Alpha Insights: Factor Performance Update & What's Coming Next

December 2025

Alpha Insights Ratings Continue to Deliver

As many of you know, since launching Alpha Insights in August, the product itself has evolved tremendously.

We have gone from an AI platform that maintains a growing coverage of stocks - starting with ASX-listed companies before branching out to much larger markets like the US.

And for those that have been following me since the start of the year, you were privy to the many iterations and improvements made to my framework along the way, the change in direction of Alpha Insights' strategy, the endless platform updates/redesigns. And... both the good, and also the horribly incorrect outputs generated by my AI process.

Glad to say now, after thousands and thousands of iterating, and testing the framework across the February, May, August, and November reporting season across - well... 300+ companies from all kinds of industries/sectors - I have. well... more confidence in the framework (never say never, right?)

So, with the framework now properly bedded down with a much higher degree of stability, we are finally moving onto the good stuff.

The Crème de la Crème. [First one anyways, got a few more surprises up my sleeve, pending direction of company].

Alpha Insights' proprietary ratings.

If you had missed our first article on this ratings system, you can read more here - https://www.alphainsights.com.au/analystcorner/alpha/.

This update is a continuation from that first note.

While the first note highlighted a longer time horizon stretching back to the start of the year, from here on, we will be referencing a set/established rating system that begins in October. And given the sell-off we have seen in recent times, no better time to really stress test this framework.

Over the past 10 weeks, the Alpha Insights proprietary rating system was tracked across 466 individual stock assessments, covering 345 ASX-listed companies.

This would help to reinforce whether what we've been building - a sophisticated, systematic, data-driven institutional analysis framework (lol, a rather descriptive framework i'd say), that identifies winners and avoids losers.

Findings:

For a period where the broader market fell by 3.13%, our BUY-rated stocks generated +2.84% alpha — meaning they outperformed the market by a meaningful margin. Meanwhile, our SELL-rated stocks underperformed, exactly as the system predicted.

The spread between our best and worst ratings? +4.70%.

While our testing duration may be shorter than I would have hoped for, given the sample size that we are working with, these results are promising.

Why a 10-Week Test? Because We Don't Believe in Backtesting AI

This is a live, forward test. Every rating in this analysis was generated before the price movement occurred, then tracked in real-time. No hindsight. No curve-fitting. No cherry-picking.

We take this approach because we're sceptical of backtesting AI-driven strategies.

Why?

AI models are trained on vast amounts of historical data — including news, market events, analyst commentary, and price movements. When you backtest an AI strategy against historical periods, there's an inherent risk of that the model already contains embedded knowledge of what the outcomes are.

So. An AI might appear to predict the past brilliantly, but that's not really prediction. More like memory.

Backtests of AI strategies can be fundamentally contaminated by training data leakage.

The only honest way to validate an AI-driven approach is to test it on data the model couldn't possibly have seen: the future. That's precisely what we've done. Every assessment in this report was made in real-time, with outcomes measured as they unfolded.

Ten weeks of live results may seem modest compared to a 20-year backtest, but we'd argue it's far more meaningful.

What the Numbers Tell Us

We measure our factors against a simple question: does the rating actually predict what happens next?

10 out of 12 factors performed exactly as expected.

When we combine our top factors — specifically, stocks rated BUY that are also Undervalued — the results strengthen further: +3.17% alpha with a win rate of 44% (compared to the market's 36.7%).

In plain terms: more of these stocks finished in the green, and they lost less (or gained more) than the average stock.

What This Means for You

We recognise that Alpha Insights, in its current form, provides detailed company-by-company analysis.

Valuable?

Absolutely. But I understand. Synthesising all that information quickly into something actionable isn't always straightforward.

There is a lot of content is onto the Alpha Insights platform. (and even much more on my database...haha)

Hence, several new features are being developed as we speak.

Coming Soon: Curated Stock Lists & Model Portfolio

1. Factor-Screened Stock Lists A regularly updated list of companies that meet our top-performing rating criteria — the same combinations that have demonstrated consistent alpha generation.
2. Model Portfolio A composite portfolio built from our proprietary factors, designed with four objectives:
   • Alpha generation — targeting outperformance against the market
   • Higher win rates — more positions finishing positive than negative
   • Lower volatility — a smoother ride through market turbulence
   • Sufficient liquidity — ensuring positions can be entered and exited without difficulty

Both features are currently being finalised, and will represent the practical application of everything we've validated. Realistically, I should be able to deliver this by January 2026.

The Bigger Picture

This milestone represents a significant step forward in how you can utilise Alpha Insights. Rather than sifting through individual reports and piecing together the puzzle yourself, we're building the tools to do that synthesis for you — grounded in data, tested over time, and designed for real-world application.

The depth and variety of our AI-driven analysis remains. What's changing is how efficiently you can access the conclusions that matter most.

We'll share more details as development progresses. In the meantime, the factor performance speaks for itself: the system works, and we're making it work harder for you.

Questions or feedback? We'd love to hear from you.

Alpha Insights | December 2025