Any good books/sites to learn algorithmic trading?

[Anonymous No.52464028]

[Anonymous No.52464122]

Try algorithmic trading by Ernest P. Chan.

[Anonymous No.52464560]

Does this unironically help trading? Dont things happen unrelated to charts/the market?!

[Anonymous No.52465771]

Read a bunch. This has been by far the most useful.

[Anonymous No.52465788]

>>52465771
So why aren't you rich?

[Anonymous No.52465813]

>>52465788
he's an effective altruist

[Anonymous No.52465942]

Das rite!

[Anonymous No.52465980]

>muh logic

[Anonymous No.52466039]

>>52464122
Have that, only read some of it, but it looks solid.

[Anonymous No.52466082]

Haha, op, you know python?

Look into some ml algos.

I won't say much but I'll say you are on the right path.

[Anonymous No.52466091]

>>52465788
People with bots don't talk about them

[Anonymous No.52467531]

necro skull bump.

[Anonymous No.52469947]

>>52465788
How do you know hes not rich?

[Anonymous No.52469968]

>>52466082
Thanks annon. Will do.

[Anonymous No.52470920]

>>52464122
> Build models that try to predict the correct buy/sell signals from indicators calibrated to the current market climate instead of models that predict future price changes

This is the only useful bit in that whole book. Now you don't need to read it.

[Anonymous No.52471244]

>>52470920
Yeah I have this book and his quantitative trading book and their both underwhelming. It’s alright but I thought there would be more of material for the prices I decided to just buy the other book that was posted in this thread and see if it’s useful.
The most useful thing though is trial and error. Just have a goal to build something and learn as you go. You’re gonna get stuck a lot but once you’ll figure everything out you’ll be on a whole other level than people that just followed tutorials and didn’t build anything themselves

[Anonymous No.52471327]

>>52469947
because hes here lol

[Anonymous No.52471347]

>>52471244
>I decided to just buy the other book
> buy
bruh....

[Anonymous No.52471446]

>>52471347
I like hard copies and ain’t a poorfag

[Anonymous No.52471510]

>>52471244
> The most useful thing though is trial and error. Just have a goal to build something and learn as you go. You’re gonna get stuck a lot

That makes sense. This has pretty much been my programming journey so far. Everything is hard at first. Keep trying and failing until one day its not failing anymore.

Thanks for the info annon. Ill skip the first book for now and buy the second one.

Besides books, Any groups/sites/forums that you recommend ?

[Anonymous No.52471735]

>>52471446
Kek

[Anonymous No.52471752]

>>52471347
Man. You cant gain something for nothing my guy. If you are a poorfag and download books make-sure if you make it to pay back those who helped you.

[Anonymous No.52471822]

>>52471327
If i was rich I would be here all day

[Anonymous No.52471823]

>>52471752
>poorfag
Is algo trading then even something for poorfags? It won't work with $20. And w/o experience and track record nobody would hire him to play with their money.

[Anonymous No.52471874]

>>52466082
based artificial intelligence dev

[Anonymous No.52473384]

>>52466082
I did it for my bachleor thesis, predicting a tradin signal for the next day. Used KNC, DTC, RFC, MLPC, CNNCs, LSTMC (C as in Classifier). Has portential. Also am looking into transformers and HMMs right now. Also i just used basic features, no fancy layers etc. Could be a no brainer lmao. Read a lot of papers. Some are good, some fail to handle the data properly they feed into the ML Algos. Also many try to predict the price which well maybe possible but not like they do it.

[Anonymous No.52473404]

>>52464028
it's a meme

[Anonymous No.52473408]

>>52464028
After looking around, found these books.

https://fxgears.com/index.php?threads/recommended-books-for-algo-trading-in-2020.1243/

[Anonymous No.52473429]

>>52473404
Nigger your mom's a meme, make an nft and sell her on opensea.

[Anonymous No.52473455]

>>52473384
That's interesting. Any papers you suggest?

[Anonymous No.52473476]

>>52473384
Also Mandelbrot can be very interesting, thinking of it like gans and therefore possible trainign data. Feels nice being deep into deep learning learning deep truths iykwis

[Anonymous No.52473592]

>>52464028
There is no cure all book for algorithmic trading. If you want to get into algorithmic trading go finish a PhD at a top 10 program in math/physics/stats/cs and then go work at one of the hedge funds that dominate the field for a year or two.

In that environment, hope you are lucky enough to discover a set of strategies and leverage your track record to launch a firm.

Worst case scenario you are wagecucking for 300k+/year total comp minimum.

[Anonymous No.52473678]

>>52473476
Ohhhh now this is getting more interesting.

>>52473592
I mean, you gotta start somewhere right? Yeh I can see the phd --> hedge fund route making sense for someone.

But from what I have been by reading for the past 2 days what I see is, its a lot of work, but many people have found a way through on their own.

Even without the prospect of making it, this stuff looks really interesting so, I just wanna start somewhere and keep improving, hell who knows 10 years work down the line I could turn this profitable.

[Anonymous No.52473721]

>>52473455
Suggest not really, because most of those i have read do not k-fold cross validate.

A Paper that stood out was, one about transforming market data into image data and applying a CNNC. Altough the paper is full of spelling errors and they confues some terminology and im not sure whether they normalised the testing data with the stats of the training data or not:
https://www.researchgate.net/publication/340303825_Deep_Learning_for_Financial_Time-Series_Data_Analytics_An_Image_Processing_Based_Approach

If you are into RL there is a library and paper about it, one Author has a medium explaining how it works. I myself have coded a PPO agent but, need to further evaluate it also i dont posses the computing rescourses for stress testing:
https://ai4finance.medium.com/

There is also a survey paper about predicting crypto with ML and other Aproaches. I have overflown some of them, not the deep learning section yet doe. Some of the surveyd papers are also utter trash imo:
https://arxiv.org/abs/2003.11352

Right now my focus is on getting a deeper understaniding of some concepts i think are key and necessary.

[Anonymous No.52473763]

>>52470920
What does this really mean

[Anonymous No.52473793]

>>52464122
I only trust Ernest 4chan

[Anonymous No.52473863]

>>52473763
It means that if your goal is to predict the actualy price of the next day or next Intervall, then it's like saying the price of tomoorow (or the next Intervall) is more or less the same as today.
Suppose you want to predict the price of tomorrow, with a function f(x)= x * c.
Now no matter how complex the model c will take on a values close to around 1. So that if the price of today is 1k tomorrow it will be more or less 1k +-some%.This has also to do with the fact of how you optimize the model. Since you will try to minimize the error in neural Networks the value that gives the least error will be one to yesterdays price. if you use other models like KNN or DT or RF, then the price of tommorow will be an average of todays price..

[Anonymous No.52473953]

>>52473863
Ok this is really helpful. Thanks
I'll Google too but what are KNN or DT or RF

[Anonymous No.52474099]

>>52473953
They are k-nearest-neighbour, Desicion trees and Random Forests. There are also models using averages ARIMA of the like but i haven't dwelled much into them. If you are really interested into this topic for the basic algorithms u can use the sklearn library for python. You don't need anything installed on your device as you can use google colab.

An article writing about this topic but giving the wron reasons is this one here, but I assure you the comment section is more insightful and interessting than the article itself:
https://towardsdatascience.com/how-not-to-use-machine-learning-for-time-series-forecasting-avoiding-the-pitfalls-19f9d7adf424

For any ML Algorithm all you have to do is google a tutorial. This will get you started.
Also i would just focus on classifcation tutorials for predicting trading signals. The turoials will give you everthing you need to get started and i recommend to really think about the thinks you do and to understand what you are doing an using.

[Anonymous No.52474300]

>>52474099
I dabbled with ml to predict direction and it was not very useful. The markets seem to change to much. I'll review the resources you shared. Thanks fren

[Anonymous No.52474688]

>>52474300
>I dabbled with ml to predict direction and it was not very useful.
Same here fren. I didn't find ML very helpful while I was making my bot since there are just too many factors with crypto in particular that are impossible to predict. I've had much more success running a linear regression to determine whether we're currently at a local top or bottom and then buying or selling accordingly. Some days when we're really crabbing my bot makes 40% of whatever initial I plug into it for the day. It's pretty comfy.

[Anonymous No.52475616]

>>52473721
>>52473863
>>52474099

Annon thanks so much, this is really helpful stuff man.

[Anonymous No.52475960]

>>52473793
oh was he the person who gave a interview on cnn ?

[Anonymous No.52476712]

>>52474099
>and i recommend to really think about the thinks you do and to understand what you are doing an using.
What did you mean by this?
I feel like a kid that overhear adults talking about something, but want to go try what they were talking about himself. I know next to nothing about coding, but have been trading for a while and am not tech illiterate. If I do not know any Python like you mentioned to the other dude should I just steer clear of these bot? I can learn pretty much anything I have put my mind to and am leaning towards watching the tutorials like you mentioned and giving it a shot.

[Anonymous No.52477394]

im just starting to get into this. i have a math degree and been interested in ML for awhile. i dont have a job currently so i said fuck it, why not learning algo trading? currently doing an algo trading with ML course on coursera and reading whatever i can. i hope i can make money one day lol.

[Anonymous No.52478571]

>>52464028
The algo trading red pill is using stochastic differential equations and control

[Anonymous No.52479296]

>>52464122
I love that people replied to this as though you were serious.

[Anonymous No.52479693]

>>52473476
>Mandelbrot
My main takeaway from his work is that timing is indeed relevant, most significant market movements happen in only a few days or short periods.

[Anonymous No.52479773]

>>52464028
Stop the cope. You won’t compete with Wall Street quants. Get back in the wage cage.

[Anonymous No.52481612]

>>52479773
Just because you will never be a girl and your parents will never love you. Doesn't give you the right to go around posting dumb shit on the internet annon. If you got nothing to add to the discussion go fuck yourself.

[Anonymous No.52481620]

>>52464122
> p. Chan

[Anonymous No.52481630]

>>52465788
/thread

[Anonymous No.52482968]

You won't find any algo trading systems written down that still work. Only as soon as the edge is gone people write a book about it.

People kinda have the wrong view on algorithmic trading or what those big HFT firms do. It is just an edge in the market and the trades are executed by a computer. This edge could be simply based on speed of the trade or some scalping strategy. Which means it is based on action by others. As a retail investor you should look at a longer timeframe, don't waste your time and try to compete with the speed and processing power of the big players. The most important part is finding this edge! No book is going to tell you how to do that. You have to sit down and do your reasearch the old way. Looking for signals/conditions that give you probabilities of a potential event in the future. If you found something interesting you need backtesting. If backtesting confirms it you have now to put those signals/conditions into a formula with a few variables (time, size, exit...)

And only now you build an algorithm to execute the edge you found above. There was a thread a couple days ago about this one succesful Japanaese trader. His signal/conditions were firms that traded below a 25 day moving average and the overall price movements of the sector itself, because specific firms followed the overall price movement with a lag. Not sure if this is still the case but you can assume he lost his edge as soon as this became public.

[Anonymous No.52484189]

>>52464028
you know there are commercial bots you can just rent as a service right?
what makes you think that you can code a better bot yourself as a one man dev team?
chances are you will end up with just another version of least squares buy low/sell high scalping bot, of which there are thousands out there already - as a result your contribution will add to market instability, amplifying local boom/bust cycles and ultimately worsening the eventual black swan bust

>>52474688
>really crabbing my bot makes 40% of whatever initial
so the LTCM approach of picking up dimes in front of a steam roller
you would at least need a very good decision system to know when to switch the bot on and when to stay on the side lines
one unexpected high vola event could wipe out days of gains just like that
then there are scam wicks or if you get big enough: people starting to take notice of your order patterns who will try to take you out

>>52482968
that's Takashi Kotegawa and his strat may have worked in the 90s/00s, but obviously the gains from something as simple as that have reduced due to competition

[Cryptos No.52484335]

>>52466091
I do.

>>52471822
i am

>>52473429
she is

>>52484189
.need a very good decision system to know when to switch the bot on and when to stay on the side lines

yeah it's called set a ceiling and a floor limit.
Crab market is the best to scrape gains.

[Anonymous No.52484477]

>>52484335
Nice well im not going to speak about my bot.

All I'm going to say it's working g really fucking well and I'm about to be rich. I'll come back on here and make a thread after the fact telling how

[Cryptos No.52484534]

>>52484477
only one question... is the program free or does it cost you?

[Anonymous No.52484577]

>>52464028

[Anonymous No.52486637]

>>52476712
I just wanted to say do less right, than more but wrong. Understand the mathematical concepts of ML algorithms, the coding part isn't that hard appart some specifics, that if you test properly won't you will notice 'errors' or behaviour that deviates from what you intend to do. ML is the theory of applying math, mainly statistics, linear algebra and computing graphs to model a process that would be diffcult just doing by sequential instructions ( code). It also is beautiful in the sense, that if reality itself relies upon math or math is a good of enough model to describe reality, then ML will definitely tap into waters concerning the way reality functions..
Anyways if you are really into it, then be aware that it will or could take some time. Im not saying that it works a 100%, my best models are able to predict 52.5% true positives. Meaning that thos 52.5% are actually true predictions that could be used as trade signal. Of course one has to compare Returns and weight them, the easiest way is to compare against the hold return, which depending on how you train a predict can be outplayed. And mind i did really basic stuff, and have used simple features as inputs (log returns). Thats why i think they have Potential. Also ideal and theoreical returns don't consider the fact that you will not necessarly match the price, given the orderbook is constantly moving the price.

Model the Problem you want in a logic manner, just like you were explaining it to someone (what steps to take, in the simplest manner possible) and use ML as tool to reach the goal, the ML is only as good as your Modelling and understanding of the Problem. But don't get lost to much in some details, as working on the problem will always give insight. And backtest the shit out of whatever you will be doing.

If your results are too good always triplle or double check them. If you are able to predict the market it means you have understood concepts that are yet not known to humans

[Anonymous No.52487374]

>>52478571
What are you controlling for anon?