Awesome idea @shearman. Something not possible prior to AI.

And I loved the article you linked. Great reminder to ensure a focus on continued learning, not just performance.

I definitely don't apply all or even most of those 34 rules. And there was a lot of repetition of ideas, so the 34 can probably be collapsed down to a dozen or maybe even half a dozen.

One of the key ideas in the article is write down your predictions, and then reflect. That's one of the biggest insights and benefits I've experienced from being part of Strawman community. Being an active member prompts me to craft and document my predictions. I'm then accountable. And I look back on past predictions and reflect on whether I was too bullish or bearish. I can then calibrate future predictions.

I strongly recommend members to post their valuations. Worst case, you get no thumbs, but you still put in the thinking and presented it to the world, and you can reflect back on it. But more often than not you also get some great feedback and ideas from other members. Takes courage, but that's the best way to learn.

@shearman, what it sounds like you are doing is a form of screening, where you put in a bunch of variables that have worked in the past (eg earnings growth, debt levels etc), then screen for the stocks that fit that criteria. qualitiative is more difficult than quantitative data, so many use a rating system for that (subjectivity is usually involved). I've seen criteria that go to hundreds of inputs, often that conflict, and you need stable and accurate coefficients, which can be a challenge. That is, how to weight each variable, and does that change over time? (it does) ok ive been down this path.

of course is depends on how much the future is like the past. You must assess that. It is a well-known and efficient way for people to identify investment candidates, especially if you have a large data set.

it is the base of quant systems that diffuse the risk by holding a large sample to try and eliminate stock speicif risk, but i dont htink that is where you are going here.

so to sum up, screens are useful at directing your effort, finding stock opportunities. The skill comes in the extrapolation of what is relevant.

The only thing i would add, the future is uncertian and it pays to have estimates ( flight path) of what you envisage the future looks like for each investment and track this for compliance over time.

certianly what you are talking about is more useful than someone asking Claude what stock will outperform, I've come across that as well. lol

@Solvetheriddle

Reading your response perhaps Im not sure if I communicated clearly what I was doing

Ill try and elaborate:

Im not not referring to using AI for screening. Although using it for both qualitative screening and monitoring of position thesis are good potential areas I hope get added to various tools out there in the future.

What Im referring to is when you have done some research and have an investment opportunity (to buy X for $Y) - you have to do a valuation, weigh up risks vs opportunity and make a decision. When you do you expect to get Z% return.

i.e. what to buy, at what price

However as humans its hard to improve this skill as

• The outcome often takes years
• Luck can be involved - It can be a good/bad decision but bad/good outcome
• When we look back on why we bought - even with journalling - we forget/remember exactly why we made the decision

If we can get better at this it can make a big difference.

So what if we have a learning tool that presents us with anonymous but real case studies with what ever you want to know at point X in time (from at least 5 years in past)

Then you review, create a hypothesis, valuation and expected return - and then submit it

Then instantly you get graded/feedback:

• What the outcome really was +5 years after (Rev, EPS, Share price, + what happened) + actual return if you bought at that price 5 years before
• Feedback on your reasoning (from the AI) as to what perhaps you missed or could have thought better about

So its like an investing game - and Claude even tracks how you are doing

This is what Ive been playing with

After getting the instant feedback to know what happened on 20 situations over a few hours (vs 5 years) feels like it has the potential to speedup and improve learning

I was just wondering if anyone has tried this as I want to pursue this more but would be good to collaborate with people who are also interested

@shearman ok, so its like a review of what went right and wrong from the initial assumptions, like i assumed eps grwoth of 10% but it was 2% becasue the moat ended up not as strong as i thought, profits disappeared? Am I close? anyway maybe i need to read the article.

btw thats my biggest blunders, betting big on a moat that was breached, or what you know for sure that just aint so

@Solvetheriddle

Here is an example of the interaction output from using Claude in this way.

I wasnt taking too long to think things through before responding - but seeing how well it worked.

This will give you a better idea.

https://claude.ai/share/2bb7bd37-0251-41d8-bab9-547c756124e0

@shearman ok now i think ive got an idea whats going on. i cant disagree with the Substack since it reminded me, in uncomfortable familiarity, of the tome i wrote a few weeks ago on SM. (40 years....). ok i see it appears you are attempting to learn as many variations of investment outcomes without going through the actual real money experience. Fair enough, whatever works, if you can replicate the real life experinces with all that entails, then it is a powerful exercise. good luck with it