, well you could overcome the nervous bits, then it becomes a rush to do it correctly.

I've never programmed a system, but I hear they can be over "optimized" to previous data.
Good luck! market was choppyyyyy today.

 

[rnicoll]

Haha, well you could overcome the nervous bits, then it becomes a rush to do it correctly.

I've never programmed a system, but I hear they can be over "optimized" to previous data.
Good luck! market was choppyyyyy today.

Testing for over-optimisation is what this week has been all about. That said, the live version isn't behaving all that much like the same strategy backtested over the same period, which is more than a little weird. That goes on my things to resolve list.

Funny thing is, it turns out I was calculating profit insanely on Wednesday, using pips instead of percent. Given it was across a basket of currency pairs with percent based stop-losses, this meant the problems with GBP/USD were trashing the system. Fixing that meant it was more or less breaking even on Wednesday. Current live version still loses over the week, overall, but an improved version with stake management manages to make a small profit (1% before margin).

Obviously, the stake management may be optimised to this week, though, so that needs testing too. Could be here a while!