Whycliffes @Whycliffes

🚨BREAKING: A new book for Quant Traders & ML Researchers ML for Trading 3rd Edition Start reading here:

In just a few days I built a fully functional options backtesting platform for an options ranker, connectivity to Portfolio123 models and genetic optimizers like MT5. Guys...this is really mind-blowing what you can do!

When I started trading, I was completely lost. Too many assets. Too many strategies. Too many people telling me different things. I didn't know where to begin. I tried everything. Lost 30% of my first €1,000. Felt like the market was designed to take my money. Then a friend named Juan took me by the hand. He walked me through the fundamentals of quantitative finance, showed me what actually matters, and helped me cut through the noise. Thanks to him I discovered @P123Finance . And that changed the trajectory of my career. Looking back, this is the roadmap I wish someone had given me on day one: Step 1: Learn the basics. YouTube, blogs, papers. Understand what assets are, how markets work, and what drives returns. Step 2: Pick ONE asset class. Stocks, bonds, crypto, whatever pulls you. But just one. The person who tries to master everything masters nothing. Step 3: Go deep. Move from "I know what momentum is" to "I can build a momentum strategy and explain why it works." This is where most people get stuck. Step 4: Backtest properly. Use a platform or build your own infrastructure in code. But test your ideas with real data before risking real money. Step 5: Risk management. A great strategy without risk controls is a ticking bomb. Position sizing, drawdown limits, hedging. This is what separates surviving from thriving. Step 6: Go live. Deploy your strategy in automatic or semi-automatic mode. Monitor performance. Track signal quality. Iterate. The hardest part isn't any single step. It's doing them in the right order without someone guiding you. I was lucky. I had Juan. If you're interested in systematic equities, my P123 coaching is designed to accelerate steps 3 through 6. You don't have to figure it out alone. If you want the full breakdown, just let me know in the comments and I'll send it to you.

@gladstein @LynAldenContact OK

I'm trialing a new concept of generating new factors/features for P123 using Symbolic Regression with Python and Julia as math engine. It's a bit fiddly to work with Julia but I'm getting very promising results. First factor created with out of sample results. Yes, one factor only! Next is to try to take any none linear factor/feature and straighten them out to be monotonic, which is in theory should be superior for ranking system.

keyboard_arrow_left Previous keyboard_arrow_right Next

We spoke to diplomats, advisers, scholars, experts and current and former officials in China. Almost all of them see the war in Iran as a grave American error. Register for free to read why: econ.st/416vZDK

real 💀

There is a popular belief that buying on fear and selling on greed (contrarian) is profitable. My back test proves that historically that has not been true. Below is the result of holding SPY when Fear and Green index is below 15... Fear and Green index is 9 today.

EXTREAM FEAR... still. Stay safe next week.

x.com/i/article/2038…

Fukuyama: "The world has become a very dangerous place, because the most powerful country is under the control of a 10 year old boy..." Frank isn't holding back. We're all here for it.

"The US is strong, Europe is weak" You hear it everywhere: in the US national defense strategy, in speeches by Trump, Bessent, Lutnick, etc., in Davos this week. You hear it everywhere but it's a myth. The US has not grown faster than Europe since 2010, not by one iota. 🧵

Since the Cold War, successive US governments have gaslit Americans into thinking Europeans get free healthcare, free education and decent parental leave, while Americans don’t, because the US pays for Europe’s defence. It’s a myth that also helps them justify 750+ US bases worldwide. Europe spends $550bn a year on defence, more than China and Russia combined. Even without the US, Europe is the most heavily defended continent on Earth. US bases in Europe cost approx 3.5% of the US annual defence budget and deliver outsized strategic power. That isn’t charity. It’s central to US strategy.

@travelingflying

The US doesn’t “pay for Europe's defence”. Of a $968bn global defence budget, only $33bn is spent in Europe, mostly on US bases which are not NATO bases. The actual US funding of NATO is approx $600m a year, about the same as Germany. The myth dies on contact with arithmetic.

Most foreign policy issues are difficult and complicated. Greenland isn’t one of them. Let’s have a look at seven points: 1. The U.S. needs Greenland for its own defense - Golden Dome, radars, basing. ▶️ The United States can do virtually anything it'd like in Greenland, security-wise, without taking possession of it. The 1951 Greenland Defense Agreement, which was renewed in 2004, allows the United States to build bases there, station troops, and more. 2. Greenland is about to fall into the hands of Russia and China, and the U.S. can’t let that happen. ▶️ The U.S. once had 10,000 U.S. troops in Greenland; now there are around 200. If there is an imminent threat of Chinese or Russian takeover (there isn’t), perhaps start by increasing that number? 3. Russian and Chinese ships are swarming Greenland and the Danes can’t fend them off. ▶️ If Russian and Chinese ships are really menacing the island, the U.S. Navy could sail around it right now en masse. It isn’t. 4. The U.S. needs to own Greenland because "you don't defend leases." Even if Denmark allows full access, there’s a difference between owning and renting. ▶️ This is the No One Washes a Rental Car theory of international relations. In reality, the United States is committed to defending many allies whose territory it does not own. Trump himself defended Israel just last year. The whole point of alliances is mutual defense of one another's territory. That doesn't require seizing it.  5. The Danes are bad allies, so they should hand over Greenland. ▶️ Denmark has been a model ally. Not so long ago, Danes fought for America’s defense rather than the other way around. Among 40-plus allies and partners in Afghanistan, Denmark lost the most soldiers as a percentage of its population. Our allies defended the U.S., which, by the way, none of them owns. 6. This is the new Manifest Destiny. We’re an expansionist, frontier people. Greenland should, one way or the other, join the ever-growing Republic. ▶️ The post-1945 order is predicated a prohibition against conquest. Countries don’t acquire the territory of another without their consent. Iraq doesn’t get Kuwait, Russia can’t have Ukraine, Canada won’t be the 51st state, and the U.S. doesn’t compel Greenland to join. We’ve seen a world before in which conquest abounds. It’s the law of the jungle. 7. This isn’t real, just some fun administration trolling of the ever-nervous Europeans. ▶️ It is at a minimum a major distraction from real issues the transatlantic allies should focus on: Russia, Ukraine, Iran, China. Prodding allies to distrust our word and intentions does not amount to good policy. Most foreign policy issues are difficult and complicated. Greenland isn’t one of them. The sooner this manufactured crisis fades, the better.

Quants! Am I missing any analyzing method? The toolkit is built to analyzes datasets of financial features - Pre M.L. Training I made a much simpler toolkit earlier that has been very helpful for me building AI models, expect this one to be extremely helpful.

@GfI_Himmelreich My first none linear live (real money) is reaching 100% (in SEK). Without the USD slaughter this year it would have been even better.

AI Factor Portfolio Strategies, in general --> > 200 Portfolio Strategies (always skip and date on the feature level) > 100 AI Factor Base Models - all rank & date based Small Cap Strategies behaved well OOS Live (3MRel, 3MTotal) - all ZScore & date based Small Cap Strategies behaved well OOS Live (3MRel, 3MTotal) - all ZScore & Dataset Large Cap Strategies behaved well OOS Live (3MRel, 12MRel, 12MTotal) - all ZScore and Date Large Cap Strategies behaved well OOS Live (3MRel, 12MRel, 12MTotal) The only combo that did o.k. (e.g. underperforming but with less vola against the SPY) but not well was: rank and date on large cap models. The most interesting learnings: - LightGBM and Extratrees do not care about collinearity and also not about lean feature sets (130 - 180 is the sweet spot). Think non linear, a linear regression (for example linear regularized) is something completly different than LightGBM or Extratrees. - Use ranking systems from P123 as your base feature set (about 80 Features) and then add single features from there (Greedy early splits need strong anchors!). - Train wide, then filter smart on the portfolio strategy level, AI Factor Models via ExtraTree or LightGBM will not build a Polynomial and therefore is extremely robust: Y = (1/N) * Σ Treeᵢ(X). Even Portfolios, with below 10 stocks are extremely stable. See example below, OOS Live since 03/21/25. - A model with 179 Features and conservative Hyperparameters will build up to 25.000 Rules in 500 Trees, so forget comparing AI Factor models with traditional ranking systems, it is something completely different. A traditional ranking system with 179 factors is a single, linear model, while an AI model (via LightGBM or ExtraTrees) is a dynamic ensemble of thousands of non-linear decisions. AI Factor is a Ferrari, learning is needed, but if you master it, you have the better tool - Combine ExtraTrees (Buy Rule) / LightGBM (Ranking System) Predictors in one portfolio strategy is the absolute killer so far. - Use Deepseek --> tell it you use @P123Finance --> it knows us well. Describe what you are doing --> it takes up to 5 Minutes, copy screenshots of your AI Factor model to Deepseek (also all the features), tell it you want to capture nonlinearity (otherwise it might sell you OLS Textbook stuff) via AI Factor. Share the results and ask questions about why something is working, copy most important stuff into a word document, or start a blog ;-)