Trend Engine: AI-Powered News & Trends

I’m not here to play the guru or sell a training. Nor to reveal all my findings as it would results in the alpha to vanish quickly. For info, i’m trading live with it since december 2024 and went from 15 to 100ke. But I wanted to share a few takeways:

\- I am only using market data. And not even live ones. I use the low-tier polygon offer to get the data I need before I position myself.

\- I use simple rule-based approach with a few rules in the style “if X1 > (or <) t1 and X2 > (or <) t2 and etc ” to filter out tickers of interest.

\- I only buy stocks and play with all my stack all in. To avoid bad surprise (because it can be very volatile), I try to diversify with at least 4-6 stocks. The more my stack grow, the less I am restrictive with my conditions. This in order to avoid incredibly bad draw-down on a single stock.

\- I use a fix entry point and a fix exit point. Even if it is not optimal, in average, it is not so bad and it remove a lot of overhead and simplify a lot my backtest. No stop-loss, no take profit. I am usually in very volatile tickers, its not rare to stop loss hunt before a break-out. Fixing myself a window is a good way to not be biased by market manipulations.

\- I tryied to apply machine learning to my features. Funny enough, anything I attempted around that degraded at best the ROI of my backtest, proving, one more time, that good old handmade rules outperform everything.

\- How did I came up with my rules ? Actually… simply by observation and logic. My strategy is actually very simple on the paper and very logic once you think about it, and I am surprise it still work so well.

\- In order to tests my ideas, I have a very straight forward methodology. I have stored in .parquet files the tickers (including delisted one – to avoid survivor bias) in .parquet files from 2003+ (I used 1 month of high tier polygon to get all the historical data I needed). All the daily data fit in memory (I have 64gb CPU, bought on purpose). Then I filter out based on my long term filters that don’t need more than a daily granularity, and then I iterate to build the lower order features. My stack is simply python + pandas/numpy. I should use polar which is more optimized for the exercise, but I am a bit lazy to learn…

\- Having a backtest on which I can rely has been the most important thing. When you finally have a strategy that work, this is the ultimate thing that will help you in hard moments (and there will be). In my case, after an incredible x2.5 in december 2024, I got horrible jan/feb months where I lost about half of my gains. But that was fine, because I knew this kind of scenario could happened thanks to my backtest. (Actually in my backtest, I am seeing a maximum period of time without gain of 1 year over the full period (2003 +).

\- Despite of the simplicity of what I am doing, my worst enemy remain myself. I actually throwed 10’s of k simply because I tryied to deviate from my strategy or take revenge trading on bad days.

\- In order to calculate my average reward, I am not using a simple algebric average. Due to beta-slippage, and particularly with my highly volatile strategy, big drawdowns can really biased the results. Instead, I am using the geometrical average which perfectly account for those drawdowns ( exp ( mean ( log( gains + 1) ) ) -1

\- In term of visualization: I like to calculate the cumulative sum of the log of the gains over time. This is a very nice way to see breaks of trends. In my case: it showed that my arbitrage is actually improving over time, particularly since the covid. My main assumption is that this is due to more retails getting involve in trading.

I think that’s all I had to share. Feel free to ask questions, I’ll answer if I could.

Image: my live gains on IBKR. I used to be on another broker before which explains the lower % compare to what I announced earlier.

Hey everyone,

NeurIPS 2025 reviews should be dropping soon (July 24th AoE), and I thought it might be a good idea to start a thread where we can share our thoughts, experiences, and reactions.

Feel free to post your initial impressions, any surprises (good or bad), questions about rebuttals, or just how you’re feeling about the process this year. Whether it’s your first submission or your tenth, you’re not alone in the rollercoaster.

Let’s keep things constructive and supportive. Good luck to all!

I recently ordered a car dashboard camera from a suspicious Shopify store ([https://eqhin1-7n.myshopify.com](https://eqhin1-7n.myshopify.com)) which initially showed a return policy. The product was available via Cash on Delivery (COD), so I thought it was safe.

🚨 Instead of what I ordered, I received two unmarked bottles filled with unknown liquid. There was no invoice, no proper packaging, and zero communication from their side.

After delivery:

– The store **blocked my IP address**, so I couldn’t access my order page anymore.
– **No return or refund** option was available even though it was promised earlier.
– The liquid contents are sketchy and **potentially dangerous**.

I’ve already submitted a fraud report to Shopify with all the details, including the store link and suspicious activity.

👉 If anyone else is a victim or sees this same pattern, please report the store to Shopify [here](https://www.shopify.com/legal/report-aup-violation) to help get it taken down.

Stay safe and don’t trust random Shopify stores offering COD unless verified.
Proof – https://drive.google.com/drive/folders/1H6IzJsuMFYn33YqaGbFlzNoYWcRg1s09

I didn’t buy Ethereum in 2015. I wish I had.
What I did instead was spend years observing it watching it grow from an idea into infrastructure. By 2018, I was already involved in the markets, but mostly traditional ones. My job at the time gave me exposure to how institutions think. And while most dismissed crypto, I kept asking myself: What if Ethereum is misunderstood the same way the early internet was? I didn’t go all in. I started slow with money I could afford to hold for years. But as I saw the developer activity, the emergence of DeFi, and eventually NFTs, my conviction grew. Not because of hype, but because of network effect. Ethereum wasn’t just a coin it was a platform. Today, ETH is my core position in crypto. I hold it staked, I use it onchain, and I believe in its long-term role as digital infrastructure. It’s not perfect. Nothing is. But conviction, in investing, is about having a framework you trust enough to stick with through the noise.

Would love to hear:
What gave you conviction in ETH (or crypto in general)?
What would it take for you to hold for the next 5+ years?

# Experts support Massachusetts bill to ban weaponized robots – Robotics experts testified at the Massachusetts State House last week in support of legislation promoting the safe, ethical use of robotics statewide.

Hello , available to supply you vintage/recent tees , streetwear… if you re around europe feel free to DM for more infos

While it’s clear that tone, reasoning, and compromise have their merits, partisanship often does the opposite. The study highlighted how sticking rigidly to partisan lines tends to derail conversations and entrench divisions. This is a crucial insight: people didn’t change their minds, but by moving away from partisanship, they managed to create a more empathetic and understanding atmosphere.

### Final Thoughts

These findings have significant implications for anyone looking to engage in better political conversations. By focusing on tone, reasoning, and compromise, we can transform how we discuss politics, leading to more respectful and constructive interactions. While minds may not change overnight, the improvement in conversation quality can lay the groundwork for future understanding and collaboration.

Incorporating these elements into our dialogues might not dissolve all our differences, but it certainly paves the way for more meaningful conversations. Next time you find yourself in a political discussion, remember to keep it calm, reasoned, and open to compromise. It might just change the way you see the world – and how the world sees you.

I KNOW people aren’t going to post their working algos online. I was curious if there were examples of full systems online. Like I said they could be total failures from a strategy perspective. Basically just trying to look at the general structure of what a full system might look like.