cam

@SectorX_AI Is your allocation amount based on signal only or also its beta/how much it makes up of SPY?

🚨BILLY MAYS HERE FOR RELATIVE ROTATION GRAPHS! Are you TIRED of comparing sectors to an index that's 30% TECH? Do your rotation signals look GREAT on screen but deliver ZERO predictive power? THAT'S BECAUSE YOUR BENCHMARK IS BROKEN! See, with traditional RRGs, you measure XLK against SPY, but SPY is already 1/3rd XLK! You're comparing a sector TO ITSELF and calling it "relative strength"! That's like weighing yourself while holding the scale! And it gets WORSE! XLU is only 2.5% of SPY while XLK is 30%. So every sector has a DIFFERENT amount of self contamination baked into its signal! Your "rotation" isn't rotation, it's just cap weight distortion sloshing around! BUT WAIT, THERE'S MORE! What if I told you there's a method that uses the GROUP ITSELF as the benchmark? Equal weighted. No circular reference. No sector competing against itself. Every asset measured against its REAL peer group. And here's the BEAUTIFUL PART... once your benchmark IS the group, YOU'RE NOT LIMITED TO SECTORS ANYMORE! Crypto? Precious Metals, Energy? Commodities? Doesn't matter! Completely different asset classes, same framework, same clean signal. The method TRAVELS because it doesn't depend on some external index that only exists for one corner of the market! And LOOK, if you INSIST on using an external benchmark for US sectors, at LEAST use RSP, the equal weighted S&P 500. SPY gives you a benchmark where tech is TWELVE TIMES more influential than utilities. RSP gives you a NEUTRAL playing field. That's not even controversial, that's just MATH! But WHY use any external benchmark when the group measures itself PERFECTLY? You get CLEAN signals! You get ACTUAL relative strength! You get an IC that ISN'T ZERO! And you get a framework for ANY asset class, not just the one S&P happened to build a sector taxonomy for! STOP comparing sectors to a benchmark they're trapped inside. The benchmark should be neutral ground. SPY isn't neutral... it has favorites! That's your RRG. You're welcome. Now CALL in the next 10 minutes for ANOTHER GREAT DEAL! 🎪📉🗑️

@SectorX_AI At the moment i tried the sectors that make up SPY

@SectorX_AI By the steady freddy do you mean that you use the sectors which make up $RSP? I tried demeaning % returns each day and going long for top 60% decile bucket but it doesnt seem to beat equal weight. Maybe i am demeaning wrong, should demean price instead of rets?

There's a standard econometrics technique that converts raw group prices into a zero sum relative signal in one step. It's called 'cross sectional demeaning'. I admit, I built a trading system around the concept and didn't know the formal name 🤭 It removes common shocks (beta, macro tides, micro noise) by subtracting the group mean at every tick. What's left is pure relative movement. The trading world never adopted it because it fell between two literatures. Too academic for traders. Too financial for economists. It doesn't sell subscriptions. It replaces them 💥 🧩 🔑 📈

@SectorX_AI Hey sorry i took ages to reply! Yep i think ive got something decent. How do you use absolute rets? I know you use demeaned (rel) rets to allocate, curious about the absolute rets involvement

@HHidden10746 Howdy! curious if you found anything interesting

Love the story this tells, Claude and I spun this dashboard in the last couple days, 5 minute scrub replay, circle sizes are the target allocations in my account. Gauges on the right show IC (correlation between the signal and returns that day) and allocation share of each sleeve. $XLE dominated this week, followed by a trio of crypto miners, looks like a road trip! $LIT Lithium leading Energy. 🔭 🫧 📡 Fidelity account report: +10% Jul-Dec +9% Jan-Mar

@SectorX_AI Thanks! Whats the Y axis on the left? I see theres 6 of them for sectors on this given day

@HHidden10746 However many circles you see on the chart is how many are allocated. The size of the circle is the percent of the current sleeve share each asset is assigned.

@SectorX_AI Are there any articles youd recommend id read or posts to take a look at to get started with rotation systems?

@HHidden10746 yah I made a spreadsheet backtest that works for me, AI makes all the formulas much faster than I could.

@SectorX_AI My dms are open if you wanna chat, i got some cool ideas which would fit well with your model I think

@HHidden10746 I haven't found a good tool to backtest this kind of model, curious what you are using?

@SectorX_AI Thanks!

@HHidden10746 Long only. Tried a few ways, strength only keeps winning. Open project if that's regime or structural.

@SectorX_AI Ah i just tried all of the SPY sector etfs for it

@HHidden10746 I do continuous allocation across the range. The basket matters more though, demeaning needs behavioral coherence to surface anything.

@SectorX_AI I tried relative (demeaned returns against peer group) and combining that with absolute but I couldnt get any good results in the backtest yet. I just divided it into returns when going long > 0.5 relative and short < 0.5 but it seemed bad. Maybe my problem is in the allocation

@SectorX_AI If youd ever want to work on something together hit me up! Im a software dev by profession

@SectorX_AI Did you ever try other factors apart from momentum?

@HHidden10746 thank you! I went so long and finally found something interesting and kind of foundational. To accept what I'm doing puts a mirror up to many assumptions I've had for many years.

@SectorX_AI Got it, thanks! Love your posts by the way, most of the time I read silently haha

@HHidden10746 Thanks for digging in. I demean whatever group makes up each sleeve. The group mean already contains SPY beta (and sector beta, and every other shared driver), one pass removes all common shocks. No rolling windows or estimation errors, the peer group IS the factor model.

@dSystems_ai Also for someone new to machine learning what markets would you recommend for them to try and trade or look for alpha?

@dSystems_ai Is the surprise % in relation to a market consesus forecast? Is your model forecasting market expectations based on seasonality patterns? How do you evaluate if the model is good or not (is the target the EIA survey number?)