Live performance updates for PRC Marketplace strategy: High Probability Reversal

Forums ProRealTime English forum General trading discussions Live performance updates for PRC Marketplace strategy: High Probability Reversal

Currently, there are 0 users and 1 guest visiting this topic.
View all attachments
Viewing 15 posts - 76 through 90 (of 131 total)
  • #189678 Report

    Last year in June I started experimenting with the idea of pausing a strategy for a short while if performance starts deviating from the mean. Typically this is done to prevent trading when performance is slacking off, but the way I envisioned it was to pause the strategy when performance has been unusually good. This was based on my observation that just like the market itself, strategy performance also tends to ‘mean revert’. So in order to do this I devised the use of a linear regression slope which tracks the equity graph and will signal the strategy to pause trading untill performance falls back below for example 1 standard deviation from the linear regression line.

    After today seeing the High Probability Reversal strategy giving back the exceptional gains for March it might be time I start looking into adding this logic to the strategy. Overall the performance remains excellent so this will just be another potential enhancement to the strategy going forward. I will do some tests over the next few days and if satisfied with the results anounce the date of a possible update.

    Attached is a backtest of teh strategy where I projected the linear regression line on the equity graph on 3 different points to demonstrate how performance typicaly relates to linear regression.

    Attachments:
    1 user thanked author for this post.
    #189680 Report

    Plugged in some code and already like what I am seeing on the backtest! Will continue testing and perhaps even push the update through at the end of the week. Clearly it’s a good idea to have a cooling off period after a very good run. Trading while performance is above 1 standard deviation of linear regression is simply counter productive.

    Attachments:
    #189684 Report

    How do you rule out over-optimization when adding another filter to the strategy?

    #189685 Report

    Primarily in the way I approach filters.

    Filters should first of all be logical, in other words you shouldn’t just be looking to add filters or conditions to ‘enhance’ performance (this will inevitably lead to over-optimization).

    Secondly, filters should be based on a sound mathematical or statistical premise.

    So in this particular instance, based on my observation over the years of how strategies typically perform and what I am now once again seeing, it would make perfect logical sense to include an out-performance filter based on the expectation that performance should remain within a statistically significant range. Also in this particular case it is a very straightforward filter that doesn’t even require optimization.

    #189722 Report

    all these fancy words and explanation aside.

    you still optimize the strategy, and launch it without testing it a few months first?

    #189723 Report

    In essence this is not a functional change at all. So the strategy entries, exits and rules remain exatcly the same, hence no need for months of testing.

    It is simply a condition to tell the strategy to take a short break when there has been an exceptionally good series of trades. This provides a simple statistical edge.

    Also when an update is made available, users obviously have the choice to first test it in a demo enviroment however long they want, while still running the previous version.

    #190095 Report

    Hi Juanji,

    I have a couple of questions, if you don’t mind.

    1. I read that €1.000 starting capital is enough. What’s the maximum number of contracts of each trade? I’m asking, because if you are registered as a professional trader by IG markets, your margin requirements might be different than mine.
    2. What’s the initial risk of each trade in GBP or %? Do you have a trailing stop?
    3. My understanding is that I buy a lifetime license. What will happen if you don’t want to continue for example in 1 year. Do we then get the whole code so that we will be able to run the automated trading system on our own device going forward?

    Thanks for your answers in advance!

    Best regards,

    Christian

    #190110 Report

    Hi Diegol

    Here is the answer to your questions;

    1. You only require $1000 euros as the strategy with the multiplier set to 2 (default exposure) will only take up to a maximum of 6 contracts with a margin requirement of ~115 euros for each contract opened. This means that even with 6 contracts margin requirement will stay below 700 euro and also the maximum drawdown is also not much more than 200 euro. So even in worst case scenario of having the maximum drawdown from the beginning and taking into account the max margin requirement, you will be fine with 1000 euros. But it is even possible to set the strategy multiplier to 1 (instead of 2) and it will half the maximum exposure to 3 contracts and hence half the drawdown and margin requirement.
    2. The average loss is usually around 20 GBP with the worst losses around 50 GBP per trade. The stop losses are dynamically calculated based on volatility and time in the trade.
    3. It’s a lifetime liscence meaning it never expires. It will always remain connected to the liscence as this is teh only way to ensure you get updates on the strategy.

     

    I hope this answers your questions?

    #190117 Report

    Thanks that answers my questions!

    One last one, do plan an easter sale?

    #190345 Report

    I will launch an update to the strategy this weekend and along with it a 20% sale (lasting untill the 1st of April).

    The update as discussed in my previous posts will allow the algo to take a break after a significant winning streak to effectively ‘hold onto profits’ while the market conditions that led to the unusual gains get a chance to ‘cool off’ so to speak, this will significantly reduce the chance of the algo giving back gains such as what we saw happening with the current version this month. This ‘break’ will be based on a standard deviation from the linear regression of ‘normal’ performance metrics.

    #190502 Report

    Thanks for the discount offer!

    If you don’t mind, I have three additional questions, before I’m going to take it:

    1. How long is the usual duration of a trade?
    2. Depending on 1, do you hold overnight and over the weekend positions?
    3. What’s the latest performance of your system?
    #190509 Report

    Hi Diegol,

    I appreciate your questions as it helps evryone reading this thread and who wants to understand the algo better.

    1. The average time in the market is just under 4 hours
    2. By default yes, but you have the option to disable holding long or short positions over weekends
    3. Attached is the final update of the current version I have been running live. You will see we had a spike in performance at the beginning of the moth which then mormalized with the last 3 trades resulting in wins.

    March has truly been an anomaly considering the Russia/Ukraine conflict and it’s impact on commodity markets. But the algo seems to have handled it well with performance back on track and still positive for the year to date.

    The new version will obviously handle these types of anomalies much better and even allow the strategy to lock in any gains possibly resulting from such events.

    Attachments:
    #190520 Report

    The new version has officialy been launched and all existing clients will now have received a notification regarding the update.

    For new clients, I am offering a 20% discount untill the end of the month (coupon code: launchpromo)

    Attached is the full backtest results for the updated version.

    I will also starting today once again be running the new updated version on the $1000 live account with the strategy multiplier value set to 3.

    Live results will be posted on a monthly basis with occasional updates in between.

    Attachments:
    #190654 Report

    Is it normal that I can not call the backtest correctly? I can see 15000bars, but already 25000 bars he will not wait after 15 minutes. So I only see the last 2 months.

    #190658 Report

    Look like your equity curve doesn’t look like mine at all…

    How is it that the gap in mine at mid of March is not shown on yours which is completely flat?

    Attachments:
Viewing 15 posts - 76 through 90 (of 131 total)

Create your free account now and post your request to benefit from the help of the community
Register or Login