Live Trading "Alpha Portfolio" Has Gone Live
12/02/2019 at 2:05 AM #114029
Having seen a bunch of questions about portfolios and profitability I have put together a £5k live trading portfolio of strategies that I hope will perform well and not make me look silly. Around £3.5k is assigned for margin and £1.5k for draw downs/P&L.
I am aiming four 5-10% return per month with around the same draw down (Based on the £5k portfolio not just the £1.5K allocated for P&L. I will post regular updates on the performance of the strategies and performance as a whole as well as that of up and coming systems in the development pool on Demo
There are currently seven strategies begin applied to multiple products and time frames, they all have money management and will retain 75% of their profits with the remaining 25% being used to bring in new systems probably at each month end
I will created a more detailed post later this week describing the systems and how I arrived at the relative weightings of the individual systems
How many strategies/product and strategy combinations do you run?
What market/system type would you like to see developed for the portfolio?
Ruark12/06/2019 at 8:52 AM #114305
Im interesting in knowing: Have u done a correlation analysis? Like creating a correlation matrix on all your algos, based on for example the weekly / Monthly returns on each algo?
Im just wondering because looking at all those strategies im wondering if maybe 5K£ portfolio is not enough money to be running all those strategies?
I havnt seen your backtest so obviously i have no idea tho! Im just thinking for myself that my strategies usually have a max drawdown around 800-1200€, if just 3 of your strategies hit max drawdown at the same time, and the other strategies are not making money at the same time, its game over if you have ish the same max drawdowns as me!
I started all my algos literally days before the huge october 2018 crash and i have seen a few of my algos hitting a new all time max drawdown during that time, it was really scary, luckily i would say, for me at least the correlation is and was low enough, i still had a few algos making money while some had their biggest drawdown at the same time. If none of my algos had made money during that time it would have been game over for me.12/08/2019 at 8:27 PM #114448
I have attached an analysis of the back tested portfolio results for 2019 and the raw data taken from the prorealtime results (the portfolio has changed slightly since I posted the image)
My formulation for portfolio size is Margin (unless trading once a week or less) plus 50% of the max drawdown to date (based on a system with no money management)
I have found that the Dax 30 and EUR/USD systems tend to hedge each other well and having a couple of ATR or volume based systems can hedge out those system operating on longer time frames in the occurrence of a news based event so for example on Friday with non-farm payrolls my Wall Street system was triggered short by the early action and then went on to have a loss, however the EUR/USD systems based on ATR hedged it quite nicely
You can see in the attached example that the portfolio doesn’t really ever go below £0 P&L
The maximum net drawdown is £150 ish on a balance of £5k to start and £8,750 at the time of the drawdown
I have attached a graphic of the performance of the portfolio in November with and without Alpha 10 as this system is trigger often
I see a fairly healthy mix with internal hedging in the portfolio, I was in the past running several 5 Minute EUR/USD systems with the bulk of my capital and then the correlations really hurt the portfolio with the market moving from mean reverting to trending
One thing that I did find that I don’t know if you have is that if you code to buy/sell a position then the minimum unit size is 1, however if you code the positions size then you are then limited by the broker so for example the Dax and EUR/USD can be traded for 0.5 per point on IG.com, Wall Street for 0.2 per point which makes it much more accessible than the 1 per point that hard coding the buy signal would require
I found the code below via a link on the forums to a snippet library https://docs.google.com/spreadsheets/d/1rgboqj7sVwsP9ZRhOduOefye48QMWC07jWVXCl-KJPU/edit#gid=0
This code starts with a position of 0.6 per point and will increase the positonsize by 0.1 for every 113 earnt (this is 0.1 x (margin per point + 50% Max DD per point))
Or you could just code positionsize = 0.5 then open positionsize in the system1234567891011121314151617once multiplier=1once fraction=113once newlevel=113once oldlevel=113once startpositionsize=0.6once positionsize=startpositionsizeif strategyprofit>newlevel thenmultiplier=multiplier+0.1oldlevel=newlevelnewlevel=strategyprofit+multiplier*fractionpositionsize=multiplier*startpositionsizeelsif strategyprofit<oldlevel and multiplier>=1 thennewlevel=strategyprofitoldlevel=strategyprofit-multiplier*fractionmultiplier=multiplier-0.1positionsize=multiplier*startpositionsizeendif
One thing that helped me massively was seeing that you could get the test results from the back test report and click and hold to drag them into Excel, prior to this I was keying them out manually
How are you looking at your correlations and building your portfolio?
12/09/2019 at 12:39 PM #11449512/23/2019 at 2:29 AM #115400
Update Week Commencing 16/12/2019
Full details at the link below:01/12/2020 at 1:43 PM #11650701/12/2020 at 8:47 PM #116541
Far from it, going strong and a number of higher time frame (4h Hour) break out strategies are going live this week
I am still doing the proper write up for this week (it did not trade from New years Eve to 2nd Jan) but I have added a couple of screenshots below to wet your appetite in the mean time
1 user thanked author for this post.01/12/2020 at 8:58 PM #116546
Thanks for the update. Nice pics from edgewonk!
How many strategies you need to have when it starts its usefulness using edgewonk for oversight you think?
Could you also show a screenshot as in post 1 plz?01/13/2020 at 12:54 AM #116583
I couldn’t really say how many strategies for Edgewonk, personally I kind of fell into using it because I had it anyway. I use it in a very different way to the discretionary traders that would normally use it as the factors that I monitored when trading in that style, were the kids at home, what mood was I in, analysing my management of the trade and impulsiveness to get in or out don’t really come into it.
At the moment I find that it is very good for measuring a couple of metrics (I use it for demo trading as well as live):
PRT loses the results if the system is discontinued – Edgewonk can capture these details without me needing to leave systems on stop but not removed from the PRT autotrading queue, so for example at the moment I am going through most of my strategies and splitting them into long and short versions using a trading journal allows me to keep the presplit data
If I override the system and close a position PRT doesn’t track it, in Edgewonk I can record the trade and review the gain or loss caused by my actions
I love an equity curve and the underwater and simulator features
It allows me to review the performance of systems by time frame, product, underlying indicators
Essentially its a solution for recording a lot of live and demo trading and analysing it out, if I wasn’t using it I would need to design something that does the same job. You can for example extract the PRT back testing data and format it for import and see how all the systems would have behaved as a portfolio without resorting to Excel.
I run 4 instances, Live, Test, Pension and Portfolio Building
I have attached the current list but as I said some changes have been made as I split the systems01/13/2020 at 11:34 PM #116705
Thanks for the pics. Indeed edgewonk gives much more and better information and that it works for the demo is nice too! At the moment I’ve 8 strategies and as you did, adjusting, adding or deleting as we go but then oversight is gone on which strategy caused profit or loss.
With more strategies oversight becomes more important, but it’s hard to find/create strategies which you can thrust. I agree splitting is probably better and perhaps then there are fewer variables to optimise.
Looking forward to your next update!
1 user thanked author for this post.