You are using an out of date browser. It may not display this or other websites correctly.
You should upgrade or use an alternative browser.
You should upgrade or use an alternative browser.
Frosty's auto-trading bot goes live with REAL money
- Thread starter frostengine
- Start date
now this is the first time i would have a real concern.Quote from frostengine:
... this is the BIGGEST down day in all of my 4 years of data....
...
maybe you can do something in order to get back to an objective
view on the subject: calculate the likelihood of observing such a
day in the first weeks of trading. first i would make a distribution
of daily returns, than add some additional tail to it - just to cover
some higher losses that could be, yet never happened so far. now
calculate with which likelihood you experience what you currently
experience. and hold this likelihood against the possibility that
your system is actually a statistical fluke and not tradeable.
then name a trigger and pull when your likelihood falls below.
do not get tempted by things like "that happened before", it is
the likelihhod of you experiencing it once you are really trading
which counts.
most important, even more than losing or winning money right
now, is that you make it back onto the driving seat.
frosty,
Here' some advice from somebody who has been developing and trading systems for years.... After an outlier like that you need to shutdown your system and analyse what went wrong.
Don't take it too hard. You will only truly begin to understand how to develop and test systems properly as a result of learning from your mistakes. Everybody starts with an experience like yours. You can learn from it and move on or you can blow-up your account.
Find someone who knows how to develop a system review every step of your development and testing process. Learn some basic stats (distributions etc). Study the concept of R and expectancy. Learn how to do some simple monte carlos.
stocksniper's advice was excellent and deservant of re-reading .......
Gut it out frotsy. That which doesn't kill you makes you stronger.
All the best,
bolter
Here' some advice from somebody who has been developing and trading systems for years.... After an outlier like that you need to shutdown your system and analyse what went wrong.
Don't take it too hard. You will only truly begin to understand how to develop and test systems properly as a result of learning from your mistakes. Everybody starts with an experience like yours. You can learn from it and move on or you can blow-up your account.
Find someone who knows how to develop a system review every step of your development and testing process. Learn some basic stats (distributions etc). Study the concept of R and expectancy. Learn how to do some simple monte carlos.
stocksniper's advice was excellent and deservant of re-reading .......
changing systems or contract in the middle of the action is equivalent to quitting and starting a new system. From experience developping ATS (just over 3 years, not much but still I think I can give you some feedback) I can tell you that second guessing your system will end up costing you a lot.
You should have PREDEFINED criteria for:
- when to change the trading contract, stock,...
- when to move from system1 to system2 to system3 to ...
- when to stop completely...
The whole game should be PRE-planned, PRE-defined, and preferably coded before you do anything.
Gut it out frotsy. That which doesn't kill you makes you stronger.
All the best,
bolter
Frosty, sorry about your bad day.
It looks like you got some good advice here on the statistical level, but I am more puzzled by the action itself since it seems that your system is a trend reversal system, and yesterday thats pretty much all the mkt did. It opened flat, traded down, reversed up to unch, traded up, and finally reversed back down to unch. What went wrong? How does yesterday compare to other similar days?
It looks like you got some good advice here on the statistical level, but I am more puzzled by the action itself since it seems that your system is a trend reversal system, and yesterday thats pretty much all the mkt did. It opened flat, traded down, reversed up to unch, traded up, and finally reversed back down to unch. What went wrong? How does yesterday compare to other similar days?
If every automated system got shut down because it hit a maximum historical drawdown day. Very few would be running right now. It has to happen sometime. and I can assure you that there will be a worse day sometime in the future. Let say you have 100 yrs worth of data from 2000 to 2100. Logically there is a 4%~ chance that the max drawdown exists in the first 4 years assuming equal probability from day to day.
frost,
End-of-months always do weird things to the markets. I either have really good days or really bad days on the last day w/ my system, so don't sweat it.
Is it possible to run your bot in *both* papertrading and live mode simultaneously? It would be interesting to see if you get the exact same results, or if you get different slippage.
End-of-months always do weird things to the markets. I either have really good days or really bad days on the last day w/ my system, so don't sweat it.
Is it possible to run your bot in *both* papertrading and live mode simultaneously? It would be interesting to see if you get the exact same results, or if you get different slippage.
A lot of messages here to reply to, so I'll try to give a small summary. Concerning having predefined times when to quit, change contracts etc.... in hindsight that looks like it would have been a good idea to do. Unfortunately I did no such thing. I did however, have a rough estimate of how low I would let my equity go before pulling the plug. When reached my plan was to pull the plug, go back to sim trading, and wait until I had another solid strategy to try and that way I woudl still have enough equity to give it a shot.
With that said the bot has NOT hit that area yet, so I will not pull the plug entirey let, but on the same token I will pull it if it gets there because I need enough equity to start a new strategy in the future if needed.
Yesterday I went into crazy testing mode, and was really looking in depth not at how my strategy performs, but more about WHY my strategy performs... what edge is it really exploiting..... I especially looked at the second strategy in regards to the first.. they are basically the same strategy but one uses some market internals to help decide when not to trade. What I found is the second strategy's filter makes VERY good sense for this type of strategy...... However, looking at it the way I am using the filter may not be the BEST way... It opened my eyes up to some new ideas on improvements, but thus far nothing good has come from it, but the wheels are turning....
I have decided today to run the second strategy. I believe this strategy is MORE sound. I may not be able to test this strategy as far back as the first due to the amount of data I have... but the principles its built on makes me feel confident that it is a more sound strategy. If nothing else comes out of switching at least the strategy will trade a lot less for the next few days which will give me some time to continue exploring some ideas I had.....
Concerning if I can run the live mode and sim mode at same time.. the answer is yes... I have not done it many times... especially not in this mode of live testing, but back when I had my old auto trader from several months ago I did it a few times to see how reliable the sim mode was and the results were very similar.
With that said the bot has NOT hit that area yet, so I will not pull the plug entirey let, but on the same token I will pull it if it gets there because I need enough equity to start a new strategy in the future if needed.
Yesterday I went into crazy testing mode, and was really looking in depth not at how my strategy performs, but more about WHY my strategy performs... what edge is it really exploiting..... I especially looked at the second strategy in regards to the first.. they are basically the same strategy but one uses some market internals to help decide when not to trade. What I found is the second strategy's filter makes VERY good sense for this type of strategy...... However, looking at it the way I am using the filter may not be the BEST way... It opened my eyes up to some new ideas on improvements, but thus far nothing good has come from it, but the wheels are turning....
I have decided today to run the second strategy. I believe this strategy is MORE sound. I may not be able to test this strategy as far back as the first due to the amount of data I have... but the principles its built on makes me feel confident that it is a more sound strategy. If nothing else comes out of switching at least the strategy will trade a lot less for the next few days which will give me some time to continue exploring some ideas I had.....
Concerning if I can run the live mode and sim mode at same time.. the answer is yes... I have not done it many times... especially not in this mode of live testing, but back when I had my old auto trader from several months ago I did it a few times to see how reliable the sim mode was and the results were very similar.
I have a question for this thread thats been on my mind for a while. Knowing what we know about the state of world affairs, climate change, oil peak and a general consensus that the economy is heading towards the toilet. Is now a really dumb time to be spending time designing an automated system / getting into the market?
And for those of you who have a more positive outlook than I. if gas prices next summer increase the amount they increase over this summer, I would definitely bet on several recessions / depressions in the next 10 years. It is surprising how gas prices effect costs / profit margins across the whole board.
And for those of you who have a more positive outlook than I. if gas prices next summer increase the amount they increase over this summer, I would definitely bet on several recessions / depressions in the next 10 years. It is surprising how gas prices effect costs / profit margins across the whole board.