@joshuaaalampour: Not financial advice. I wanted to create this to complement my neural net strategy from about a month ago. I needed something with weak correlation and so a topological graph diffusion arbitrage algorithm was perfect. I have visions of combining different strats together to form an ensemble, where the capital allocation is dictated by a meta classifier of some sort. I want to try and tackle that next, but it’s gonna be really hard. But also really elegant!! Here are my backtest results (all net of costs, 3 yr rolling walk forward out of sample with purged and embargoed tuning, and t+1 execution): 2020: +10.86% CAGR, 2052 trades, 1.09 Sharpe, 9.62% max DD 2021: +12.91% CAGR, 2397 trades, 1.38 Sharpe, 8.3% max DD 2022: +16.24% CAGR, 2701 trades, 1.31 Sharpe, 11.24% max DD 2023: +13.08% CAGR, 2488 trades, 1.34 Sharpe, 10.51% max DD 2024: +15.47% CAGR, 2579 trades, 1.29 Sharpe, 9.69% max DD Research preview only. #quant #math

Xora :

This sounds so cool how do i even begin to understand what im hearing

mrmatsson11 :

Risk to return? You did a discounted cashflow for every stock?

𝕝𝕝𝕖𝕤𝕝𝕚𝕖𝕖𝕖𝕖 :

saving this in case i wanna get into some new rabbit hole of knowledge

Waffles :

Interesting work but I have some fundamental questions. When you run Laplacian diffusion on returns, what exactly is diffusing and what’s your baseline for the residuals? A residual just tells you idiosyncratic return vs neighbors, but that’s not a mispricing without an underlying pricing model. What defines ‘correct’ price here?

Logan Nissen :

Dude that’s crazy how did you learn

didyousoup :

Underperforms Nancy Pelosi tho

Lucas :

DeVoDetEne :

did you test it with fees?

pirateangel13 :

all that to lose to spy 😭😭

PeaBrane :

if it were profitable you wouldn't be telling us this

Patch :

3Blue1Brown would be proud

keshaun :

just for me to make 50% shorting a biotech company that’s diluting their shares

Aus Styled :

I want to understand this so bad

PenithThlammer :

How much backtesting have you done? Specifically with Monte Carlo methods.

Enthusiast🫆⚙️ :

i did this too

1474674 :

Im not your target audience, just wanted to say that I like your outfit.

umami :

Overfitting

Bowen :

Wait, I am amazed that you found an actual use for this stuff. Professor Takashi Owada at Purdue pursues maybe a bit more theoretical works in this topic area if you want to read more. The field in general is relatively niche but I am glad I finally made it to this side of TikTok.

Pausch :

how did you make the animations?

Ember :

What did you do to avoid overfitting?

fearow69 :

Is this better than multilinear regression?

matt_evans5 :

Dude, that’s actually insane, I’d love to learn more about this

Shaurya :

Do you think I could get into quant with a finance and CS double major (bs + masters)?

Rasmus Ahrenkiel :

Whaat in the matrix are you saying

munkeren1 :

Yeah so if you could just email me that, would be greatly appreciated.