Quant Detective

Managing two automated trading algos • Data-driven strategies • Backtests made simple • Python • R • Market breadth tracking • Machine learning

[Quant Detective @QuantDetective]

Breakout strategy had a very bad performance so far in 2023.
For those interested, the exact entry and exit criteria of the backtest are described in the following thread.
https://t.co/rIyIAG8ZeF

[Quant Detective @QuantDetective]

1/💰 Has the decision to NOT trade Kristjan Kullamägi's (@Qullamaggie) inspired breakouts in 2023 been the right call?🤔We'll delve into the NASDAQ and NYSE stocks universe, focusing on the quantitative validation of breakout setups with these high level characteristics:👇(1/n)

[ Q-Cap  @qcapital2020]

This is an excellent interview btw

Nicolai (Norwegian Sovereign Wealth Fund CEO) asks the IBM CEO if AI a bubble

Listen very very carefully to his answer

[Quant Detective @QuantDetective]

LLMs are now often used for data processing or modelling projects. The starting point should always be to visualize data transformations in your mind and have a plan in your preferred language on how to execute the queries.

Only then proceed to use LLMs. Otherwise the chance of a bug or incorrect interpretation of the request by the LLM is high.

Even if successful, it limits ability to interpret or develop over time, if there a zero grasp of how data has been generated, transformed or if it has any biases.

[Gappy (Giuseppe Paleologo) @__paleologo]

I recently bought this book on Amazon, mostly out of curiosity. It goes next to Polya's, Hilbert & Courant, and other classics. It covers the equivalent of the infamous math 55a/b Harvard sequence. Written by *amazing* Russian mathematicians (I studied analysis on Nikolskii's two volumes--awesome). It's 60 years old. Not a vice, these days. It is also unique in that every chapter has a historical context in the historical materialism tradition, very interesting and useful. And it's instructive to read how a great mathematician presents standard material. Some parts (the use of computers in mathematics) are dated. It's ok. Most aren't. And there is a lot of geometric intuition in Russian analysis descriptions! Which I like.

1120 pages. $42 paperback (recommended), $17 kindle.
If you are a high school student, or an autodidact, I can't think of many other books that will make a better intro. Polya sells for $25, Hilbert-Courant $27. They can be found on the internet, but paper is so much better. Per dollar spent, these books, and the hundred hours you'll spend on them, will make you an incommensurably more cultured and happier person.

There is no AI for that.

[Quant Detective @QuantDetective]

@pedma7 Agreed. Reduced immigration + aging population should eventually come into play (see Italy)... but when? 😁

[Quant Detective @QuantDetective]

Market is currently stretched; apparent when compared to its state on Apr 7.

[Quant Detective @QuantDetective]

Approximately 45% of US stocks decreased by more than 5% in one month.
~10% decreased by more than 20%.

[Quant Detective @QuantDetective]

https://t.co/1u8wYprKEn

[Quant Detective @QuantDetective]

Good interview with Tom Lee. Summary points are below. I view several of them as good bets.

- US market resilient. High oil prices actually benefit US stocks since the economy is a net exporter.
- Do not panic sell during geopolitical crises. Timing means missing the few days that create most annual returns.
- Market drawdown might still happen later this year. A new Fed chair will have to handle sticky inflation and weak private credit.
- AI spending is small compared to the broader global economy.
- Software stocks likely bottomed out. Companies still need regular SW instead of building custom AI tools from scratch (eg necessaity to maintain SW).
- Private credit faces trouble. Large traditional banks will survive this cycle and likely benefit.
- Bitcoin and blockchain remain strong inflation hedges. They will serve as the core micro-payment network for future AI agents.

[Quant Detective @QuantDetective]

Approximately 45% of US stocks decreased by more than 5% in one month.
~10% decreased by more than 20%.

[Quant Detective @QuantDetective]

@therobotjames Interesting article!
What are your thoughts on ELO based systems/predictions?

I've developed something very naive ~12 years ago and we'll... it did not perform.

[Quant Detective @QuantDetective]

Significantly more boys start playing chess than girls. The pool of young female players is not large enough to produce exceptionally strong players.

That is the primary effect. Personality traits may play a smaller role, as surprisingly one needs a killer instinct to be good at chess.

[Quant Detective @QuantDetective]

Good article. Widespread productivity increases in expanding economies are unlikely to make things worse on average.

"The bearish loop creates a simplified linear model: AI gets better, businesses reduce headcount and wages, then buying drops, businesses invest in AI again to defend their margins, and the downward cycle repeats. This assumes a completely stagnant economy."

[The Kobeissi Letter @KobeissiLetter]

The bull case for the AI Revolution is abundance.

The “AI will end the world” narrative has gone viral and has quickly become the base case assumption.

However, if you look back at any time in history, innovation has largely yielded a net positive long-term outcome for humanity.

Innovation kills tasks, not people. In the bull case, humans work differently, productivity rises, costs fall, and scarcity declines.

As a result, the world can see less conflict over currently scarce resources, lower poverty rates, and improved living conditions.

A world full of abundance, peace, and efficiency is possible.

Humanity has always prevailed.

[Wall St Engine @wallstengine]

Fed Chair nominee Kevin Warsh on Bitcoin:

"Bitcoin does not make me nervous"

“Bitcoin is the newest coolest software that will allow us to do things we could never have done before."

[Quant Detective @QuantDetective]

@daytradingzoo @peterkto Lol 😂

[Quant Detective @QuantDetective]

@GoshawkTrades Congrats! Best situation to be in - past performance resembling overfitting or some form of data leakage

[Quant Detective @QuantDetective]

I often defend the value of extreme and combat sports.

My argument is that almost any meaningful achievement lies beyond fear (and life there often feels brighter).

Sport becomes a way to train oneself to take calculated risks and think rationally in extreme situations. Risk can break, but avoiding it can do even greater damage to one’s life and potential.

[Quant Detective @QuantDetective]

I am grateful for 2025 - it has been a great year (~25% on my own + borrowed capital, which translates to ~44% on my avg capital).

Ideas that helped me in 2025:
- it is useful to understand how market participants think, but commonly accepted “truths” can be incorrect and harmful to performance depending on the strategy. They should not be trusted blindly, instead analyze the data and don’t be afraid to invert fintwit's wisdoms when backtests support it.
- this leads to a broader point of being comfortable with contrarian positions when models or an investment thesis justify them. The key questions then become how to define what is a contrarian view and at what point it offers a sufficient risk:reward.

[Quant Detective @QuantDetective]

@predict_addict In contrast, engineers are often good modelers- inclined to understand tradeoffs, implications of the data generation process and take acceptable shortcuts.

[Quant Detective @QuantDetective]

Good post. I have many similar experiences and really dislike the single-minded effort to squeeze some extra performance from a dataset.

All too often such people then completely miss business objectives, deliver an unstable model or do not think through how the model will be used by the business.

[Quant Detective @QuantDetective]

@pedma7 @ScottPh77711570 @weaponizedFOMO @therobotjames @ALBACapMgmt 😂