/biz/ - Business & Finance

>>16922067
its just rotor wash

>>16922615
> rebuild
cringe

>>16922067
yes

>>16922067
where is this from anon?

>>16922067
this is my algo from way back in feburary of last year. heres an overlay of jan 17th price

>>16922916
heres an updated prediction from back in august. you could probably search the post archives for an explanation on how it works/how to interpret it that i wrote back then
if theres enough interest ill post an update in the coming days.

>>16922922
Looks good, do you have a long term graphic ?

>>16922916
>>16922922
My Negrodamus!

>>16922996
i do, but i cut it off intentionally. backtesting shows the accuracy drops exponentially as the algo extrapolates further into the future. i tend to keep it at 2 years max.
there is also a short term version that can predict up to within a $100 error margin unlike the one i posted, but the predictions are profitably valid only up to a week max.
such is the way of machine learning.

Under rated

>>16922067
>>16923054
>machine learning
You can't use gaussian processes to extrapolate off-sample, numb nuts.

>>16922922
>>16923054
What are the main variables you use to derive price?

>>16923054
teach me your ways senpai, what kind of algo are you using? indicators?

>>16923089
>sees curves
>gaussian processes!!!

>>16923101
Explain what you're doing then, because it looks like total bullshit to a trained professional.

t. data scientist

>>16922080

>>16923112
except it obviously works you literal nigger

>>16923125
Except no it doesn't. Look at the fit on the 2017/2018 data. It's literally just bullshit random lines.

>>16923125
>>16923143
Also
>making 4 year forecasts on less than 2 years of training data
ayy lamo

>>16923097
>>16923099
it doesnt predict price, it predicts momentum. i did a writeup a few months back so you could probably google it, but to summarize, the algo finds clusters of momentum and regresses a model on those. i could probably fit the algo into a neural net framework, but at this point its just regression on a custom model.

>>16923112
>trained professional
>sees curves
>gaussian processes!!!

cmon now anon

>>16923167
You're using some fucked up version of basis function regression that GP's use. Don't tell me you're not, because you clearly are.

>>16923143
>trained professional
>doesnt understand resolution vs prediction accuracy tradeoff

whew anon

>>16923181
i smell desperation so ill give you a small hint.
i dont use anything gaussian or anything remotely related to probabilities.
that is unless you start interpreting wave functions^2 as probabilities.

>>16923182
>>16923200
Go look up how maximum likelihood GP's work. You created a fourier basis GP. Also nice 100% MAPE scores bro.

>>16922922

So what you're saying is, we will NEVER be under $13000 after July 2020?

>>16923212
>extrapolation
>use GP
sorry fren, but it doesnt work like that

>>16923233
no, its momentum, not price.
so the price might not even be at 13k or it might be at 20k.
regardless of where the actual price is at the corresponding overall momentum will be heading up so your best bet is to buy

>>16922922
>>16922916

i was looking to re-enter at 6700. is that never going to happen again?

>>16924167
Naw you’ll get to buy lower just keep waiting on the sidelines bro

>>16924180
i gave u a big tittied girl plz no bully

>>16922067
It can't be the new holy prophesy, because it's not associated with a smug anime girl post.