Dump it Here

[Richard Dale]

@dis4ever - Yep, the blind faith of asking any GenAI to provide any sort of financial "advice" (aka stock selections) is akin to putting your trust in a model that will postdictively change without you knowing because it has newer samples.

It's akin to somebody saying (after the fact) that they would have shorted the market prior to Covid/GFC/Tech Boom etc.

But it also has various "stale" information being incorporated too - not too dissimilar to fundamentals.

This is where it's very important to understand the staleness of the information being presented/used. If you have no visibility of the source data staleness then you're just throwing darts.

None of this is particuarly new - I have a good friend who is at the top tier of managerial level for a major (well top 30 worldwide, actually much higher but I won't say any further) mining firm that, several years ago, said they were incorporating heuristics to interpet verbal mining shift managers reports to identify anomalies versus production levels. Production levels, for the shift, might be with parameters but if they shift manager went "off-script" versus their usual reports, it was a sign that things needed to be looked at from an operational perspective before it became a signficant problem. The "off-script" was a very basic keyword system (eg. total amount of swear wods) intially but moved to sentiment analysis fairly quickly.

Interpretation of earnings calls meetings has also progressed along these lines, but the companies are also cognisant of this too.

Interesting times.

 

[Skate]

@dis4ever - Yep, the blind faith of asking any GenAI to provide any sort of financial "advice" (aka stock selections) is akin to putting your trust in a model that will postdictively change without you knowing because it has newer samples.

It's akin to somebody saying (after the fact) that they would have shorted the market prior to Covid/GFC/Tech Boom etc.

But it also has various "stale" information being incorporated too - not too dissimilar to fundamentals.

This is where it's very important to understand the staleness of the information being presented/used. If you have no visibility of the source data staleness then you're just throwing darts.

None of this is particuarly new - I have a good friend who is at the top tier of managerial level for a major (well top 30 worldwide, actually much higher but I won't say any further) mining firm that, several years ago, said they were incorporating heuristics to interpet verbal mining shift managers reports to identify anomalies versus production levels. Production levels, for the shift, might be with parameters but if they shift manager went "off-script" versus their usual reports, it was a sign that things needed to be looked at from an operational perspective before it became a signficant problem. The "off-script" was a very basic keyword system (eg. total amount of swear wods) intially but moved to sentiment analysis fairly quickly.

Interpretation of earnings calls meetings has also progressed along these lines, but the companies are also cognisant of this too.

Interesting times.

@Richard Dale, you raise some excellent points.

Hindsight bias and staleness of free “AI”data is a significant challenge if you use it in your decision-making process.

IMHO, “AI” should be seen as a supplement to the process rather than a substitute.

Skate.

 

[Skate]

Critical thinking rather than as a definitive solution

I find that “AI” is a great tool when thoughts need answers. The responses it provides help refine and hone the next question, guiding the thought process rather than replacing it.

Skate.

 

[Richard Dale]

IMHO, “AI” should be seen as a supplement to the process rather than a substitute.

Aka "I'm going to supplement my decision process with information from an opaque source with undefined sources with varying degress of unknowable staleness"

Good luck! LOL.

 

[Skate]

Aka "I'm going to supplement my decision process with information from an opaque source with undefined sources with varying degress of unknowable staleness"

Good luck! LOL.

@Richard Dale, you make another great point.

However, the same argument could be made about any gathered information whether from news reports, expert opinions, or even traditional research.

Every source has its own biases, limitations, and degrees of “unknowable staleness.”

The key is to approach all information critically and weigh it appropriately in the decision-making process.

Skate.

 

[Richard Dale]

The key is to approach all information critically and weigh it appropriately in the decision-making process.

You clearly can't make any consistent decisions with your historically unverifiable approach (with motherhood vibes).

Quid pro quo - lol.

 

[divs4ever]

You clearly can't make any consistent decisions with your historically unverifiable approach (with motherhood vibes).

Quid pro quo - lol.

inaccurate data ( or stale data ) has always been problematic in research , but if using AI one might be tempted to ignore the potential inaccuracies in sourced data ,

possible work-arounds are running the AI two of three times but using a different data source for each run and look for disagreement in the results

AI is basically a fancy version of computer chess , you have the current position and the program needs to analyze the probable next moves and ( try to ) select the best foreseeable outcome , BUT how does AI cope with 'black swans ' , maybe AI does better in giving options after an unexpected event ( when the average trader/human is paralyzed with indecision

maybe expecting AI to give you a solution is the wrong application of it , maybe we should be asking it for a range of potential outcomes and options for the next step in each , in theory the best asset of AI is the removal of human emotions in the situation analysis

BTW i have had some nice outcomes using straight instinct ( automatic reaction ) but i am aware others are not so good at that

 

[TimeISmoney]

AI is just a man-made machine; if you use it for the purpose it was intended, you shouldn't have too many issues.

It's like a washing machine, if you just plonk the clothes in there and overload it and don't level it properly from the start it will never do its job properly because you're using it out of its parameters of design.

 

[Skate]

The Dangers and Pitfalls of Coding a Trading Strategy with Limited Experience
Many aspiring traders believe purchasing a charting program like Amibroker and data from Norgate, then coding or buying a strategy is the key to easy profits. However, the reality is far more complex and filled with potential pitfalls. Here are some critical points to consider:

Correct Coding
Ensuring your strategy is coded correctly is essential and it's harder than it sounds. Even a minor error can lead to significant losses. Understanding every aspect of the code functionality is required to avoid costly mistakes. Refining your code should be an ongoing process, as continuous improvement is key to long-term success.

Backtesting Limitations
Backtesting is a valuable tool but has inherent limitations. Historical data offers insights, but it doesn’t guarantee future performance. Market conditions evolve, and what worked in the past may "fail in the future". Approach backtesting with caution, and validate your results across various market scenarios to ensure robustness.

Mindset and Drawdowns
Trading with real money is vastly different from paper trading. Large drawdowns will test your emotional discipline in sticking to your strategy during tough times. Emotional control often determines whether a trader succeeds or fails.

Insights from Experts
@Richard Dale's participation in the “Dump it Here” thread offers a unique opportunity to gain insights from someone with decades of experience. His contributions always provide invaluable insights that offer a valuable framework for traders at any level.

Moving Beyond the Basics
As your experience grows, it’s essential to go beyond simple backtesting. Forward testing with both in-sample and out-of-sample data is crucial for evaluating the robustness of your strategy. Carefully consider the length of your backtest as "shorter tests" may miss key market shifts, while "longer tests" could incorporate outdated conditions.

Skate.

 

[Skate]

Trading versus Investing: A Balancing Act of Risk and Reward
Investing and trading offer two distinct paths as one provides steady, consistent returns with minimal involvement, while the other promises potentially higher rewards at the cost of greater risk and volatility.

Over the past year
Investing has proven beneficial, yielding returns with little effort or excitement. It’s a calm, passive approach that can sometimes feel like being a spectator, watching from the sidelines rather than actively participating in the game.

As the new year approaches
I’m at a crossroads - should I stick with my current investment portfolio, enjoying the stability it offers, or dive back into trading, where the thrill of the market comes with both higher stakes and the potential for greater rewards?

At times good enough is good enough
I've refined my approach and the initial results aren't too shabby.

Backtest
Using Amibroker and Norgate "Current & Past" Data.






Skate.

 

[divs4ever]

the temptation ( for some ) is to reduce the holdings ( some or all ) in the months where the falls were 5% or more

there would be some urging you to buy ( something ) in those two DOWN months , now sure buying something ( carefully ) in May 2023 might have improve the returns ( say be using dividends to boost selected discounted socks ) but what of June 2024 , although the drop was worse , the aggregate holdings would have fallen from a higher peak ( and might not have been bargains

selling in May and returning in September might not have been a great strategy in 2023 and 2024 with ASX listed stocks

but with investing all you can do is buy , assess and adjust if needed ( or desired )