Trading for a living | Page 33 | Forex Factory

Options

Go to
Permalinks
Print Thread

Similar Threads

Living in China, trading for a living 18 replies

1 trade per pair per year 6 replies

multiple systems per pair/multiple pairs per system? 3 replies

Trading 24 hours per day, 5 days per week 14 replies

Attachments: Trading for a living
Exit Attachments

Trading for a living

  • #641
  • Nov 13, 2025 6:20am Nov 13, 2025 6:20am
Bit early for stats, but this is so far
Attached Image (click to enlarge)
Click to Enlarge Name: screenshot.png Size: 3 KB
  • #642
  • Nov 13, 2025 9:44am Nov 13, 2025 9:44am
Quoting M31
{quote} You can automate this method, but bring supertrend in is a bias check{image}
What is the entry logic behind the circles marked in red?
  • #643
  • Nov 13, 2025 9:51am Nov 13, 2025 9:51am
Quoting M31
This is a typical trading desk, loads of emas, drawing recent highs as support across, and an RSI with a moving average like a TDI. TradingView {image}
Haha. Yes... That's me in the future.

Attached Image (click to enlarge)
Click to Enlarge Name: unnamed.jpg Size: 158 KB
  • #644
  • Nov 13, 2025 9:54am Nov 13, 2025 9:54am
Attached Image (click to enlarge)
Click to Enlarge Name: terminal64_TUHID1XDEc.png Size: 11 KB

Attached Image (click to enlarge)
Click to Enlarge Name: terminal64_7wqiACaDqi.png Size: 22 KB
  • #645
  • Nov 13, 2025 10:04am Nov 13, 2025 10:04am
Quoting Noisy
{quote} What is the entry logic behind the circles marked in red?
Middle was divergence, Bollinger band B,

first one, actually 3 at the bottom, all 20-sma cross with candle and supertrend as direction.

Top one is a condensed in a hurry wave 5 with belt hold line, off the 4hr FVG.


RSI with period 4-6 ish will get these as signals to enter with supertrend as direction and or candle patterns as a second check.

Supertrend is period 9, multiplier 3 but with adaptive code

double ratio = currentATR / avgATR;

and then

inpATRMultiplier / ratio

ATR average is 100

inpATRMultiplier is the supertrend output

With that it's currently 2.22
  • #646
  • Nov 13, 2025 11:14am Nov 13, 2025 11:14am
Quoting Noisy
The small experiment is progressing. {image} {image}
Well done! Perplexity and I would like to reward the good work by suggesting a performance improvement.
Attached Image (click to enlarge)
Click to Enlarge Name: pyramiding_comparison.png Size: 129 KB
Attached File(s)
File Type: pdf pyramiding backtest.pdf   608 KB | 1,064 downloads
File Type: csv pyramiding_strategies_comparison.csv   1 KB | 65 downloads
  • #647
  • Nov 14, 2025 4:02am Nov 14, 2025 4:02am
Quoting medici
{quote} Well done! Perplexity and I would like to reward the good work by suggesting a performance improvement. {image} {file} {image}

Good idea. I've always wanted to try pyramiding anyway. The problem I've had most often with it is that when doing it manually, I completely messed up adding at the right time. Managing stop-loss orders manually can also quickly become messy. The art lies in doing this without increasing the risk of the initial position. I'll try to incorporate that pyramiding functionality into my little experiment. I wanted to finish the trailing profits this weekend; maybe I can manage it all at once.

The current version of the system appears to be working as intended. Currently showing a slight loss.

Attached Image (click to enlarge)
Click to Enlarge Name: terminal64_o2m3c5Xej9.png Size: 12 KB


Attached Image (click to enlarge)
Click to Enlarge Name: terminal64_3B3XPuPr6H.png Size: 31 KB
  • #648
  • Nov 14, 2025 4:16am Nov 14, 2025 4:16am
Quoting medici
Why not play with this toy instead: https://www.automated-trading.ch/NT8...range-breakout? It uses FVGs to find its trades.
I've now got this and will test it out as well. I've been waiting ages for my live account with NinjaTrader to be activated, but support seems to be having trouble getting it sorted and keeps asking me for more documents. I also have a live account with Interactive Brokers, which I think can also be used with NinjaTrader, but it currently has no balance. The margin requirements are also much higher there. You don't get real-time futures data that way, and I can't even test a demo account that doesn't expire. I had one with NinjaTrader that included real-time market data, but those expire after 14 days. I'd like to try buying a futures prop challenge only as a second step, once I have a better understanding of how the system works under live conditions.
  • #649
  • Nov 14, 2025 5:59am Nov 14, 2025 5:59am
Quoting M31
{quote} Middle was divergence, Bollinger band B, first one, actually 3 at the bottom, all 20-sma cross with candle and supertrend as direction. Top one is a condensed in a hurry wave 5 with belt hold line, off the 4hr FVG. RSI with period 4-6 ish will get these as signals to enter with supertrend as direction and or candle patterns as a second check. Supertrend is period 9, multiplier 3 but with adaptive code double ratio = currentATR / avgATR; and then inpATRMultiplier / ratio ATR average is 100 inpATRMultiplier is the supertrend output With that...
Thanks for your explanation. I'll take a look at it as soon as I've finished the random based system. I'm guessing mid/late next week.
  • #650
  • Nov 14, 2025 9:35am Nov 14, 2025 9:35am
RandomOrderEA 1.30

- Trailing Take-Profit and Pyramiding implemented but not yet tested.

Three trades so far today, two losers and one winner. Current profit: +$182. I'm going to switch off for two days now. Wishing everyone a relaxing weekend.

Attached Image (click to enlarge)
Click to Enlarge Name: terminal64_pj9PHSK76H.png Size: 8 KB

Attached Image (click to enlarge)
Click to Enlarge Name: terminal64_yL0QQmTgtX.png Size: 19 KB
  • #651
  • Nov 15, 2025 10:03am Nov 15, 2025 10:03am
Quoting medici
{quote} This is a well-known puzzle among traders, simply because there are many general lessons about trading that can be learnt from it. I asked Perplexity AI to make its contribution. See attached. Comments?
Quick question. Have you ever tried to verify a backtest report like this one for a strategy that comes from Perplexity? And could you perhaps recommend some resources where I can learn more about Perplexity, especially in relation to financial analysis? I'd like to read up on it a bit.
  • #652
  • Edited 11:18am Nov 15, 2025 11:03am | Edited 11:18am
Quoting Noisy
{quote} Quick question. Have you ever tried to verify a backtest report like this one for a strategy that comes from Perplexity? And could you perhaps recommend some resources where I can learn more about Perplexity, especially in relation to financial analysis? I'd like to read up on it a bit.
No, I have not. i know the backtests are not done using tick-by-tick data, so they are pointers rather than statistical facts. Which is good enough given the context. Further analysis would have to be made on appropriate platforms.

As to Perplexity, it would be rude not to let it speak for itself. Response to your questions in the attached files.

Also, this is worth watching:

Inserted Video
Attached File(s)
File Type: pdf perplexity AI for financial analysis.pdf   864 KB | 1,017 downloads
File Type: pdf the best sources for learning about perplexity Ai.pdf   564 KB | 710 downloads
  • #653
  • Nov 16, 2025 2:13am Nov 16, 2025 2:13am
Quoting Noisy
RandomOrderEA 1.30 - Trailing Take-Profit and Pyramiding implemented but not yet tested. Three trades so far today, two losers and one winner. Current profit: +$182. I'm going to switch off for two days now. Wishing everyone a relaxing weekend. {image} {image}
Excellent work! Perplexity and I respond with another suggested performance improvement:

What if trades are still entered at random times, but now in the direction of the slope of an Exponential Moving Average?

Answer in the attached report.
Attached File(s)
File Type: pdf random time but with slope of EMA.pdf   447 KB | 155 downloads
  • #654
  • Nov 16, 2025 4:19am Nov 16, 2025 4:19am
Quoting medici
{quote} Excellent work! Perplexity and I respond with another suggested performance improvement: What if trades are still entered at random times, but now in the direction of the slope of an Exponential Moving Average? Answer in the attached report. {image}
Thanks a lot. I'll try to incorporate that as well.

However, the values that Perplexity spits out as a result of its "test" seem a little too good to be true.

Total Return: +786.875% | Win Rate: +48.1% | Max DD: -6.31% | Profit Factor: 3.60

I mean come on? I would die for these values of only over 1000 trades right now, hehe.

It might depend on how the test is conducted, in particular on which data. For example, it mentions "5,000-bar synthetic NQ price series with realistic properties." So it appears that Perplexity used AI-generated data for their test results. It might explain the overly optimistic theoretical values instead of testing the whole thing on real historical market data (tick-by-tick as you already said), especially when execution delay, spread, fees, etc. are added.

Regarding the strategy description itself, the information is somewhat confusing and not entirely clear, which in turn makes it difficult to understand which exact criteria the results are based on. For example, in the upper part of the report, Perplexity mentions: "Entry Timing: Still random (every 4-6 bars)". In the lower part, it then says: "ENTRY: - Random timing every 30-45 minutes".

This is of course a significant difference whether a random order was executed every 4 minutes, every 6 minutes, every 30 minutes or every 45 minutes.
  • #655
  • Nov 16, 2025 5:22am Nov 16, 2025 5:22am
Quoting Noisy
{quote} Thanks a lot. I'll try to incorporate that as well. However, the values that Perplexity spits out as a result of its "test" seem a little too good to be true. Total Return: +786.875% | Win Rate: +48.1% | Max DD: -6.31% | Profit Factor: 3.60 I mean come on? I would die for these values of only over 1000 trades right now, hehe. It might depend on how the test is conducted, in particular on which data. For example, it mentions "5,000-bar synthetic NQ price series with realistic properties." So it appears that Perplexity used AI-generated data...
Oh, absolutely, as said before the backtests and reports are pointers and indications, rather than something to be taken literally. Good pointers, however.
  • #656
  • Nov 16, 2025 6:22am Nov 16, 2025 6:22am
Quoting medici
{quote} Oh, absolutely, as said before the backtests and reports are pointers and indications, rather than something to be taken literally. Good pointers, however.
Yes. I've just started to explore Perplexity a bit. Perplexity Labs, in particular, seems interesting. I assume you used this to create these kinds of PDF reports?

Quote
Perplexity Labs is a project-based AI tool that goes beyond simple information retrieval to create complex outputs like reports, dashboards, spreadsheets, and even mini-web apps.
  • #657
  • Nov 16, 2025 6:54am Nov 16, 2025 6:54am
Quoting Noisy
{quote} Yes. I've just started to explore Perplexity a bit. Perplexity Labs, in particular, seems interesting. I assume you used this to create these kinds of PDF reports? {quote}
No, Perplexity Research is sufficient for this type of study. Labs is agentic and for bigger projects where the creation of interactive apps and tools, and/or multi-file output are required. Research can also also produce multiple files, but usually less than a handful.
  • #658
  • Nov 16, 2025 7:33am Nov 16, 2025 7:33am
Quoting medici
{quote} No, Perplexity Research is sufficient for this type of study. Labs is agentic and for bigger projects where the creation of interactive apps and tools, and/or multi-file output are required. Research can also also produce multiple files, but usually less than a handful.
Thanks for the tip. I just took a quick look, and ChatGPT seems to be able to perform historical backtests, even apparently on Tick-by-tick data that you provide. It might even be possible to get the data via some API by now which is then used by the AI directly for its backtesting. I've pretty much avoided AI so far, but it's really amazing what's possible these days. Unfortunately, all sorts of errors still occur.

Quote
AI hallucination is when an AI model generates incorrect, nonsensical, or fabricated information while presenting it as factual and confident. It occurs because the AI, particularly a large language model, fills in gaps based on patterns in its training data rather than on verified facts, leading to errors that can range from minor inaccuracies to seriously misleading statements.

I find Perplexity super interesting, especially since it also uses real-time data from the internet, and it's incredibly fast. Google urgently needs to catch up, otherwise they'll soon be in real trouble.

Asking ChatGPT for a Tick-by-tick historical backtest for a trading strategy:

If you want a real historical tick backtest (now we’re getting practical) I can run one for you — here’s what I need and what I’ll deliver:

What I need from you (pick any combination you prefer)

  1. Tick data (CSV, Parquet, or similar) containing at minimum: timestamp, price, volume. If you have bid/ask or order-book levels, include them (that allows more realistic execution).
  2. The strategy definition in executable form OR a clear rule set (e.g., “enter long when price crosses above 50-tick SMA; exit when crosses below 20-tick SMA; fixed size 500 shares; stop loss 0.25%; take profit 0.5%”). If you want, I can implement typical order types: market, limit, IOC, TWAP slicing.
  3. Execution assumptions: commissions (per trade or per share), slippage model (fixed ticks, fractional slippage, or simulate order book), order size / sizing logic, and whether shorting is allowed.
  4. Time range (all data or a date window), timezone preference (your account timezone is Europe/Berlin — I’ll use that unless you say otherwise).
  5. Desired outputs: equity curve, trades CSV, P&L per trade, metrics (CAGR, Sharpe, Sortino, max drawdown), per-day stats, and/or plots.

What I’ll deliver when given real data

  1. A reproducible Python backtest (tick-level loop) that: ingests your tick feed, applies signals, executes orders at next-tick (or via limit/order-book if you provide book), and records every fill.
  2. Trade log + full equity curve CSVs you can download.
  3. Performance report: total return, annualized/period return, volatility, Sharpe (with your chosen risk-free rate), Sortino, max drawdown, win rate, average P&L per trade, and latency-aware stats if you want.
  4. Visuals: equity curve, P&L histogram, trade markers over price, per-day P&L.
  5. Optional: parameter sweep (walk-forward or grid search) to tune SMA lengths, thresholds, etc.

If you want me to run a backtest now

  1. Upload your tick file(s) here (or paste a small sample) and describe the strategy & execution assumptions (use defaults if you want me to pick sensible ones). I’ll run the backtest and return results in this chat (plots + downloadable CSVs).
  2. If you prefer not to upload data, I can provide a ready-to-run, commented Python script you can run locally (with instructions and options for different execution models).

Which would you like?

  1. Run the backtest now on your tick file (upload file + strategy)?
  2. I’ll provide a ready-to-run script and explain how to adapt it to your data?

(If you want me to run it, just upload the tick CSV and paste the strategy/execution rules — I’ll take it from there and run a tick-level backtest in this session.)

  • #659
  • Nov 16, 2025 7:35am Nov 16, 2025 7:35am
It seems like you are a beginner/learner when it comes to research and the use of AI for it.

I thought that I'd give you some more substance on that, so I asked Perplexity to produce a Strategic Framework for Using Perplexity in Futures Day Trading Strategy Development. Attached, 21 pages.

As that's a bit lengthy I asked it to produce an article suitable for publication in Futures Magazine, discussing the strategic framework for would be professional users. Also attached. Maybe the place to begin.
Attached File(s)
File Type: pdf strategic framework.pdf   433 KB | 459 downloads
File Type: pdf perplexity-for-traders-article.pdf   150 KB | 294 downloads
  • #660
  • Nov 16, 2025 7:38am Nov 16, 2025 7:38am
Quoting Noisy
{quote} Thanks for the tip. I just took a quick look, and ChatGPT seems to be able to perform historical backtests, even apparently on Tick-by-tick data that you provide. It might even be possible to get the data via some API by now which is then used by the AI directly for its backtesting. I've pretty much avoided AI so far, but it's really amazing what's possible these days. Unfortunately, all sorts of errors still occur. {quote} I find Perplexity super interesting, especially since it also uses real-time data from the internet, and it's incredibly...
Oh, but ChatGPT assumes that you are providing the tick data... You can upload that to Perplexity too, and ask Perplexity Labs to create an interactive backtesting app for you, for example.
2 traders viewing now