Maybe!!Don't tell me you are going to let your winners run?
The myth of letting your winners run
- Thread starter Flowfollower
- Start date
I have been using several optimizers like Adaptrade, stratagyquant
and Stratasearch and I bought GSB which is about the best and has very good OOS especially for ES. Adaptrade was useless, it has a mind of it's own and I could not get something usefull out of it.
To my surprise I have found good OOS performance running it on 430 bigger stocks of the SP500. I use a complex custom criteria formula only, so not all the rules with the weights and exclusions because my experience was that this did not work well. The systems that have a trade% of around 2%, a profit factor of more than 2 and a RRR of bigger than 2 and less than 20% drawdown and thousands of trades over 2002-2019 do work fine OOS in general and pass the covid crash well. The trades should not be much longer than 1-2 weeks.
I typically see the last 2 years OS (after testing 2002-2019) as much profit as the SP500 but with only 30 -50% in the market and with much lower drawdown. I also check for shifted (the S parameters for parameter stability) returns both by excluding and by having it in the custom criteria to find systems that have a broad area where they work even when parameters are shifted by +/-3.
I have bought a lot of software and was surprised to find a free software like Stratasearch that works quite good for me. When using SS on a single ticker it is difficult to avoid curvefit and I also find it performs bad OOS. I have put quite some time in it and have found out everything on my own.
If somebody could help me with old Stratasearch material like video or userforum material I would be happy to pay for that.
I also use Amibroker with an Elliotwave application from an Italian programmer and that has been very precise in predicting turning points in sectors. Running that ap on a sector and seeing which stocks all have wave 3's in several timeframes gives a precise indication.
I look for confirmation as broad as possible which is easy to do with the Amibroker explorer that shows the waves for all the stocks.
and Stratasearch and I bought GSB which is about the best and has very good OOS especially for ES. Adaptrade was useless, it has a mind of it's own and I could not get something usefull out of it.
To my surprise I have found good OOS performance running it on 430 bigger stocks of the SP500. I use a complex custom criteria formula only, so not all the rules with the weights and exclusions because my experience was that this did not work well. The systems that have a trade% of around 2%, a profit factor of more than 2 and a RRR of bigger than 2 and less than 20% drawdown and thousands of trades over 2002-2019 do work fine OOS in general and pass the covid crash well. The trades should not be much longer than 1-2 weeks.
I typically see the last 2 years OS (after testing 2002-2019) as much profit as the SP500 but with only 30 -50% in the market and with much lower drawdown. I also check for shifted (the S parameters for parameter stability) returns both by excluding and by having it in the custom criteria to find systems that have a broad area where they work even when parameters are shifted by +/-3.
I have bought a lot of software and was surprised to find a free software like Stratasearch that works quite good for me. When using SS on a single ticker it is difficult to avoid curvefit and I also find it performs bad OOS. I have put quite some time in it and have found out everything on my own.
If somebody could help me with old Stratasearch material like video or userforum material I would be happy to pay for that.
I also use Amibroker with an Elliotwave application from an Italian programmer and that has been very precise in predicting turning points in sectors. Running that ap on a sector and seeing which stocks all have wave 3's in several timeframes gives a precise indication.
I look for confirmation as broad as possible which is easy to do with the Amibroker explorer that shows the waves for all the stocks.
There was a backtester/optimizer called Stratasearch which is no longer around...You could throw in as many rules as you liked with money management,and it would run thru hundreds of thousands of permutations and of course curve fit...More often than not,the exact opposite of what I believed backtested better than the holy grails...And of course,when I walked it forward,it all fell to shit
Last edited:
Any rule based system providing long/short signals has an important metric called E-Ratio of edge ratio that shows a favorable movement of an instrument after a signal. The Build Alpha software has this metric among others and it shows if the edge is real or not. Also it shows how many bars move in favorable direction after a signal after which holding a trade any longer becomes worthless.
A system based on pyramiding/scaleIn-scaleOut can be allowed to run for a longer time. But that requires diversification and posiiton siziing rules like Turtle traders.
https://www.buildalpha.com/eratio/
A system based on pyramiding/scaleIn-scaleOut can be allowed to run for a longer time. But that requires diversification and posiiton siziing rules like Turtle traders.
https://www.buildalpha.com/eratio/
In Strataserach I use $V_AvgPosDrawDPct in a custom criteria which optimizes MAE.
It is the average position drawdown. I don't often use advanced metrics like sharpe
and many more but keep it simple to the metrics that I want to see. It works for me.
I know a experienced user who went from GSB to build Alpha and after a year
went back to GSB mailing me the systems were hard to reproduce in reality. He did most ES.
How long did you use Buildalpha and with what systems/markets do you have good results?
Thanks for your interesting Corona links. I red about the cave theory.
Searching for "The Case Is Building That COVID-19 Had a Lab Origin" will give some very
interesting insights from 2 researchers who investigated people dying in a cave with many bats 2012 having similar sympons as Covid19. They translated the original chinese rapport of those supposed first cases. That material was brought to the Wuhan lab and they suggest
that could have been the first cases. It is an interesting theory next to the lab engineering one.
It is the average position drawdown. I don't often use advanced metrics like sharpe
and many more but keep it simple to the metrics that I want to see. It works for me.
I know a experienced user who went from GSB to build Alpha and after a year
went back to GSB mailing me the systems were hard to reproduce in reality. He did most ES.
How long did you use Buildalpha and with what systems/markets do you have good results?
Thanks for your interesting Corona links. I red about the cave theory.
Searching for "The Case Is Building That COVID-19 Had a Lab Origin" will give some very
interesting insights from 2 researchers who investigated people dying in a cave with many bats 2012 having similar sympons as Covid19. They translated the original chinese rapport of those supposed first cases. That material was brought to the Wuhan lab and they suggest
that could have been the first cases. It is an interesting theory next to the lab engineering one.
I purchased BuildAlpha about a year back....I am using mostly EOD data on Indian and US markets to work out strategies through data mining....Things I like are that it divides the data into test and train sets so the curve fitting problem is eliminated...and edge ratio as I mentioned along with other metrics help a lot.....I have done strategies coding on Ninjatrader7-8 and MT4 before but BuildAlpha saves a lot of time....I am currently making a portfolio of non-correlated strategies...Having said that, its hard to find working strategy on intraday data.....also short signals are very hard to find on indices and ETFs.
Why?
I can see having a stop hit and re-entering.
How do you decide whether to scale in or scale out if the price is trending in your direction?
You should always let winners run and cut losers. But in many cases you only know in hindsight what would have been best. To try to solve that problem I did an extensive test daytrading futures ES.
I did over 1,000 daytrades. For each trade I put in an Excel:
What I learned from this experiment:
My conclusion was to enter and exit based on indicators and a well placed stop. In the long run nothing else that I tried could beat that.
I did over 1,000 daytrades. For each trade I put in an Excel:
- Entry
- Exit
- Max open profit
- Max open loss
- Stop (that I could change to see the impact of that change)
- Take profit (that I could change to see the impact of that change)
What I learned from this experiment:
- I was able to see where the optimal place was to put a stop as I had the max open loss for each entry.
- I saw the distribution of the profits and losses, and the importance of letting profits run. Letting profits run was important as these profits represented a larger part of the net profits. So putting take profits instead of getting out based on indicators had a negative impact on the net profits.
- Changing the stop resulted in less losses in bad trades, but also in being stopped out in a trade that would have been profitable. For each profitable trade that was stopped out I had to check if I could reenter again and still take some profits. At the optimal stop I did not get stopped out of winning trades.
- I had a clear view on the expectancy.
My conclusion was to enter and exit based on indicators and a well placed stop. In the long run nothing else that I tried could beat that.
To me that sounds like you don't know how far the trend will go.
It is very difficult to get out of a trending trade and to get in again at a better price. Most of the time you will not succeed as the retracements are in general small, can stop suddenly and prices can move fast in direction of the trend again. So you will miss a lot of real good exits and also the real good reentries.
A good example is the cannabis stocks. I had small positions in them a couple of months ago and scaled in during the run-up, but scaled back out as price dropped. I'm still in the initial small positions before the move up and down, and will scale back into them when their prices start to go back up again.