1% a day consistently: possible?

[Thunderdog]

Quote from stephencrowley:

...I was just trying to gather input from other experienced automated traders whether 1% a day was within the realm of possibility (not possible as in "anything is possible" but actually possible).

Fair enough. But when it comes to trading, may I suggest that you try not to let yourself be either limited or misdirected by the thoughts and beliefs of others? Especially when you don't know the validity of the source. Against this background, however, be advised that you are looking to immediately place yourself in the top 0.1% of traders with the performance numbers you're talking about. Just something to think about.

P.S. I just made up that 0.1% number, but I'm guessing it's probably optimistic.

 

[OddTrader]

How do you know when to re-activate the system after encountering a stop-loss?

 

[stephencrowley]

After stop-loss I won't trade for the rest of the day. I'll pull in all the recorded data for that day (3GB/day usually) and take a peak at the model to see how/why it broke down. It shouldn't happen that often in theory as I re-estimate all parameters over a rolling 1-week period.

Quote from OddTrader:

How do you know when to re-activate the system after encountering a stop-loss?

 

[PitchBlack]

Quote from stephencrowley:

After stop-loss I won't trade for the rest of the day. I'll pull in all the recorded data for that day (3GB/day usually) and take a peak at the model to see how/why it broke down. It shouldn't happen that often in theory as I re-estimate all parameters over a rolling 1-week period.

You NEVER stop trading after a stop-loss, because a stop-loss is part of the trading system. You have to take ALL the signals that are generated by your system. If you don't do that the testing that you did before starting to trade has no statistical value at all.
Maybe you miss the biggest move of the century after your stop-loss was hit.

Many traders develop and backtest a system and take the results of the tests as the picture of what will happen in real trading. But once they start trading they don't trade as they did in their simulation. So the real results are completely different from the testing results. Most of the timles results are worst because of the different behaviour in real trading.
You never know which trade will be good or bad, otherwise you would only take the good ones. Therefore you have to take all the trades.

 

[stephencrowley]

Perhaps that is true of directional trades, but mine don't fall into that category. My stops are designed such that if they are ever hit then it means something fundamental has changed in the things I'm trading and I should re-evaluate the model before resuming.

Also, limiting drawdown and being consistent is more important than catching massive swings. The things I trade barely move at all but they move consistently.

Quote from spike500:

You NEVER stop trading after a stop-loss, because a stop-loss is part of the trading system. You have to take ALL the signals that are generated by your system. If you don't do that the testing that you did before starting to trade has no statistical value at all.
Maybe you miss the biggest move of the century after your stop-loss was hit.

Many traders develop and backtest a system and take the results of the tests as the picture of what will happen in real trading. But once they start trading they don't trade as they did in their simulation. So the real results are completely different from the testing results. Most of the timles results are worst because of the different behaviour in real trading.
You never know which trade will be good or bad, otherwise you would only take the good ones. Therefore you have to take all the trades.

 

[OddTrader]

Quote from stephencrowley:

My drawdown has been practically 0 (less than 0.5%) in testing, even in out-of-sample data. I can't stand drawndown.. every trade I make is 'optimally' hedged and has very strict stop losses. My stop loss has never been hit at testing and its only set at 1% drawdown from entry. If it is hit then the system will pretty shut it self down so it'll only incur a full 1% once which wont wipe out the many profitible trades before it. That'll give me time to re-examine the relationships it was trading.. etc, hopefully that'll never happen either as my models are constantly updating to new relationships.

I guess many traders usually would not think a system like this a FULLY (developed) automated system.

Quote from stephencrowley:

General question for intraday automated traders.. no manual intervention except in extreme cases.

???

 

[stephencrowley]

Why would it not be 'developed'? Also, why the confusion? Manual intervention can only screw things up if you start to see patterns in nothing. Our brains are wired to see things that aren't there, reasoning, logic and stats can help you decide when if something is real or not.

Quote from OddTrader:

I guess many traders usually would not think a system like this a FULLY (developed) automated system.




???

 

[OddTrader]

Quote from stephencrowley:

After stop-loss I won't trade for the rest of the day. I'll pull in all the recorded data for that day (3GB/day usually) and take a peak at the model to see how/why it broke down.

Since your backtesting has not encountered any stop-loss yet, how do you know (what's the underlying theory?) there will not have consecutive losses after the very first one?

Quote from stephencrowley:

It shouldn't happen that often in theory as I re-estimate all parameters over a rolling 1-week period.

If it happens, how often?

 

[stephencrowley]

I can't give away the underlying theory without giving away my strategy.

Basically, I won't get stopped out many times in a row because the 1st time I'm stopped out the model will be reestimated with data from the day that I got stopped out, and also with data from when the model worked fine. The models output will be consistent with all the data it observes and if it isn't stable across days then it will tell me so and I'll have to find something else to trade.

Quote from OddTrader:

Since your backtesting has not encountered any stop-loss yet, how do you know (what's the underlying theory?) there will not have consecutive losses after the very first oen?




If it happens, how often?

 

[OddTrader]

Quote from stephencrowley:

My stops are designed such that if they are ever hit then it means something fundamental has changed in the things I'm trading and I should re-evaluate the model before resuming.

How much lead time (Hours, days or months? ) do you need to modify the system due to this kind of fundamental changes?

How many times did you modify your (Dynamic/ Constantly Updated) system during your backtesting?