Vibe Trading | Page 3 | Forex Factory

Options

Go to
Permalinks
Print Thread

Similar Threads

Day Trading vs. Swing Trading vs. Position Trading 86 replies

Attachments: Vibe Trading
Exit Attachments

Vibe Trading

  • #41
  • May 7, 2026 8:03am May 7, 2026 8:03am
Does your code place the trades automatically, or do you still have to place them manually? Also, which programming language are you using to build it?
  • #42
  • May 7, 2026 9:59am May 7, 2026 9:59am
Attached Image (click to enlarge)
Click to Enlarge Name: Screenshot 2026-05-07 at 15.57.31.png Size: 42 KB


Another range trade, but I have to admit I am not sure how this worked out as ES was dumping at the time, I was definitely thinking NQ would follow, but then ES found support at the VAL and NQ and ES ripped higher
  • #43
  • May 7, 2026 10:06am May 7, 2026 10:06am
Quoting bennigaertne
Does your code place the trades automatically, or do you still have to place them manually? Also, which programming language are you using to build it?
The code is doing everything - I just fed it some ideas, it then built the trade execution scripts, connected up the API and sorted out the trading permissions with Topstep.


Attached Image (click to enlarge)
Click to Enlarge Name: Screenshot 2026-05-07 at 16.02.41.png Size: 193 KB


I get to watch this all day .

From what I can tell the code is Python
  • #44
  • May 8, 2026 1:56am May 8, 2026 1:56am
Quoting Rfastrat
{quote} The code is doing everything - I just fed it some ideas, it then built the trade execution scripts, connected up the API and sorted out the trading permissions with Topstep. {image} I get to watch this all day . From what I can tell the code is Python
Really cool to see how good its working!
  • #45
  • May 8, 2026 6:24am May 8, 2026 6:24am
Quoting Rfastrat
{quote} The code is doing everything - I just fed it some ideas, it then built the trade execution scripts, connected up the API and sorted out the trading permissions with Topstep. {image} I get to watch this all day . From what I can tell the code is Python
I read in the Topstep API documentation that API activity must come from the users local device and not from a hosted server, VPS, or remote environment.
Have you found any way to place trades through a hosted backend, or does it only work with local execution?
  • #46
  • May 8, 2026 11:08am May 8, 2026 11:08am
I personally have a Mac mini, which is always on during the trading week. Pretty sure if I had it on a VP server with a fixed IP address based in the UK, where I am, we could run multiple accounts, but it needs to be all associated with each other, so 1 Topstep Account per IP address. Billing and Payouts the same. I am happy to have it on my local machine, as it is less complicated and less expensive.

Once I have proven the concept with Topstep, I will take my money and put it on a broker account - no point paying other people a % of the winnings, when I am taking all the risk
  • #47
  • May 8, 2026 11:13am May 8, 2026 11:13am
Quoting bennigaertne
{quote} Really cool to see how good its working!
thanks, the hardest thing has been finding strategies that actually work. So many strategies that I was taught that were supposed to be mechanical are totally discretionary. Not saying there is anything wrong with discretionary, but currently Ai is good for systems, not for new ideas. Hoping to train the Ai on my thoughts and actions going forward and see if it can find a signal within the noise. Ai has helped me build out enough of a portfolio of strategies that are not correlated, so I am happy that it will do what the back test says it will do, as we keep going forward.
  • #48
  • May 10, 2026 9:21am May 10, 2026 9:21am
Quoting Rfastrat
{quote} thanks, the hardest thing has been finding strategies that actually work. So many strategies that I was taught that were supposed to be mechanical are totally discretionary. Not saying there is anything wrong with discretionary, but currently Ai is good for systems, not for new ideas. Hoping to train the Ai on my thoughts and actions going forward and see if it can find a signal within the noise. Ai has helped me build out enough of a portfolio of strategies that are not correlated, so I am happy that it will do what the back test says it...
can you share us some hints on tech stacks and framework for steup?
  • #49
  • May 10, 2026 3:42pm May 10, 2026 3:42pm
Quoting phrixusong
{quote} can you share us some hints on tech stacks and framework for steup?
If you go the LLM route to build out a trading system, you can't ask it to generate a profitable trading strategy. It doesn't know how to trade.
So you have to find strategies that work (don't reinvent the wheel), you have to show it how you trade successfully, and you have to expect to bin >90% of strategies that you test. You have to back-test properly on clean data. I give more detail here https://www.rfastrat.com/module1 and will update with my latest findings shortly.
  • #50
  • May 11, 2026 10:45am May 11, 2026 10:45am
Attached Image (click to enlarge)
Click to Enlarge Name: Screenshot 2026-05-11 at 16.42.26.png Size: 143 KB


Fixed some bugs last week and also updated the look of the UX/UI as the Claude code dashboard was starting to look old and like everyone else's who vibecodes.
The bug fix was one which has had an immediate effect. The bug had been suppressing quite a lot of the live signals from a few of the strategies that place a Limit Order. In the backtest, this actually saved me some money, as the trades that didn't trigger would have been losses. Today, we're also running a small DD after 4 trades.
  • #51
  • May 12, 2026 7:44am May 12, 2026 7:44am
Did you start by using paper trading first to test your system and place orders, or did you go directly with the Combine/live execution? And have you passed an Account yet?
  • #52
  • May 12, 2026 8:22am May 12, 2026 8:22am
Quoting bennigaertne
Did you start by using paper trading first to test your system and place orders, or did you go directly with the Combine/live execution? And have you passed an Account yet?
I started with backtesting, then I always paper trade (fwd test). Signing up for a combine is the only way to get the account linked to the API. Passed the combine and now trade an Express-funded account XFA.
  • #53
  • May 12, 2026 8:44am May 12, 2026 8:44am
Quoting Rfastrat
{quote} I started with backtesting, then I always paper trade (fwd test). Signing up for a combine is the only way to get the account linked to the API. Passed the combine and now trade an Express-funded account XFA.
Ahh i always wondered how to get access to the API keys... makes sense now.
  • #54
  • May 12, 2026 1:36pm May 12, 2026 1:36pm
Quoting bennigaertne
{quote} Ahh i always wondered how to get access to the API keys... makes sense now.
the API was through ProjectX but is now solely through TopstepX
  • #55
  • Edited 7:39am May 14, 2026 2:27am | Edited 7:39am
Attached Image (click to enlarge)
Click to Enlarge Name: Screenshot 2026-05-14 at 08.21.29.png Size: 65 KB


Trading small has never been so profitable. I am usually trading 1-mini, scalping in and out. Invariably taking a massive loss after my ego can't take it that I am wrong. Yesterday was my best-ever trading day for execution ( i have earned more in a day before), and I wasn't even at a computer. This follows the previous day, where a phantom trade lost me $450 because of some crap code Claude had written. So what have I learned? I definitely want the algo way of trading to be my primary way of making money. I will still have to accept losses, but now, if it is due to bad code or crap logic, I can fix and patch better code to make sure those mistakes never occur that way again. Something I tried to do with my own psychology, but never accomplished. If I could have just learned from every mistake, I wouldn't need to go do the algo route. If I had self-control, discipline, better ideas, if I could rely on myself, I'd be a very, very happy trader now.
  • #56
  • May 14, 2026 6:54am May 14, 2026 6:54am
Quoting Rfastrat
{image} Trading small has never been so profitable. I am usually trading 1-mini, scalping in and out. Invariably taking a massive loss after my ego can't take it that I am wrong. Yesterday was my best-ever trading day for execution ( i have earned more in a day before), and I wasn't even at a computer. This follows the previous day, where a phantom trade lost me $450 because of some crap code Claude had written. So what have I learned? I definitely want the algo way of trading to be my primary way of making money. I will still have to accept losses,...
you've definitely got the hang of the vibe coding - not withstanding Claude cocking up orders. What did it do so we can all avoid?
  • #57
  • May 14, 2026 7:06am May 14, 2026 7:06am
Quoting triphop
{quote} you've definitely got the hang of the vibe coding - not withstanding Claude cocking up orders. What did it do so we can all avoid?
Claude probably fucked state tracking or error handling, thought you were flat while a position ran wild.
Drop the exact symptom/snippet and I’ll give you the fix in 1 line. Ego stays off, code stays cold.
  • #58
  • May 14, 2026 7:48am May 14, 2026 7:48am
Quoting triphop
{quote} you've definitely got the hang of the vibe coding - not withstanding Claude cocking up orders. What did it do so we can all avoid?
it's simple stuff that humans would have spotted if they had placed an OCO / bracket order. The system is supposed to delete the SL if the TP is hit and vice versa. The system is supposed to delete the SL and TP if we get to the end of the trading day. The system is supposed to delete the orders if nothing triggers after a period of time. One of these things didn't happen. So my system was at the time, supposed to delete/cancel/flatten anything that wasn't recognised. Including manual discretionary trades. Unfortunately, I had a couple of phantom trades, and the fail-safe didn't do its job. One trade went long the MNQ and the market was dumping, so took a big loss there when I finally realised. The second phantom trade had gone into profit, mitigating some of the damage. With them being phantom trades they were placed at the broker, but were not recognised by my dashboard. The only tell would have been seeing the code fire off the trades and that would mean me sitting at the desk (which is not what I want to do), the other tell was that a legit trade had pushed past TP in profit but hadn't closed. Because the rest of the trading engine was now out of whack. I noticed the profits on my dashboard, investigated why it hadn't closed out, discovered the previous problems, flattened everything. Then set about patching everything up and asking the system to check its own checks and now we have a harness where multiple eyes are looking for odd lots and actions. There could still be a trading mine, sat in the code, just waiting for the perfect time to blow up, but I am more confident today than Monday that I have put extra measures in to help when it does explode in my face.
  • #59
  • May 14, 2026 7:52am May 14, 2026 7:52am
Quoting MarsForex
{quote} Claude probably fucked state tracking or error handling, thought you were flat while a position ran wild. Drop the exact symptom/snippet and I’ll give you the fix in 1 line. Ego stays off, code stays cold.
This is a very kind gesture, thank you. I spent Tuesday going through everything and I think I have vibecoded my way back to safety. The trading day yesterday went very well. All of the strategies that fired off were the same ones that tried to screw me over a day or so before. If anything ever pops up again, I will message you if Claude can't help. I wish I had stuck with coding from my early start with Spectrum 48k and BBC micro computers. I would be a frickin whizz at this now. Ai is filling in several decades of knowledge, or so I hope it is.
  • #60
  • May 14, 2026 8:14am May 14, 2026 8:14am
Quoting Rfastrat
{quote} it's simple stuff that humans would have spotted if they had placed an OCO / bracket order. The system is supposed to delete the SL if the TP is hit and vice versa. The system is supposed to delete the SL and TP if we get to the end of the trading day. The system is supposed to delete the orders if nothing triggers after a period of time. One of these things didn't happen. So my system was at the time, supposed to delete/cancel/flatten anything that wasn't recognised. Including manual discretionary trades. Unfortunately, I had a couple of...
Eeesh, that's grim. well done for getting it back on track - I think you've expressed my worst fears of automation right there. Perils of vibe coding but if its any consolation a mate of mine works for a quant fund. He wouldn't appreciate any details but imagine that x a few thousand orders sub 2 secs. By mistake.
0 traders viewing now