Daily reminder that TA is the biggest meme of all time. Go ahead and press play

When I write up the position of the moon when I'm taking a shit I bet I get results that doesnt show a normal distribution along all 360 degrees of its possible positions. I can work with that and make assumtions where the moon might likely be when im shitting next time.

An accurate forecast might then be explained like that there exist some unknown function that connects the moons position to my shitting intervals. Its that function that caused the correlation.

> It doesnt look like a random walk
> Posts a pattern matching a markov process to a tee
user you do realise that the scale, peaks, tops and distribution of a random walks is affected by some hidden state? As in you could model any financial series with a makov progression by changing some variables.
The problem is inferring these variables with TA.

> thinks this somehow disproves it being a random walk distribution

ta works. but less is more.

Read a book idiot assuming fractals and random walks on price fluctuations hasn't worked, people tried it and lost money

While of course people tried TA and are just swimming in fucking money. Also, do you get that theres a difference between approximation/modelling and just trading on it?
Obviously you cant trade on it directly since it doesn't account for the heteroscedasticity of most financial time series. However its pretty obvious that any lag t+1 is dependent on tH where tH is current lag with some hidden state. Again, the problem is inferring this hidden state with TA, which you have not really shown to be possible, instead you are talking about some bullshit fractals who nobody would ever use lol

t. scientist
I know what a random walk looks like, we use them to model all sorts of shit eg. protein folding, chemical diffusion. But apparently you don't because it doesn't mean there's no patterns to analyse

TA is useless with Crypto. It's mainly dependent on mass shilling.

How do you think DOGE coin got to be so popular?

> what is stochastic calculus
> what is markovian modelling
> what is bayesian stochastic volatility
> what is state space modelling
where did I say there's no way to analyse it? I'm saying TA is not the right tool or do you depend on subjective methods too in your science work you retard?

People who believe in TA are the same kind that think astrology and homopathy works.

Wouldn't TA be a extreme simplification of these models though?

Like a huge return one day would be considered an outlier in our stochastic model and we can expect it to return to normal, similarly the TA babby sees a huge jump in prices that diverts from the previous jumps and concludes that the price will move back to the trend line.

I'm not saying its good but there is some validity to it.

...

If anything, TA would be an extreme, subjective (because you are still relying on some interpretation) simplification of classical time series analysis. By that I mean only working with various auto correlative models in the ARMA family (AR, MA, ARMA, ARIMA, etc)
And you can test your example of mean reversion really easily. Take the BTC time series, apply ARIMA on it, plot the residuals. If the autocorrelation is gone you just explained it away and the series is indeed mean reverting. The problem with TA is, its not reproducible this way and relies on subjective interpretation

Well I'm not saying TA is entirely useless. If there aren't major events happening, TA may work because traders believe it works. It's a self-fulfilling prophecy. But hype and news can fuck up the trend most of the time. Most alts have been taking a huge dip lately not because TA had predicted it but because BTC price soared. On september 22 TenX went up 30% because of a picture posted on Twitter, not because muh falling wedge. Lumen has been pumping because of major news, not because TA predicted it. "Buy the rumor sell the news" is the only method that has worked for me so far.

You're not far off.

Just a tip, you can use the bollinger bands as a good indication of future performance.

You'll see that they tend to tighten whenever the volume increases. Volume can indicate an imminent price crash or moon, to determine that you can use other factors to predict it such as market sentiment or all the shilling that goes around here.

So yeah, it can be very simple but you will need to look deeper into the fundamentals.

>"Buy the rumor sell the news" is the only method that has worked for me so far.

Yeah, that or "wait till everyone else loses all their money and watch them all bitch to teach me a lesson".

And to expand on this, simple mean reversion is usually engineered as it is not common due to volatility clustering (which is the defining future of financial time series data). For example applied stochastic optimal control is usually about finding pairs that mean revert and finding the optimal, dynamic hedge ratio to trade them on.

this user just described the violation of homoskedastic assumption, as in the volatility is itself an unknown

> defining feature
no future lol

>Again, the problem is inferring this hidden state with TA, which you have not really shown to be possible

TA is using time series analysis (indicators). You can find correlation between them and the price by using machine larning tools, the approx coverges exponentially fast towards the optimal solution learn 2 autism

But of course, opening paint and drawing meme lines does nothing, you have to do it proper

You cant use volume just because your TA book says so because exchanges are faking volume.

Well I see you haven't really read what I wrote so I'm not even gonna bother. I'll just tell you this, TA is fucking far off from anything related to time series analysis, otherwise this board would be full with phillips-perron tests, dickey-fuller, etc. Instead we have meme charts and statements like
> the approx coverges exponentially fast towards the optimal solution
I take it you are making a billion bucks just trading on your python+keras setup then? If not, why not?

That is the right answer to "literally ta but not for dummies" however it's still better than purely emotional trading. And the meme lines people do here are not real lol

>TA is fucking far off from anything related to time series analysis

a technical indicator is a mathematical calculation based on historic price, volume, or (in the case of futures contracts) open interest information that aims to forecast financial market direction.[1] Technical indicators are fundamental part of technical analysis

Its from wiki. Its literally the opposite of what you've wrote. TA is NOT drawing meme lines

Also keras an python lmao

The machine learning tools are self written and I use C#, python is to messy for large apps. We will see if my approach gives me millions

Oh man... This is pretty hopeless I see. Those mathematically calculated indicators can be charted like every other fucking math function resulting in meme lines posted here and everywhere on youtube, twitter, etc.I hope I don't need to link those for you.
I can come up with a fucking function so that F(x) = price * your homosexuality levels. Doesn't mean its predictive of anything besides your homosexuality levels. Time series analysis would you proving said function explains away some feature of the time series, e.g. said function can induce stationarity for example, through a reproducible, objective process.

are you a bit slow or something? the statement was sarcastic. Also wow C#! You must be fucking Gennady Korotkevich himself. We'll see I guess. But, my guess is, you gonna buy high and sell low.

TA works to some degree because people trading thinks TA works. It's why broken support or resistance creates runaways.

Regardless, you should learn TA because people trade on it. Specifically bots.

TA is for guiding intuition
if you don't think it helps, then you're just retarded
the lines help you visualize support levels and such... its not an objective science, it helps with decision making

This guy gets it. Moving away from forecasting with TA and forecasting in general and instead focusing on exploitation. Thats how quants make money

I dont care if its a function that is absolute shit as long as it leads to some correlation to future data and thats what I want.

stupid

To exploit TA it must be correct first even if its "man-made"

Cuck

haha
>I dont care if its a function that is absolute shit as long as it leads to some correlation
Keep relying on subjective correlation, give the roulette wheel a try while you are at it
> stupid
great comeback
> To exploit TA it must be correct first even if its "man-made"
You are a real retard arent you? Lets say you believe that your homosexuality is somehow curable. I sell you anti-buttsex tea which you should take every night before bedtime for $5k a pop. That makes the tea "correct" right? Even though its "man-made".
> Cuck

I bet your handwritten machine learning implementation is a bunch of if..then..else soup too. lol what a brainlet, no wonder making money here is so easy

what if there is no rumor, but only news?

If TA works reliably we would all be rich by now

But user, all of modern computing boils down to conditional gates

keyword being "soup" user, not conditional gates

You dont understand. If the price of btc goes up everytime you have buttsex and you buy btc everytime you get buttfucked then things doesnt need to be correlated in any way as long as you make money.

When you have 100 things that have nothing obv to do with your model but the graphs correlate you can use these things to forecast prices. The reason this works is the law of large numbers. The more correlating data you have the less it is wrong

lol nice dodge
> law of large numbers
hahahaha I know about basic probability theory thank you. Thats not how it works though. LLN is applicable to a time series if the process is ergodic. The vanishing point can be established as | t - s | increases. Otherwise you are just shovelling shit around and will overfit. Do you seriously think you are the first one to come up with this shit? LOL
But hey, since you are so successful why don't you post the drawdown% and the sharpe of your algorithm and we'll go from there. I'll wait.

In my algo overfit is prevented by throwing in random values. Also you only need to be right 60% of the time to make significant money assuming what you are trading has enough volume.

Of course my idea is not new, as I said I read books

> if it overfits I'll just throw in some dropouts and batch normalisation
ah, the classic student approach. I don't know whats going on, but I'll just throw in some shit in there and hope things work out.
> as I said I read books
oh I can see that user, but I guarantee thats all you do. You wouldn't be as spastic as you are otherwise. Still waiting for those stats my man, since you are making significant money apparently. I'll keep waiting I guess.

And I know you gonna say "I'm still working on it, but you'll know the power of my million billions soon" of course. But all I'm asking are some backtest results.

I have a nice proof that I wont share. the program isn't even finished so no stats

lol as predicted sure you do user, sure you do. Anyways thanks for humouring me, back you go to fitting those random forests, good luck

Yeah you wish I would post something substantial because you're obv so pessimistic and out of ideas that holding btc is the best strategy for you.

Lmao enjoy those measly 10-20% gains every month poorfag

Blown the fuck out
How will h81opB59 ever recover?
Downgraded to literal nigger status

TA tells you probable future price action. This is why technical traders have stop losses.

user you realize there are people who are good and bad at TA right? lmao

> Lmao enjoy those measly 10-20% gains every month poorfag
C O P E
O
P
E

real talk though user, you and I both know that you are completely out of your depth right now and all of this is just imaginary stuff in your head. I mean its good fun, but dont delude yourself too much.

surely you'll agree more (much more) bad than good though right?

Interesting discussion but, can anyone provide an actual method that works?

actual math works user. If you are interested look up "Quantitative Equity Portfolio Management" by Qian a good introductory treatment of the subject. Then go from there.

I don't understand a single word from the "stocharithemtisticlslaslda" talk here but how can TA work in a market with such a low marketcap where any whale can pump the price at any time?

Threads like these make me glad I'm an Intelligent Investor (TM)

hey Veeky Forums newfag here what strategies *if any* do you guys employ to know when to exit a trade if it goes sour? I'm trying to daytrade and the ONLY thing keeping me from success is not knowing when to cut my losses and i end up getting fucked way harder than i should be

you're a fucking retard. please kys. QUANTITATIVE ANALYSIS requires sophisticated math skills, TA is for poorfag retards who want to larp as the next wolf of wall street. chances are that you don't even know how to solve a differential equation

>however it's still better than purely emotional trading.

It's no different actually.

The way i see it is TA cant tell you when to buy or sell on its own, but if you know you want to buy a stock and youre looking for the right entry point then TA will help you find a general low point or a general trend on the chart

What everyone seems to be forgetting is that day traders might as well use a magic 8 ball according to you TA bashers

>he doesn't understand elementary cryptopsychohistory

meh, I lost money when I just guessed, and make "shitty 10-20% gains" with ta so whatever, it works to some extent. but my maths is good enough probably to do some of these things you mentioned, so thanks for sharing a bunch of ideas at least. i'd never really looked into it before.

This. If bitcoin reaches a double top and stagnates for a few days, it's usually going to crash and level out. TA will help you work out when to enter then