I use market-neutrality in crypto to cleanly capture alpha with min. to no drawdowns//variance; targeting minimum 10 sortino -per strat, actual 20 sortino no MDDD (or min. daily max. DD). The bonds intraday MR E.E. (Electronic Eye) offer the main allocation while my total accumulation strat port/versioning for cryptoassets + liq. reserves (vs fiat) offer the total customized package firm offer! ML families (Google Colab//Drive, same shared bt sim working codebase; same shared bt sim opts. out of working sub-folder, my I.P. original methodology), systematic statistical arbitrage & technologist approach (ex. colo NY-5) as single blended mixture models for bt sim as same shared proxy to live trading! Looking to collab (O.S.S. style in closed-source quant (100-1000x improvement in same shared codebases)) with other limited capacity traders ((30-40% PnL split//separate -per strat mgmt)) that are looking for true scalability w/ sig. CF's, non-traditional firms//tradfi-to-non-tradfi PM's looking to move, and a partner in me!
Despite open-sourcing with (still shared same given single email) open pipe to PM's/Traders/Partners/Allocators, no change whatsoever in same shared limited capacity intraday custom mean-reversion bonds with 0 oversight//resources over period of nearly 8+ months building it out myself; so I can port bonds MR to crypto MOMO. Simply edit params.txt, tickers.txt, and algo pre-loaded key "MR" to "MOMO" (or few functionalized locs for returns, vol space) to port my custom intraday mean-reversion bonds algos to crypto momentum (ex. Blockworks ETF publicly available list <> algo sniping or HFT to UHFT Electronic Eye (E.E.) on single blended tradable signal pre-created//pre-generated rolling alpha distributional params algo key 'cumulative residuals' per symbol aka 'SYMBRES' algo key). Original I.P. protection (root sys. ((original creator//original works)) over usr.) open-sourcing in hardest corner of marketable edge in quant trading (vs. simply removing imports at top of program!), same shared codebases for 100-1000x actual improvement in underlying given ("chicken-scratch shit .py to .ipynb notebooks") reverse-engineered ML model families to bt sim outputs (same shared backtest sim sub-folder <> backtest simulation//ML model families outputs). No UNAUTH algorithmic execution ((nor AI reveng//auto-prompt for random algo generation put into AI prompt//model (e.g. o1-o3 model, claude 3.5 sonnet, cursor.AI,...) to guess//execute trades on my underlying main strat shared <> unified algorithmic design documentation system on Notion)): new class-first citizen functionalized live-pythonic functionalized functions live to colo exchange NY-5 (see FunctionalvsOOP README docs here: https://lnkd.in/e5jaauhy) hybridized class-first citizen <> functionalized docstring functions to confidently share codebases for theorized 100-1000x improvement in same shared algorithms (within my main strat algorithmic design documentation on Notion) without any loss in same shared low-capacity MR bonds to port to crypto MOMO.