Probabilistic

Uncertainty is everywhere.

[Probabilistic @probabilistic_]

@bookwormengr What prompt did you use to get this result?

[Probabilistic @probabilistic_]

@alexcabral71 Onde esses dados sobre os volumes dos leilões são publicados, @alexcabral71 ?

[Probabilistic @probabilistic_]

@Burgos Essas configs de personalização acabam sendo embutidas no prompt final, então as instruções para o ChatGPT foram marginalment melhores. Ainda assim, a qualidade da resposta do ChatGPT é superior, objetivament melhor. Para os seus casos de uso ele tem sido consistentemente melhor?

[Probabilistic @probabilistic_]

@Burgos Concordo que precisão é mais importante. Melhor uma boa resposta mais demorada, que uma ruim rápida. No entanto, minha dúvida é se o ChatGPT identificou automaticamente que precisaria usar mais compute no problema, apenas com o prompt, sem direcionamento adicional. Foi o caso?

[Probabilistic @probabilistic_]

@Burgos Pergunto pq uma resposta com latência de 20 mins incomodaria muitos usuários, então fiquei curioso como foi a experiência de usuário neste caso.

[Probabilistic @probabilistic_]

@Burgos Muito interessante! Que tipo de configuração foi necessária para o ChatGPT processar e pesquisar por 20 mins? A quantidade de compute utilizada certamente ajudou na resposta, mas fiquei curioso de entender a nuance de como ele tomou a decisão de usar essa quantidade de compute.

[Probabilistic @probabilistic_]

@agpatriota Pode compartilhar o artigo?

[Probabilistic @probabilistic_]

@caglar_ee X is a matrix and β a vector, right?

[Probabilistic @probabilistic_]

@__paleologo Could you list some of his papers that you consider very good/important? Besides this one.

[Probabilistic @probabilistic_]

@choffstein Ok, but in this case, valuation bets could in theory have medium to high breath, as they might be built on a x-sectional basis. So valuation bet -> low breath, not necessarily true. Unless, you assume that these x-sectional bets dont change much over time. What am I missing here?

[Probabilistic @probabilistic_]

@TheOneRealPK @__paleologo Awesome, tks a lot! He also made great contributions with standard tweets, replies, etc. Wish we could recover those as well...

[Probabilistic @probabilistic_]

@cryptohound666 @macroarb @NaijaTechBro @0xfdf DMs closed indeed

[Probabilistic @probabilistic_]

@pedma7 @cryptohound666 @0xfdf Can you share these prints pls?

[Probabilistic @probabilistic_]

@FormulaDeltaOne Do you apply convex optimization besides least squares linreg in any context of your job?

[Probabilistic @probabilistic_]

@saplash_i_am @0xFaust12 @0xfdf @christinaqi Thought that SAPI, programmatically, was the same as DAPI with BLPAPI. I mean, in terms of functions, classes, etc, all the same but with SAPI pointing to a separate server instead of a local BBG terminal, as is with DAPI. Isn't this the case?

[Probabilistic @probabilistic_]

@saplash_i_am @0xfdf @christinaqi Which provider do you think offers the best cost-benefit for historical adjusted prices (raw prices, corporate actions, newly listed stocks, delisted stocks, etc.) FS or BBG? Personally, dont know much about FS, but pulling hist intraday data from BBG has been pretty frustrating

[Probabilistic @probabilistic_]

@__paleologo @dipeshghimire77 Just stumbled upon this discussion. Would it be too much to ask for a print screen of the response email? The answer seems like something that would be of interest to others practitioners as well. Tks in advance.

[Probabilistic @probabilistic_]

@__paleologo @macrocephalopod Giving additional reference, @macrocephalopod did a quick analysis using this combo some time ago. Would like to hear your thoughts on how you would tackle this kind of problem, if I may.

https://t.co/d1XjWMTtka

[cephalopodshop @macrocephalopod]

Instead we can do 100,000 bootstrap resamples of the dataset and fit a regression to each one, to get an approximate distribution of the regression coefficients.

That looks like this (a regression slope of 10 would mean you expect to make $10 additional profit when the starting VIX level increases by 1 point).

[Probabilistic @probabilistic_]

@__paleologo @macrocephalopod Sorry, missed this reply. Let me be more clear, I was referring to use cases where you'd want estimate the uncertainty of coeffs of linreg, and standard error is useless because of heavy tails/skewed dist. Then I see the combo bootstrap + OLS quite often being used.