thanks!
Frosty's auto-trading bot goes live with REAL money
- Thread starter frostengine
- Start date
Let's see if I understand this correctly. If the 360-minute EMA is going up and the 180-minute EMA is going down, you buy. If the 360-minute EMA is going down and the 180-minute EMA is going up, you sell short. So, you long the short-term pullbacks from the up trend, and short the "up" bounces from the down trend. Something like that?
You didn't specify how you establish the "trend", so the length of the EMAs are my guess (although an educated one, since I researched these types of strategies, too). Or perhaps you are not using EMAs at all. I'd also guess the system is always in the market, either long or short? Another thing that's not clear is when you decide to enter. I mean, how long or far do you allow the shorter term trend to deviate from the long term trend before the signal is generated?
nonlinear,
Yes, you are right in some of your assumptions. However, EMA's are not what I use. The concept is the same, the way the trends are identified is very important but not imporatnt to the overall discussion.....
The bot is not always in the market, although it generally is..... there are times when it is not.. if the adx drops below <15 it closes out is current position. So in those times it will not be in the market at all.. other than that its always in once it has entered an intial position....
I don't really have a way of describing how far I allow the smaller trend to deviate before a signal is generated.... however it is a rather big deviation... Them simply moving in opposite directions is not enough
Yes, you are right in some of your assumptions. However, EMA's are not what I use. The concept is the same, the way the trends are identified is very important but not imporatnt to the overall discussion.....
The bot is not always in the market, although it generally is..... there are times when it is not.. if the adx drops below <15 it closes out is current position. So in those times it will not be in the market at all.. other than that its always in once it has entered an intial position....
I don't really have a way of describing how far I allow the smaller trend to deviate before a signal is generated.... however it is a rather big deviation... Them simply moving in opposite directions is not enough
So, there is some threshold that specifies the minimum amount of divergence between the direction of the longer and the shorter term trends?
Look, I don't want to make it a torture to pull information from you. If this journal is just that, to simply record your daily P&L, that's fine. But for a meaningfull discussion, it would help to know the details of the strategy. Your risk is that it's a 1-billion a year strategy that you will disclose to the world. Your reward is that it might be a totally hopeless strategy (once tested rigorously), so we may save you some dough by making a case for it.
nonlinear,
I will never give absolute specifics. But yes, I have a threshold that the divergence must reach...... this threshold is only reached a couple times a day.. its not very common. Normally before the threshold is reached the longer term trend comes into sync with the shorter one.. but not always...
Over time I may devulge more information. But for conversational purposes I don't think specifics have to specfied... I think the strategy can be taken as a general concept... a class of strategies based on this general idea.. the indicators you use to determine the trends and the threshold is a personal choice... Doesn't make the general concept any more or less valid
I will never give absolute specifics. But yes, I have a threshold that the divergence must reach...... this threshold is only reached a couple times a day.. its not very common. Normally before the threshold is reached the longer term trend comes into sync with the shorter one.. but not always...
Over time I may devulge more information. But for conversational purposes I don't think specifics have to specfied... I think the strategy can be taken as a general concept... a class of strategies based on this general idea.. the indicators you use to determine the trends and the threshold is a personal choice... Doesn't make the general concept any more or less valid
frosty -
This is one torturous thread.
You obviously want and need help in producing a stable ATS but you are quite unwilling to provide any details that would allow seasoned traders to provide that help. You must recognize, if you want to be a trader, that your strategy has a very high probability of not being profitable. I have backtested hundreds if not thousands of combinations of indicators on portfolios of stocks; historical data going back over twenty years, dividends, spinoffs, mergers, bankruptcies, management, execution strategies, slippage, etc.... You get the idea I am sure.
What's the point? I can count on one hand the number of strategies that I have entrusted my equity and actually implemented.
From the discussion on this board it is obvious you have not thoroughly analyzed your strategy. My advice is shut your system off, and reevaluate how you tested your strategy.
Baby steps frosty, baby steps
This is one torturous thread.
You obviously want and need help in producing a stable ATS but you are quite unwilling to provide any details that would allow seasoned traders to provide that help. You must recognize, if you want to be a trader, that your strategy has a very high probability of not being profitable. I have backtested hundreds if not thousands of combinations of indicators on portfolios of stocks; historical data going back over twenty years, dividends, spinoffs, mergers, bankruptcies, management, execution strategies, slippage, etc.... You get the idea I am sure.
What's the point? I can count on one hand the number of strategies that I have entrusted my equity and actually implemented.
From the discussion on this board it is obvious you have not thoroughly analyzed your strategy. My advice is shut your system off, and reevaluate how you tested your strategy.
Baby steps frosty, baby steps
OK, it's your blog. But I suspect you may be left with just the cheerleaders around here, and the content of this thread would be reduced to "what a day!" and "don't give up, drawdowns are normal". A soup opera of sorts, as someone already noted.
id disagree with that. someone does not need to give up their proprietary trading system and edge to public domain to have helpful discussions about fundamental trading knowledge and ats design. i personally have learnt heaps and have had many ideas sparked by this thread long before frost even began hinting at his methodology.
i may be a cheerleader. but i definitely do not need to know the specifics to offer suggestions on how I would personally approach debugging and confronting specific issues shown by live trading. i dont think frost is looking for anyone to tweak his indicators for him. but that does not prevent us from offering ideas or experiences which may help him add or modify it himself.
and yes. drawdowns are normal. its something that has to be repeated over and over so that we dont lose faith in our empirical results from backtesting.
i just had an interesting while typing that paragraph (again one of the values of this thread). frost with your 3 month equity curve that you've posted a few times. write a quick algorithm which looks at every recorded day, and look at the relative profit 1 week after. then figure out the ratios of days that would be profitable after a week / days that would be at a loss after a week. then do the same for two week periods. then three. etc.
judging your equity curve by eye. id say if you picked a random day to start, there would be a 50% chance that you would be profitable after a week. maybe 60% chance that you would be profitable after two weeks. and so on. this should put your current performance in perspective and let you know how lucky / unlucky you've been with the random day you picked to start compared to what you would expect statistically.
you might find that your current performance is not that unlikely (maybe 10-20% chance of being at a loss after 4 weeks) . or you might find you have been extremely unlucky and not find a single example of a starting day that would be at a loss after a month of running.
id probably print out all these probabilities before i go live with a directional trading system. that way I would know what to expect as far as a realistic chance of being at a loss after X weeks. let me know what you think of the idea.