US100 mini – Overnight Swing Strategy – AI
Forums › ProRealTime English forum › ProOrder support › US100 mini – Overnight Swing Strategy – AI
- This topic has 26 replies, 6 voices, and was last updated 1 year ago by PeterSt.
-
-
11/28/2023 at 2:58 PM #22450611/28/2023 at 2:59 PM #22450711/28/2023 at 3:01 PM #22450811/28/2023 at 3:13 PM #224514
Hi Johann, I fully agree with you.
It is imperative to rather stick to one market and/or instrument and then to rather diversify over different strategies within that market and/or instrument.
A “sniper” approach works better for me than a “shotgun” approach.
4 screenshot .jpeg files were uploaded.
11/28/2023 at 3:22 PM #22451511/28/2023 at 3:47 PM #224520Yes Niel, looked at the screenshots. I for one prefer to have longs and shorts and be in and out the market more often. But, hey beauty is in the eye of the beholder. Although I agree with position sizing playing a huge part in effective returns one must be careful to use position size with back tests and development to prop up the returns. I would rather develop a successful system without changing positions and after the system has proven itself, then only will I began to develop code to support the cumulative effect. So if you take your system as is but neutralize the position changes based on strategy profits and the system stil shows promise then you know you have something you can work with.
11/28/2023 at 6:39 PM #224526Hi Johann, I fully agree with what you have said. I have just did that.
My base code have been tested since the 9th of February 2021 on the DOW, on 34 second candles…playing 1 contract with cumulate orders activated.
Attached you will find the live results of the actual algorithm running on the demo environment for the last 33 months.
I am sure you would agree that the odds of the foundations have been covered…
11/28/2023 at 7:04 PM #22452911/28/2023 at 9:47 PM #22453511/29/2023 at 2:26 AM #224536Hi Niel,
I prepared a couple of responses which merely had a “cautious” mood, but I was reluctant to post them.
I now think I did well on that (not post them).It is not easy to explain (or confirm) how good your approach is, which merely is about experience on for example the win rate (which you only now just showed – or at least just now got to me – and which is in the area that I would prefer myself for a 100 reasons).
What is also new now, is the number of trades per day, which may look “suspicious” (the biggest aim I have myself). Regarding that, here is an excerpt of one of my texts I wrote earlier today but did not post yet, but which may be applicable :
Did you take into account the spread sufficiently ? I say this because I see you using the ZAR, which probably is your local currency and thus nothing wrong with that. But in relation to other currencies – probably always in order ? – it may not work out at all.
My experience does not go further than e.g. EUR/ZAR, and that does not work out because of the huge overnight interest (and huge spread IIRC).
If you recognize that I could be correct, it is a matter of not using the ZAR *and* having your portfolio account in USD (looking at your other examples). Now there is no interest in order for the USD instruments, but for the European there still is – plus of course what you use on margin.I now post this text because to me it may be obvious that at 31 trades per day (15.5 In + Out) you could be “scalping” as such, not taking into account the spread or commission. I don’t think I could do this, hence I would need longer being OnMarket in order to gain what you do, which conflicts with the 15 full trades per day. I may achieve 7 or so.
The Gain with that is unrelated because this is all Money Management and your 90% may be a theoretical 400% in my case. This is again not important but proves the base method you use, which really would be my thing.Please notice that although I use such means in practice, in real life it does not work out because of too many times our Systems being kicked out because of IG broker issues**. So you would get crazy of restarting them with the proper Money Management parameters fed (and which you need to keep track of because nothing else does that for you). Net I now don’t use it, because it would take more than one day effort per day and one human being can’t do that (hire people for it …).
What you do is to my heart anyway … 🙂 🙂**): As we speak, see below. Not that this happens every hour, but every day would come close and restarting it with the current MM parameters would take me 30-60 minutes to do it in a decent fashion. Restarting it without MM parameters takes a minute. Don’t worry about the reasons of being kicked out for now, it is something we have to accept.
Another real life matter is the limit on position size; if that approaches USD 5M per trade, IG will be watching you and put you on “manual”. This may be fair, but now forget about the number of trades you “need” per day because regarding that nothing will work any more. So notice that IG has special triggers for AutoTrading, and nothing of those we will like for the better.
A kind of solution is make all trades Market Orders, but I suppose you will know that this doesn’t bring the best spread figures (to say it mildly and without further elaboration). And might you be on (or consider) IBKR (PRT-IB), then Limit/Stop orders are out of the question anyway because of a too slow response of IBKR and PRT not being able to deal with that (system is kicked out again). The result is once again way less profit per trade. And so in all this context I look at your great figures. I have them too. In Backtest.For a heads up : it is all doable but it may take you a year more in making things more decent and more profitable (because else you may be under water).
11/29/2023 at 9:43 AM #224543Attached you will find the live results of the actual algorithm running on the demo environment for the last 33 months.
Absolute’s performance stats he posted in (at least) his 2 most recent screenshots (maybe all) are Live running (Demo Live).
Just saying as folks seem to be under the impression / commenting as though Absolute’s stats are from backtest?
11/29/2023 at 10:01 AM #224545I now post this text because to me it may be obvious that at 31 trades per day (15.5 In + Out) you could be “scalping” as such, […]
No wait … I am observing one of my own current trades and I now realize that I do this all quite differently than “normally-You”. So what I did not think of in your case, is that each accumulated (or decumulated) trade counts as a trade as well. My own trades always are 1 from start to end (okay, 2 because In and Out). See below for an example of what happens over here (mind the column with the position size).
But this depicts the additional issue for you :
Where I can control my Money Management in between trades, you can only manage them within the single one trade (position-taking for your System) – at least for the situation you showed in your last post. If you ask me, this is “unmanageable” because you would not know the start-through data. You will only be able to see the relative position changes (one additional, three additional, five reversed, etc.).The problem is that in PRT we can’t have I/O to the outer world, so we won’t know the values of the internal variables. Notice, however, that ProRealTime has some functionality in the works which is going to be able to “start-through”, although they say themselves that it will be a challenge for them to arrange for that really. I say : then better quit your AutoTrading features all together, because nothing works for long. I could also say (and claim) : some are full with bravery and proudly etc. telling that we now can AutoTrade with IBKR, but the very few know that actually nothing works for real because of the constantly throwing out (which is for IG the same, but for different reasons which actually are better tweakable via backdoor solutions). The real problem emerges here (an open door by now) :
Because no System with normal allowed means (for program code commands) will run for very long, there is no real opportunity to let work what you (and me) so nicely worked out. I could also say : do it without any Money Management and your problems are behind you. But what did you announce in your first post ?
“position sizing is probably responsible for 90% of your performance variability”.
Right.
And exactly that we can’t do in PRT because of all the issues.On a side note, these issues are 99% broker-initiated (IB as well as IG), and other Platforms really won’t do better. But there are two huge differences :
- Other Platforms don’t work server side (which always had its advantage, but not any more).
We are at the mercy of the Gods hence in this case ProRealTime, which don’t seem to have the creativity to solve things decently (yes, I stand for these words). - Other Platforms work with far more normal programming languages, if not 100% normal (think C# or VB.Net or C++).
With those I could do anything I’d want for I/O and tracking of internal variables and starting-through means.
Let me add that anything which would make it slightly more difficult for people without any coding knowledge, will not be honored by ProRealTime. “People can’t understand that”.
- Other Platforms don’t work server side (which always had its advantage, but not any more).
-
AuthorPosts