guillem

Weirdly, getting a real job is an excellent decision if you want to be a full-time trader long-term, it solves many psychological problems. Also, ignore Twitter. 90 %+ of "elite" traders here are either fake (or we don't know the whole picture) or lucky degenerate gamblers who will quietly blow up in a couple of years.

@daytradingzoo congrats man! I might just steal your strategy of ditching X lol

I blame this result of not being on X anymore every single day. Social media in excess is just toxic. See you next week

@hackertrader @paraholik @cxxxrbon whatever... was just saying because imho in those cases is where lookahead bias that other users previously pointed out might be. cheers

@guillemservera @paraholik @cxxxrbon I don't give away the specifics of how I short...

Today I created this backtest that shorts small caps without writing a single line of code using Spikeet + Cursor. Here's how I did it: 👇

@hackertrader @paraholik @cxxxrbon where did you short those things that were 50%+ anywhere in the day? when it hit that 50%?

@paraholik @cxxxrbon I also shorted things that where 50%+ anywhere in the day (as measured from yest close), still don't see what I'm missing

@JoachimMo1985 @amphilomath is the API better than IBKR or Schwab?

@amphilomath Yep, nice API no commissions

I am creating my own order and trade management application for Alpaca using AI+python. Orderclerk from RealTest is my role model.

@CYCLEFUNDCAP qué usas para backtestear estas cosas?

SYSTEMATIC 0DTE TRADING 🤠

@SmallCapSmarts @dcbaros which IB API do you use? TWS?

@dcbaros polygon.io for historical, IB for real time

This is the strategy I mostly traded from 2018 to 2022 (fixed size here) 4 years of out-of-sample performance are unbelievably good, it was also pretty scalable too (large caps) Stopped trading it mid-2022 and now thinking to move back to it with a couple of tweaks

@nachio_134 gracias por la respuesta!

@guillemservera No, no necesariamente,en este caso PM high puede tomarse como risk, pero no influye en el setup. Incluso bien podria repetirse con un Breakout de ese PM Highs

Hace dos semanas que no operaba por temas personales. Hoy entre despues de mucho tiempo y fue una fiesta Casi 6$/Share con riesgo de 75c. Win rate del setup 64% en lo que va de 2025.

I have not felt this excited about a product like I am about @cursor_ai in forever. I spent the entire week building scripts, web applications, databases and alike. Things that used to take me a week to solve are now done in 5min. Just building a year’s worth of work within a few days. I’ll be working on these projects for the next months and it is so exciting that I can barely sleep… Worked 8am till 2:30am yesterday, it’s the type of excitement I last felt when I built a small cap database years ago. Type where you can’t stop thinking about even when you close your eyes in bed before going to sleep.

@hackertrader I 100% agree on setup/maintenance. For storage, timescaledb compression (when configured properly) is very good. Speed really depends on the use case. In my own infra I still use parquet for raw market data and to compute features, but keep curated data on postgres

Thanks man!

@mr_data_sniper muy interesante esta gráfica! es PnL acumulado de todos los setups?

La media del PNL en Rs de mis setups con el paso de las horas. Solo 6 días por encima de la media, con varios días malos y solo 1 muy bueno. Algunos de los malos bastante duros, ya que devolvieron bastantes Rs.

+20.5K Mes muy por debajo de las expectativas y con sensaciones bastante malas pese al absoluto. A diferencia del mes pasado, destacaría la inestabilidad de mis setups y el pésimo PF. Abajo dejo unas imágenes para ilustrarlo.

@buxeda_aniol @flash_research sí, la lo ideal es tener datos tuyos reales y sacar una aproximación. Puedes usar @lowcates para tener referencias y en base a eso hacer tus aproximaciones

@guillemservera @flash_research entiendo, con que parametros te basas para saber locates? entiendo que lo mejor seria tener como un año de datos reales propios y hacer una media de locates/trade pero eso es dificil cuando empiezas.

I’m a bit late to the party, but I ran some backtests today with my own engine and wanted to test this. It's definitely not fraud, but it's true that @flash_research approach can make a so-so strategy look better on paper

@buxeda_aniol @flash_research el tema del método de FR es que es muy dependiente de la distribución del PnL. Cuantos más trades alrededor de break even tiene la estrategia, peor es la aproximación... Pero lo cierto es que para estrategias donde el PnL está más polarizado un 10-20% funciona guay

@buxeda_aniol @flash_research es un backtest. El cálculo de comisiones y fees es tal cual sería en real. Locates y slippage es una estimación basada en ciertos parámetros. De hecho, tras mejorar un poco estas aproximaciones ese backtest en concreto es peor

Is a Sharpe ratio of 2 better than a Sharpe ratio of 1? Not necessarily. Sharpe ratios are not comparable, unless we control for 3 variables: T, skew, kurt. The probabilistic Sharpe ratio (PSR) provides a coherent way to compare investment performance: ssrn.com/abstract=18216…

@f1teamboss @flash_research there may be other issues with the tool for sure... but their commissions structure is not that big deal imho if everything else is correct

@guillemservera @flash_research There are so many bugs i don't even want to go into detail because we'd literally be here all day however what I can say is, real pro's have renamed it TRASH 😆

After testing different rates, IMHO a 20% PnL based fee feels more realistic. My Custom Net PnL covers locates, per‑share commissions, routing fees, and slippage on both entries and exits.