Broker data anomoly closing auto systems
-
Evening (here) all! I often read of some less than favourable situations occuring with auto systems on this forum. Whilst often they get resolved being a coding issue or in playdough accounts, today I present a case which clearly is a broker issue and very much my issue. Now I feel your pain. I have contacted both IG and opened a ticket with PRT with no response as yet. The new IG AI chatbot very kindly told me some things which could have occurred with margin yada yada…. I asked to speak to a human and got offered an email communication instead. But don’t worry, the email was clear to state that if you are ‘small town’ you will be prioritised accordingly. The Case: On Sat morning cash close I see the price has gapped up 2.11% on my market ASX200, knocking out my shorts. I think this is fine as I have some long positions, much more productive trades. So I note the Trailing stop has not moved with the price. Very concerning all weekend. This morning I get a phone ping saying the ASX200 has dropped 2.29%. As if it never happened, though it is still here on my charts and my phone app (IG) did record the highest price in the glitch. The IG charts are corrected as if trading as usual. So here, I present the charts and my resulting actions. I show with both HLOC an HA candles to highlight this perculiar behaviour. I suspect I got out favourably monetarily even though these are tiny trades. Note, the 1.5X trade is a DSP system and does not use a hard stop is unscathed. But I won’t be small forever, I am scaling rapidly and have grave concerns moving forward with my Quant business using this provider should it not get resolved. So I show this to my Euro buddies for any feedback please or at very least serve as a warning to others. On the most part I have very little issues with execution with my providers and instruments and take off 15% performance when I model portfolios due to slippage etcetera. This has served me well when it comes to performance expectations. This post is not a complaint about tiny losses as I am a trader and can lose just fine, but about TRUST and accountability.Oh, importantly I forgot to add that the Long Position closed out at a random price as part of this glitch. Giving me a loss and the trade should still be in. The auto log has recorded the result but notably NOT rejected. See image attached ( I have small image of the system BB XJO 4hr sorry I cannot delete this in edit function) The EOM shorts have their back test showing monster losses (yes even in tick data). Yet real stops were honoured.
How’s your DSP adventure going?
Hi JS. Myself and my family are forever indebted to you! I am lazy to get back and check in as they are still testing having completed 5 months now. It worked, all those sleepless nights coding and testing. I came up with 4 useful signals that I wrote 6 or so systems with. Lo and behold my win rate has climbed, accuracy is outstanding on the signal only systems (as opposed to the full time DSP models). I no longer have fomo on anything because the full time models will always be exposed and generally on the right side however slow to lock in. In fact, they are like a flock of birds changing direction and has been an amazing regime filter for my manual trading. I haven’t yet made fast models with any accuracy. They are still slowbots but I run 14 algos meaning I’m trading multiple times a day/week. Half my profits are from this new generation and largely volatility/phase models being the two major components. I’m now consistently producing 10% per month return (auto only) but yet to have a true draw down. So far 50% return with 7.5% DD. They are outperforming so much I don’t day trade any where near as much as I used to so I actually have a life. I just do one VPOC open drive scalp each morning for income if I’m around but of course I am trying to automate that also. How are yours going? Please tell. The best thing I find is none of this silly stops getting taken business as per regular trades where your stop has to go somewhere! Holding costs aren’t too bad but I run the full time systems small because they can swing around a bit. If the market ever crashes we will print money with this style of system. When I turned on the Dow15 the first move was 2000pts. I couldn’t stop laughing. I am about to do my first compound so the next 6months will be interesting indeed. I never performed an FFT. It sort of became obvious once I looked hard enough. I did reinvent the wheel on one occasion but that just helped deepen my understanding of the topic. So now I know 1%!…and it is enough! I saw you had some issues with the broker also. It’s frustrating isn’t it? I don’t expect much either but they asked for exact prices/times to investigate this morning so that is something.JS thanked this postHi Coincatcha,
Great to hear that things are going so well — and of course, you’ve got no one but yourself to thank for it, after all those sleepless nights coding and testing…
The numbers — 50% return with a 7.5% drawdown — look fantastic for a breakout/trend system…
You’re right, using the stop loss differently is one of the major advantages of such a system (I still use an emergency stop of 3% myself)…
Indeed, the first months of compounding positions are always exciting…
My own system is also performing well; of course, I’m still improving it every day…
My biggest breakthrough so far has been applying the z-score to the data…
The “z” stands for Zahl (German for “number”), and in statistics it’s used to compare different data sets…
The idea behind it is that each dataset has a mean and a standard deviation, so when you apply the formula ((Close – Average) / Standard Deviation), you obtain a scaled, dimensionless number — a normalized value…
Essentially, you’re normalizing or scaling the data, but the main advantage is that your system becomes adaptive — it adjusts itself to the average and the volatility of the current data. It works beautifully…
Of course, the system behaves like a typical breakout/trend model with a low win rate and a relatively high drawdown (16%), but it’s still the best one I’ve built so far…
At this point, it has become quite simple — it really comes down to three key steps:
-
Decomposing the data
-
Normalizing/making the data adaptive
-
Using thresholds to interpret the data
If you know even 1% about digital signal processing, it helps 100% when designing algos…
As for brokers, I now trade exclusively with IB, and what a relief it is to be done with all the IG nonsense…
My performance is also much better with IB…
Great hearing from you again, and best of luck going forward!
coincatcha thanked this post -
- You must be logged in to reply to this topic.
Author
Summary
This topic contains 33 replies,
has 5 voices, and was last updated by
11 months ago.
Topic Details
| Forum: | Platform Support: Charts, Data & Broker Setup |
| Language: | English |
| Started: | 11/03/2025 |
| Status: | Active |
| Attachments: | 19 files |