Systematic Trading Report 2025 | Cannot Be True… But It Is
Trend isn’t dead. Why machine learhing is broken. What about AI?
HedgeNordic just dropped their Systematic Report 2025 and it’s the kind of institutional brain fuel we actually care about.
Just real insight into what’s working, what’s failing, and how serious quant desks are adapting in the trenches.
I read the whole thing. Here’s the good stuff.
Paper - https://hedgenordic.com/2025/06/report-systematic-strategies-and-quant-trading-2025/
Systematic Report 2025
What’s Actually Going On in the Quant World?
1. Trend Following Is Alive, But You’re Probably Doing It Wrong
Trend isn't dead. It’s just not in equities.
“Trends in commodities and currencies were persistent… equities were more range-bound, suppressing alpha.”
My take:
Everyone wants to trend-follow SPY like it’s 2013. Meanwhile, cocoa, oil, gold, and yen trends are printing money, but only if you’re looking outside the standard “retail ETF” universe.
If your trend model isn’t multi-asset, you’re playing chess with half a board.
2. The Short Vol Addiction Is Alive and Well
Even after all the face-ripping lessons from 2018 and 2020, short vol strategies snuck back into portfolios like an ex who promises they’ve changed.
“Short vol trades crept back into portfolios as the VIX stabilized… despite persistent tail risk.”
My take:
People will always reach for yield. It’s human. But if your edge only works in low-vol chop, you’re not trading, you’re gambling on peace. Tail risk is real, and it’s not priced in until you’re already bleeding.
3. Macro Made a Comeback (and It’s Not Leaving)
One of the clearest themes: 2023 was macro’s year. Rate hikes, inflation whiplash, and geopolitical spiciness made macro quant look like geniuses again.
“Macro factor exposures delivered outsized returns… particularly during policy regime shifts.”
My take:
This is a big one. If you’re ignoring macro in a post-zero-rate world, you’re trading like it’s 2012. Price action alone won’t save you anymore, integrate macro filters, or get smoked by guys who do.
4. Sharpe Ratios Are a Lie (Sort of)
The industry still obsesses over Sharpe, but the report puts it on blast: past Sharpe does not equal future comfort.
“CTAs produced a Sharpe of ~1.4 in 2022, and yet 2023 disappointed many allocators… Sharpe doesn’t reflect drawdown behavior or crowdedness.”
5. Machine Learning Still Has an Overfitting Problem
They tested a ton of ML techniques. Result? Most ML alpha dies in production.
“Machine learning signals offered incremental improvement, but often lacked robustness or interpretability.”
My take:
Still too much curve-fitting. Most “AI” in quant right now is dressed-up brute force on lagging features. Unless you’re doing real structural modeling (e.g. non-stationary, regime-aware, ensemble-learning)... it’s just backtest theater.
6. Execution Quality Is the Hidden Alpha Killer
This one hit hard. It wasn’t the models that failed, it was the fills.
“Slippage, especially in illiquid futures and EM assets, materially degraded strategy performance in 2023.”
My take:
Nobody brags about execution, but it’s where 90% of live-trading edge disappears. Stop obsessing over signal 5th decimals. Fix your fills. If your model’s alpha is within your expected slippage band, congrats, you built a glorified randomizer.
7. Liquidity Risk: The Elephant in Every Backtest
“Liquidity risk was underappreciated… several funds had to de-gross aggressively during March volatility.”
My take:
The drawdown doesn’t come from the model being wrong, it comes from being right with too much size. When liquidity vanishes, even good ideas get margin-called. If you’re not stress-testing for liquidity gaps, you’re larping as a quant.
Summary: What the Smartest Desks Are Actually Doing
Report confirms what a lot of high-level traders already suspect:
✅ Trend still worksm, but only in the right places (yup, sounds true)
✅ Macro isn’t optional anymore
❌ ML alpha is overhyped (and overfit)
❌ Execution costs are silently killing your live performance
❌ Sharpe is not a complete picture, especially in regime shifts
🧠 Real edge is now in dynamic position sizing, vol-awareness, and robust execution infrastructure, not just signal engineering