/biz/ - Business & Finance

[Anonymous No.11155541]

I have an algorithm that can predict peaks and dips with 99.75% accuracy.

[Anonymous No.11155548]

So you're a billionaire then?

[Anonymous No.11155555]

>>11155541
my friend we are almost there

cyka

[Anonymous No.11155556]

>>11155548
S-soon I hope

[Anonymous No.11155566]

>>11155541
Sure thing bro

[Anonymous No.11155570]

>>11155555
phew, checked

[Anonymous No.11155583]

>>11155555
Wasted

[Anonymous No.11155588]

lies

[Anonymous No.11155589]

>>11155555
Holy shit witnessed

[Anonymous No.11155600]

>>11155555
You should buy a lottery ticket tn

[Anonymous No.11155612]

>>11155541
it repaints, oops!

[Anonymous No.11155624]

>>11155541
can it predict chink scams tho, and pajeet shit shillers?

[Anonymous No.11155641]

>>11155624
It only works for bitcoin because that's the only coin with enough historical data.

[Anonymous No.11155653]

>>11155541
Can you help out an anon who is in bad debt make some btc?

[Anonymous No.11155660]

>>11155541

[Anonymous No.11155697]

>>11155555

Acknowledged.

[Anonymous No.11155703]

>>11155541
I'll take it

[Anonymous No.11155724]

>>11155541
You just overfitted on historical data. Backtesting is a meme, and actually hard to do right.

[Anonymous No.11155745]

>>11155724
But it generalizes outside my training set. I agree that the accuracy seems suspiciously high, but so far everything seems correct.

[Anonymous No.11155769]

>>11155641
Did you use NLP on social media, ML or something else?

How’s your test set? You must’ve gotten 99%+ on training set, right? Did you overfit anon?

[Anonymous No.11155776]

>>11155745
is it repainting on a forward test? (tickdata)

[Anonymous No.11155841]

>>11155769
Not using NLP or neural networks. It is ML though.

>>11155776
What exactly do you mean by repainting?

[Anonymous No.11155845]

>>11155541
post pics i dont believe you

[Anonymous No.11155847]

>>11155745
>outside muh training set

Major cryptos are all correlated. They’re all the same graph

[Anonymous No.11155850]

>>11155541
/g/ pls go.

[Anonymous No.11155880]

>>11155841
what indicators do you use?

[Anonymous No.11155919]

>>11155847
It generalizes to a different part of the same bitcoin time series.

>>11155880
They're not publicly accessible. I'm doing a ton of preprocessing on the data before I feed it into my model.

[Anonymous No.11155945]

> muh ML
> generalizes outside of my training set
> 99% accuracy
> everything seems correct
just imagine being this fucking retarded... god

[Anonymous No.11155959]

>>11155541
RSI.

[Anonymous No.11155961]

>>11155945
There probably is something wrong with it, but we'll see.

[Anonymous No.11155964]

>>11155919
> I'm doing a ton of preprocessing on the data before I feed it into my model.
oh there it is
don't tell me you do standardization/scaling of timeseries data before processing it with the model

[Anonymous No.11155969]

>>11155964
There's a bit more to it than that.

[Anonymous No.11155993]

>>11155969
yeah thats not my point
I bet you are introducing lookahead bias without even knowing it

[Salty No.11155994]

>>11155969
hey bro

all the luck to you

how soon will you guarantee 100% ?

so I can try it out

[Anonymous No.11156023]

>>11155993
I don't see how I could have with the way I implemented it. I can give it another look though.

>>11155994
I probably won't be releasing it if it does indeed work like it seems to.

[Salty No.11156047]

>>11156023
so if your not going to share you are here to gloat ?

>nice

[Anonymous No.11156050]

>>11156047
No, I just think there should be more interesting discussions on 4chan.

[Anonymous No.11156082]

>>11155555
checked

[Anonymous No.11156093]

>>11156023
since youve mentioned ML you are mostly working with the mean of the distribution, as in you are predicting the mean (assuming regular cost functions, squared loss, mse, hinge, etc). Now, unless some rando on 4chan made a groundbreaking discovery, in which case you are welcome to submit your algo to M competitions and similar, the GBM still applies. Which should tell you a thing or two. Money is made predicting the tails

[Anonymous No.11156119]

>>11156023
unless you are talking about HFT and similar. There the process generally moves away from GBM/exponential kernel and takes on another kernel (powerlaw, etc). I can do 80% accuracy predicting 30sec ahead with xgboost and order book flow. But good luck bridging the spread and fees.

[Anonymous No.11156132]

>>11156050
Well what the fuck is it? Either tell us or leave.

[Anonymous No.11156144]

>>11156093
>>11156119
Yeah we'll see. It definitely needs more testing. The special sauce isn't in the ML though: I just got tired of hand tuning everything.

[Salty No.11156165]

>>11156144
thanks for sharing

to generalize

where is btc going ?

should we still expect a dump or are we green from here ?

ty

[Anonymous No.11156167]

>>11156144
lol alright man, good luck with your special sauce, thanks for the "interesting discussion" I guess

[Anonymous No.11156173]

>>11156119
>>11156093
>>11155964
>>11155919
Wtf you guys use to build this? Python? How do I get started?

[Anonymous No.11156186]

http://charts.woobull.com/bitcoin-nvt-signal/


you're welcome /biz/

[Anonymous No.11156188]

>>11156173
it doesnt matter, build it in cobol if you like. What matters is the underlying model

[Anonymous No.11156191]

>>11155541
>predict peaks and dips
>still buying high and selling low

[Anonymous No.11156193]

You guys are pathetic
>>11156173

[Anonymous No.11156211]

tradeview is full of strategies with >95% accuracy but they don't work on live data
you should test it, thats what we call repaint

[Anonymous No.11156228]

>>11156144
I, as a creator of a BTC peak/dip predictor too, am going to say that what you are reporting could be true and may be entirely possible, kek

[Anonymous No.11156239]

>>11156144
e.g. my algorithm has been working since early 2016, and so far, as of now, +60,000% in terms of BTC

[Anonymous No.11156269]

>>11156144
by the way, it's magical that great minds think alike, kek

[Anonymous No.11156325]

>>11155541
Predict outcomes you say? I wonder. What the chances are that my team wins the dark tournament.

[Anonymous No.11156332]

>>11155555
noticed

[Anonymous No.11156335]

How the fuck do you predict this?

[Anonymous No.11156339]

>>11156325
99%, care about the masked dude

[Anonymous No.11156341]

>>11156211
right? It took me 2 years to finally come up with a long term (80% accuracy, but cuts losses, and lets gains run) indicator that doesn't repaint from higher time periods that have already happened. uses pine v3 open[1] or close[1],lookahead=barmerge.lookahead_off

[Anonymous No.11156378]

If its so accurate run it out 3 years and tell us all the prices and what dates they hit on.

Otherwise nice larp and fuck off.

[Salty No.11156393]

>>11156378
ty

[Anonymous No.11156399]

>OP says he was right 99.75% of the time
>nobody asks him about the other 0.25%
Seems pretty fucking important..

[Anonymous No.11156541]

post pine

[Anonymous No.11156574]

market make or get left behind.

im profitable since last october with the same strategy and made it with an investment of 5k, which was the fee i paid to get to a decent volume to get a reduction in fees.

gl

[Anonymous No.11156657]

>>11155541
Hand it over. That's not for your kind.

[Anonymous No.11156677]

>>11155541
Mine’s only around 50% it almost fucked me 3 months ago. For the noobs, its written in python