I'm trying to make a Binance bot. I read some tips on here pointed to this strategy to determine buys:
prices = stochrsi = STOCHRSI(prices, period=60) if stochrsi < 20: macd, macd_signal = MACD(prices) if macd goes above macd_signal: buy()
It performed well on backtesting but is not doing hot while live. Does anyone have ideas what I'm doing wrong? Should my price history be in hours? Should my price history go back a certain amount (30 days)?
Bump, please help guys, trying to make money so I can buy pic related
Chase Myers
Help
Kayden Green
Anyone doing TA bots out there? Any tips? The guy who mentioned StochRSI?
Nicholas Carter
Lol, are you actually trying to implement the meme strategy in life?
1. Strategies work on some coins and don't on others; 2. Have you considered MACD/RSI divergence/convergence? That is basically how you should use them; 3. Have you considered volume cutoff?
Gabriel Flores
Also, tried making a super duper machine learning TA crypto bot. Spent 2 months, failed miserably.
Austin Anderson
Make sure it can detect a S T O M A C H
Hunter Miller
>a super duper machine learning TA Can you simulate buys with that bot? So that it learns without actually risking anything in ETH/BTC/USDT? t.brainlet
Joshua Foster
Appreciate the response.
1. Shit 2. I am only considering convergence of the macd and macd_signal. What relationship between the rsi and macd should I be looking for other than what I outlined above? 3. Yes, I set a volume cutoff at around 200 BTC in the last 24 hours because at first I was picking up some VERY stinky shitcoins. Should my cutoff be higher?
Thanks
Josiah Smith
Of course you can.... just make a simulated wallet and keep track of the balance as you make calls.
Alexander Harris
I am considering making a Deep q learning bot using futures and a few other signals as input. Sounds fun but will probably suck
Gavin Anderson
>MACD/RSI divergence/convergence Okay I am reading a little bit more about this. Yes. I literally read the meme strategy and implemented it with python and talib.
Cameron Barnes
what does your buy function do? if it uses all your capital that's a mistake, if you're actually paying taker fees that's another big mistake. From typical fee schedules I've seen I don't think anyone is making money from a bot like yours. My assumption is that you have very little experience with this stuff.
Carson Nguyen
I have absolutely no experience with this stuff. I am using my whole balance. I am pretty good at programming and am trying to make some money so I thought it would be a good project. I am very motivated to learn what I am doing wrong however, so if you can point me in the direction of some good resources that would be great.
I am paying the 0.1% fee from Binance for every transaction. This didn't seem to affect the outcome too much in backtesting. The 0.25% fee on Bittrex cut a bit more into profits.
Cameron Robinson
>using my whole balance no, no, no. you want some semblance of equilibrium.
for each time you generate a new decision you should calculate your coin equity (price times size you have) and use that to determine a size for selling. For buying you use USD balance of course. You need a function that generates a float between 0 and 1 to apply to that dollar amount to calculate the final size.
Tyler Bennett
Bump.... any more insights? I know I am a babbie and my strategy is retarded but I'm trying to learn and get better.
Elijah Cooper
use BNB on binance to save fees m8
Lucas Flores
Okay thanks for this. I am a little confused by this. The function that generates a float between 0 and 1 uses a random generator? So lets say I have $10,000 starting USD. I identify a buy in point, generate some ratio (lets say 30%) and use that to buy $3,000 worth of whatever coin I identified? Then I have USD remaining to make other buy ins so I have a few positions open at the same time. Is this right?
Caleb Williams
Just study the movements and lurk for news stories and sell the news, don't try to get all fancy and technical.
Jordan Cooper
That would be nice, but the point of this is to set up a bot that runs in the background that requires 0 input from me to generate funds. Sentiment analysis is interesting but I'm not sure I want to go there right now.
I understand no one with a successful strategy will just give it up so thats why I'm hesitant about my current one.
Carson Evans
one thing is more on the technical side make sure your program is well behaved or the exchange will literally ban you for life. That happened to a friend of mine who isn't the smartest with programming, he left a bot overnight that ignored the servers rate limit and got banned in the morning.
Logan Jenkins
I think mine is okay. It issues probably 2 or 3 api calls a minute.
Michael Martin
partially random isn't bad if the alternative is to add a slow model but completely random would not make much sense.
Anthony Murphy
Understandable it would be nice to have a bot that did all the heavy lifting and you could just turn it on like a money faucet, but that neglects the human part of the market, panic sells and such when to hold and when to abandon a turd. Or when jews are jewing.
Daniel Wright
If it isn't random, then what is the ratio based off?
Joseph Perez
so one can assume that if your hypothetical buys were actually executed, it wouldn't have that much of an impact (on other bots and so on) and the price in the end?
Ian Walker
As long as the coin has high enough volume and you're not making massive trades it should not affect the price too much if at all.
James James
I learnt it sucked with backtesting Its best performance (out of like 5 shitcoins) was FUN, on which it managed to achieve 99% ROI per trade
Aiden Murphy
RSI is also supposed to be used as a divergence signaler. Try to look at RSI/MACD maxima difference.
Oliver Hughes
for my size function it's a model that takes the base dollar amount as one input and the other as the second it returns the % of the base to use. The specifics of the model used are something you can figure out like how you selected a model to make your buy or sell choice.
Connor Garcia
Thanks guys.
Justin James
>pretty good at programming >a 6 liner will make me money
choose one, no, nevermind
Parker Martinez
That is the core logic in pseudocode and there is obviously a ton of other stuff around it to talk to the API and manage timing and such. Programming is the easy part my dude. As is evident here programming ability correlates very little with the ability to identify a winning strategy.
Thomas Reyes
You know those dummies in the pic? Do they build them with fart pockets? At the very least can they be made with a fart noise maker. I think a good idea would be to somehow attach an air tight douche into them. The douche contains actual farts from other humans which can then be dispensed from the dummy. This would be a lucrative deal not just for the makers but for women everywhere. Because you could make a buyers/sellers market, similar to a sperm bank, and men from all round the world could get hot women to fart into these airtight bags and "store" their farts for use within the dummies. Someone should really look into how this would be done.
Carson Taylor
finally, I've been waiting for the real discussion ITT to start
Cameron Wood
can someone tell me the name of those products so I can buy one?
Charles Hall
Fuck Me Silly mega masturbator and Fuck My Big Black Ass
Watch what your bot is doing. You may notice when people put in a big buy or sell order it will trigger a false alarm until the market corrects from the high or low. This is the most likely reason why it works on backtesting but not live. So either wait for the candle to complete or the signal to remain stable for a few minutes. Also check that your backtest is not using data from the future. Your stoch RSI may be accidentally grabbing data to 59 minutes in the future after the minute price data the MACD is using.
Parker Garcia
incorrect! see
William Scott
@OP: Why not just buy a crack of Gunbot and fix it to your specifications? Are you realling building a trading bot from scratch?
Nicholas Ramirez
I have seen gunbot but didn't look too much into it. I already built the bot so I just have to tweak the strategy and I already wrote all the infrastructure necessary for backtesting and stuff so unless gunbot can get consistent returns I am not going to use it
Levi Hernandez
> backtest is not using data from the future This I have seen some videos on utube, where ppl boast about their efficient bots, when their graphs scream overfitting and data leakage.
Also I might suggest trying to combine signals from different time frames. Eg, buy if you expect te price to grow both in short and in long term, but wait if it is expected to fall in short term and grow in long term.
If you leave your telegram, I might share to you my basic script.
I'm happy to share stuff that's generic and still being tested (by me) on the offchance that someone might suggest bug fixes. I was doing interprocess http much differently before and came up with some seemingly unresolveable bugs so I rewrote it last week. This seems to work ok.
you will make more money arbing the exchanges and fiat pairs out desu trends are to choppy for usual systematic trading method, I get more out of news trading than systematic
Henry Baker
>I'm happy to share stuff that's generic and still being tested (by me) on the offchance that someone might suggest bug fixes. Hey, I would suggest using this:
Much much easier than trying to do http communication yourself
Thomas Rogers
bump
Jeremiah Sullivan
oh, I do not use binance(yet) I only have an account with gdax. It looks like their fee structure isn't very good for me. Also I could not verify that binance deposit accounts are fdic insured. I always want to expand to other exchanges but gemini looks like the only option and I don't have a large enough account to do 5000 btc per month.
Xavier Morgan
hey genius, stop relying on one indicator to trade
Sebastian Powell
adding to that, you need to actually understand what the stochastic RSI actually does before you can understand my previous point.
Adrian Miller
ratio of average return signs normalized and filtered you're not smart (You)
Asher Mitchell
Why are you wasting your time (poorly) replicating the http stuff, instead of just using their libs?
No, it is not better than what I have seen from others, and it seems overly complicated.
Thomas James
>their libs gdax does not release an official python wrapper for their REST api or anything for their websocket feed. However if they did, it would be similar to the solution I had before. I wrote a wrapper with the client object and http session in the same process as the websocket client, and shared them with my trading programs using the pyro library. That didn't work well for a multitude of reasons so I had to move on to the second solution.
Brandon King
Yes, do it kek That’s how you lose your money
Asher Johnson
>$930 for fake pussy
Parker Lee
What other objections do you have to sex toys?
Luke Sanchez
What you have written is both poor quality and messy.
You will never write something as good as this. Feel free to continue to waste your time.
Ethan Johnson
>poor quality and messy It's fine for a first iteration, and I doubt anyone values your moronic opinion.
What I uploaded is not even an api wrapper for any exchange but I guess rubbing your 3 braincells together didn't allow you to determine that on your own. All it is is a way to pass requests and responses between processes, share connections to target servers and set rate limits per server. Nothing in the wrapper you poasted enables you to do that or is even related to that problem, so you can fuck off since you obviously don't have any programming knowledge.
Also, completely unrelated, but what happens when coinbase makes a breaking change to the api? do you fork that wrapper and fix it yourself, or do you wait weeks for 'danpaquin' to release a new version?
Dylan Reed
>I doubt anyone values your moronic opinion
You are the one who posted code and asked for an opinion on it, fool. Clearly you value it, judging by your long winded posts.
>what happens when coinbase makes a breaking change Coinbase along with everyone else would offer backwards compatibility for quite some time. Why do you think APIs are version numbered in the first place? They would only break anything if it has a security vulnerbility. In which case, they are doing you a favor.
Jordan King
>on the offchance that someone might suggest bug fixes. If someone had the same problem they could use something to start with and maybe they'll share improvements with me. That has nothing to do with someone opening it, reading it, not understanding what it does and commenting aimlessly. >backwards compatibility first thing to break in production, and they don't give a shit (it's only till the end of the year anyway!) When you email support about random 5xx errors that come after the upgrade they say please just switch to the new api. That's happened several times on other projects.
Camden Murphy
Like I said, feel free to waste your time.
This mentality is how we ended up with 1000s of front end javascript frameworks and other heaps of useless code.
You could be forgiven if you were writing in something obscure like Idris or Clojure.
Jaxon Allen
I didn't duplicate anything that can be easily found on the web. It's two small programs that solve an annoying problem, comparing it to browser app frameworks is retarded. And if you're still referring to the gdax wrapper I didn't use you might just be a hopeless case, since like I said before, it has nothing to do with what I uploaded.