/biz/ - Business & Finance

>>25352888
harder than you think, if exchanges had zero fees everyone would use one.

Hey, still waste your time on shitty tokens, don’t be a fools check Polkacover and CVR token. Get your piece of cake right now while they have presale

>>25352953
I cast you out of this thread, do not come back

>>25352888
Unfortunately not that easy. I've been writing on my bot as a side project for years now, but never managed to get it green. My best stats were 45% win, 55% loss. It was close, but obviously you need 50%+. Never got there, but I'm continuing to implement new strategies all the time.

>>25352953

fuck off with your pajeet scams man, seriously. FUCK OFF

>>25352888
if (price == low):
buy
else:
sell

I wrote a bot, I usually use it to set cascading buys and sells that dynamically increase the lower/higher a price drops, but right now my bot is executing $10 market buys on grt and link every hour till I run out of investment juice.

>>25353112
>My best stats were 45% win, 55% loss
Can you compare your experience with shorter-term/higher frequency strategies vs. longer-term/lower frequency strategies?

I assume if I have a "stricter" set of requirements to execute a trade I'll have less trades but maybe a friendlier chance of a profit. But I feel like that opens me up to get broadsided if I'm making less trades and the market as a whole takes a dump, since there will be less chance to recover.

>>25353254
sir can you please help take a look?

File "<stdin>", line 2
buy
^
IndentationError: expected an indented block

>>25353112
For someone who claims to be writing one for years you sure don't seem to know how to measure its performance.

>>25353254
easy

>>25353354
lmfao xD

>>25353112
Use tradingview, but becareful of inaccurate results. Real trading =/= backtest.

Tradingview backtest analysis note

1. average # bars in trade should be >2.
- low number results in overtrading
2. if using tf < 1hr, position should NOT close within the same candle.
- will cause an error of entry/exit alert because it's too quick
3. careful of using trailing stop in backtest.
- see point 2 and historical OHLC limitation
4. max drawdowns <10%, <20% acceptable
5. percent profitable > 50% ideally

Hope you doing better next time bros.

>>25353302
I tied both short term and long term trades. I know what your mindset is, I used to be there. But once you actually sit down, look at the numbers and need to implement the logic for how/when to execute trades you'll realize it's much, much harder than you think.

So, I got my best stats which this strategy: have the bot figure it out for you. Meaning: run **lots** of simulations.

1.) Collection as much data as you can: all symbols (ETC/BTC, BTH/USD, ...) from many exchanges months to years back.

2.) Write an algorithms which places trades based on moving average indicators (simple moving averages, SMA, and exponential moving averages, EMA -> https://www.investopedia.com/terms/e/ema.asp).).

3.) Generate a lot of fake/simulated trades. On each symbol, execute trades, from months back to the present. Each using different (incrementally generated) parameters, including your tolerance for loss (stop-limit orders). So for example:
> 5min SMA combined with 3min EMA (or SMA)
> 10min SMA combined with 7min EMA (or SMA)
> 15min SMa combined with 11min EMA (or SMA)
> ...
In other words, it's "trade by brute force. This will created hundreds of thousands or even millions of trades

4.) Calculate the profits on those simulations and rank those. Find the common pattern which made those trades profitable. Consider details like marketcap, volume, ask/bid spread, etc.
5.) Once you found that pattern (in an automated way), apply those parameters on actually, real-world trades.
6.) Profit (maybe...)

>>25353907
That's a man.

This thread is amateur hour max.

Not a single mention of triangle arbitrage.

NGMI.

With that said, even if you understand triangle arbitrage, you are still NGMI because of the balances required, limits on outgoing transactions, latency, etc.

>>25353606
> 1. average # bars in trade should be >2.
What do you mean by this? Are those bars based on candle views (3min, 5min, ...)? 50% profit seems **really** generous.

>>25353907
Sounds like a lot of owed tax if you end up having to pay.

>>25352888
full auto bot that gives profit is impossible to create. Manual interaction to constantly change the bot buying and selling conditions is necessary.

Better off buying manually and selling using a bot (or basically trade entry manually, trade exit with bot for short/long). Opposite can also be done, trade entry by bot and trade exit by manual. In very rare cases both entry/exit can be done by bot without leading to a loss but you constantly need to monitor the market and make adjustments to the conditions you set initially to make profit each and every trade. Eventually all your manual effort will be the same as if you never had a bot and you were trading manually.

I personally wrote and use a bot that puts trailing stop loss after I have bought manually. I select the trailing stop loss from a collection of different Moving averages or other indicator lines of those sort that closely follows current price. So if I buy at the right time manually, the bot will take care of selling it at a high enough point when price hits down on MA line. This can end up much more profitable than constantly buying and selling using a full auto bot and making 51% profits and 49% losses.

>>25354162
I tried arbitrage trading before, but as you mentioned, withdrawal times make it almost impossible. Things move to fast.

> This thread is amateur hour max.

This is /biz/

>>25354271
>Sounds like a lot of owed tax if you end up having to pay.
What do you mean? Capital gains is pretty straightforward.
>(SUM(net profit on all short-term trades in [YEAR]) * [your income-tax rate]) + (SUM(net profit on all long-term trades in [YEAR]) * [your capital-gains tax rate])
The number of taxable events does not affect the tax you pay, AFAIK

Besides, the guy you responded to was only talking about SIMULATED trades, ie paper trading, not real.

>>25354417
>Manual interaction to constantly change the bot buying and selling conditions is necessary.
I understand this point, but I don't see why this would necessitate manually entering or exiting trades. Adjusting entry/exit parameters makes sense, but the entire purpose of the bot is to monitor the market for these conditions 24/7 and execute the buy/sell when they are met.

>>25352888
Sombeody on HN coded a good CEXbot. Then Binance started to frontrun it.

>>25355183
>Then Binance started to frontrun it.
pardon?

>>25354556
>necessitate manually entering or exiting trades.

it does not , I am just saying my experience that writing several algos for a million possible conditions that market can go.. is difficult, very time consuming, worrisome, eventually unfulfilling as in how much profits earned ...compared to just entering the trade manually and exiting maybe using a trailing bot especially on 1h,2h,4h or even 1D charts.

>>25352888
Surprised nobody's mentioned hummingbot in here

>>25355352
that makes sense, i can imagine the frustration in spending time solutioning for edge-cases, waiting for simulations to finish, and then finding out there are more edge-cases for your edge-cases

bump for interest

>>25356070
tldr: it's not your "make easy money" ticket

>>25353907
>6.) Profit (maybe...)
kek, you think backtest != live.
there are many backtest millionaires out there

They work best in a crab market. Quick gains or losses can leave your bot holding bags. Pumps and dumps can trigger your bot to drop bags prematurely or late.

Basically, it does best where humans do the worst, and vice versa. After about 3 years of bot trading, I just went back to manually holding and trading again.

>>25356149
fuck wrong words.
what i want to say is that algo trading is where the new dumb money is. retards believe they just need to go into algo trading and they'll make money. nope. you need an edge. even this >>25353907 is bs. people come with profitable backtests all the time. the kick is that they dont work live because these stupid backtests don't take into account that your orders have an influence on the price.

>>25356149
The problem with backtesting I think is the order book depth. Your backtest probably doesn't have any orderbook data, so you could run a sim executing $10,000 trades, with a stoploss of -0.1% that always fills automatically.

I can't really think of any scenario where a backtest translates to real-life for this reason. Maybe it's relatively close for small trades, but totally unrealistic if you're trading a big wallet.

>>25352888
I wrote a few arbitrage bots for EOS shitcoins. That's where most of my current crypto stack came from. It was very difficult and emotionally draining due to the fact that you lose money when there's a bug or even when the network behaves unpredictably. And the only reason I was able to profit is because I was trading low volume shitcoins that no one cared about.

As far as trading with the big boys go, try it with fake money first and see how you do. 99% chance you'll lose money. Successful bots use advanced algorithms that take years to develop.

>>25356319
yes, check >>25356305
i wanted to say "backtest != live"
also implying you can fuck other participants with math PhDs + insane capital with your your little MA algo. kek

>>25355420
guaranteed profit eh?

>>25353112
you know what?
what if I told you that 40% is normal for a trend following system?

what if I told you that you should trade trend-following systems on 4hour charts at a MINIMUM and actually should just use weekly for that?
what if I told you that you are probably looking for a mean-reversion system instead if you want to trade small time frames? what if I told you you can get 70%+ accuracy by running a mean reversion system?
What if I summarized it like this:
trend-following = long timeframs
mean-reversion = short timeframes
what if I told as soon as you get your head around the difference, you should start making money soon after.

>>25353907
what If I told everyone that all this is irrelevant if you trade the weekly timeframe?

what If I also told you that fees and slippage dont mean jack shit on weekly timeframe?

if I told you all that, would you pay for a nice prostitute to come home to me for an hour?
you really should.

really, threads like these make me feel good knowing what competition I am up against. stay poor faggots.

>>25353112
I can get you a better win rate than 53% pretty easily.
Are there any services that make writing bots easier for non programmers? I have a nice feeless exchange already.

>>25356305
> people come with profitable backtests all the time. the kick is that they dont work live because these stupid backtests don't take into account that your orders have an influence on the price.
Everyone who writes those bots know that. The point is to place trades on *probability*.

>>25356905
> what if I told as soon as you get your head around the difference, you should start making money soon after.
Does it work out for you? Thanks for the advice, I will look into this.

>>25357435
> really, threads like these make me feel good knowing what competition I am up against. stay poor faggots.
Your real competition are institutions with a lot of brain power, which keep their mouths shut. I.e. BlackRock, RenTech, ...

>>25358556
since a while I am running trend following systems on weekly. and I dont care about blackrock or rentech or whatever, we dont play in the same park. I am probably too long term for them. I pass under the radar if you will.
I am not competing with them trying to scalp or something.
scalping is a meme put forward by exchanges and brokers who live off your fees since you can make as much or more money in the long run from NOT scalping and doing sensible shit and putting in 10% of the effort required to scalp properly. heck, on the weekly I even have algos for stuff like apple stock and certain commodities. forget about scalping, its way harder.

>>25358706
> scalping is a meme
Heh, fair enough.

> since a while I am running trend following systems on weekly
If you don't mind sharing, what's your average ROI in %?

>>25352888
probably better off figuring out how to make a bot for selling WOW gold

>>25358850
scalping is not an outright meme but considering the effort/time required, its not worth it considering the alternatives like daily/weekly trend-following, certain options trading strategies and a few other things


picrelated is one year backtest. I am running it live since around two months now.

>>25359105
Very impressive, not to mention how little trades you place. What's your trading volume?

this is also one year, 2019-2020 not as good.
but still amazing considering the low effort.

in picrelated, I dont know why tradingview shows 500% for buy and hold because its just not the case.

>>25359282
I am small fish, I trade around a 1000 at the moment. its growing steadily though.

>>25359392
That ROI on that small volume is amazing. I have about $210k from trading by hand, if I can somewhat decently implement that bot (finally...) then trading with $30k-$50k will be a blessing. But one has to consider the impact when trading which those larger volumes (it's not even close to whales, though).

>>25356482
>also implying you can fuck other participants with math PhDs + insane capital with your your little MA algo. kek
I know a good number of people with Ph.D's, and the only difference between us is that they enjoyed focusing on academics, where I preferred to jump into industry and make money.

Guys with a Ph.D are on average pretty smart, and fairly diligent, but they aren't quant-gods that can do anything above and beyond the typical smart guy.

>>25358518
>Are there any services that make writing bots easier for non programmers? I have a nice feeless exchange already.
not for free, but if you can't understand basic programming, you probably aren't smart enough to make it

>>25359105
>I am running it live since around two months now.
Post your stats for the live run, not the backtest, I am interested in the difference

>>25353907
Babby discovers SAA by chance and commits data dredging by accident

>>25356149
>>25356305
>>25359492
So can this be the reason why all cryptos move in tandem? That the wales need to spread their bags across all the cryptos, else they'll shoot themselves in the leg at every trade?

>>25353112
You don't need higher than 50% win to be profitable though?

>>25359711
this since sunday night.

These are my backtesting results for the past year. I ran the algo on a paper account for about 6 weeks and got identical returns. Just recently put in real cash. This is all unleveraged too, though I should be able to bump up the percentages with 2:1 or even 4:1 leverage. Can't short, however, but that's clearly much less profitable than longing.

Trading FNGU on Alpaca Markets (0% commission) btw

>>25360079
this is for sure probable.
I mean, I tried this with my measly stack, basically diversifying over many alts but stopped doing that since I got the feeling I was too diversified.
the algo works on majority of crypto so that was worth trying.
now I am back to monitoring all but only betting on a few.

>>25356905
so mean reversion for how short timeframes? a week and less?

does a bot actually give you any major advantages? like computing power, trailing sl, 24/7 activity, etc. or is it mostly just comfort of not having to manually micromanage your shit?

>>25353112
why dont you just sell whenever your bot would buy and vice versa? Ez 55% Winrate

>>25360136
You said you've been running this for two months - do you have those stats? A 4 day view says basically nothing

yeah, takes away all emotions. its great
(backtests are from 1-1-2020 to now)

>>25353112
>My best stats were 45% win, 55% loss. It was close, but obviously you need 50%+
what, why the fuck?
pic related wouldnt need that for example

>>25360184
this is good. what timeframe?
careful of the leverage though, I trade max 1.5x leverage
but never more than 1x on one position.
you can put 1x on your main positions and hedge with a short signal with 30% (the .3 in the 1.3x). algo has to work on many alts though to pull that off (or have multiple algos)

>>25354162
I did arbitrage between GBP and EUR crypto pairs on Coinbase for a while. Was making £2000 a day. Sometimes you could make 2% instant profit... but then had to withdraw the gbp and convert back to euro... leading to my bank accounts being shut down.

>>25360079
I can say that this is (likely) false. trading bots interpret market wide buy / sell signals from bigger moves in btc or eth (especially eth) to basically “arbitrage” forgotten-about trading pairs

>>25360913
I find that a bit hard to believe. Exchanges run their own arbitrage bots with zero latency, so it stands to reason that they would be getting 99% of arbitrage profits while everyone else who tries ends up losing money.

>>25360184
looks like heikin ashi fake results

>>25360429
tbu
Why wouldn't this work

>>25361022
bc its a strategy based on random backtest results, there would be no strategy whatsoever, pure luck only

also you can still lose money even if youre strat makes profit 90% of the time, or make money when youre only right 15% of the time

>>25353112
>It was close, but obviously you need 50%+
when i said dumb money is now into algo trading :D

>>25360789
>>25360853
you're aware that nobody cares about your backtests? those stats dont mean shit.

>>25361065
If his bot consistently produces sub 50% win rates then countertrading it should work?

>>25361022
>>25361209
>Why wouldn't this work
Because the opposite of a particular bad move is not necessarily a good move.

>>25361234
If he has been working on it for months and always produced sub 50% winrate then chances are counter trading his algorithms would produce >50% winrate

>>25361180
theyre a decent indicator for sure

>>25361130
What the fuck are you talking about?

>>25361342
>then chances are counter trading his algorithms would produce >50% winrate
that doesnt mean he will make money
>>25361180
and im quite happy with the results, theyre beautiful

>>25360913
Yeah you'd think that was the case - but for whatever reason it wasn't. Once made £4000 from £100,000 in an instant due to some massive sell order on the GBP pair making it cheap to buy. Only made ~20k from it before all my bank account were shut though.. was hardly worth it

>>25352888
holy fuck that picture

>>25361018
not heikin ashi, though i've fallen into the trap of calculating based off of those and deluding myself with percentage returns in the millions. these are actual candles. i had to test the strategy on a hundred different stocks, ETFs, and ETNs before finding one with returns like this

>>25360867
37 minute timeframe. as above, i tested it on every single timeframe i could, found that it worked best on 37 minutes, then tweaked all the inputs to optimize. I agree about leverage, but i'm much more comfortable with it when i leave everything in the hands of the algo. it's dispassionate, and if the backtest results are correct (which so far in real-time paper trading and live trading they have proven to be) i want to maximize gains

>>25352888
>but surely it's harder than it seems or everyone would do it.
most people give up after a few weeks, months. thats it really. you will need a few years but it will be worth it
start with the intention to learn and understand, not with the one to make as much money as quick as possible

>>25361459
>i had to test the strategy on a hundred different stocks, ETFs, and ETNs before finding one with returns like this
i hope you know that this means the strategy is shit and the one result was pure luck, right?

>>25361503
yep, that's fair. i don't really care though, i'm not here to create the world's greatest strategy for some dick-waving contest, i'm here to make money and ~2000% annual returns are good enough for me

>>25361399
why dont you provide the perf of the live trading instead of the backtest?
>>25361346
you don't need to have 51% winrate to be profitable

>>25361459
>i tested it on every single timeframe i could, found that it worked best on 37 minutes

>>25359105
>Sharpe ratio: 0.5
Kek fuck off

>>25361653
i dont share USD values on 4chan, i posted many active sexy trades tho
this strat only runs since a few months anyway

>>25361966
>Sharpe ratio: 0.5
Literally annual S&P500

so simple
so beautiful

>>25361653
Just to clarify, by this:

> My best stats were 45% win, 55% loss. It was close, but obviously you need 50%+.

I was referring to the actual profit/loss, not number of trades.

>>25360913
>>25361404
wait wtf why do they shut down bank accounts for arbitrage?

>>25353112
lmfaooo that's less than a flipped coin ahahahahaaaaa

>>25362267
it could still make massive profits you pleb

>>25361966
this is the backtest. who cares about your sharpe ratio? I care about gains.
this is bitcoin backtest from 2012 (bitstamp), the sharpe ratio is 0.2.
kekratio is not everything in trading. especially not on long timeframes

>>25353112
>I've been writing on my bot as a side project for years now
...and still doesn't understand even the basics. LMAO

>>25363013
are you serious with that? topkek

its not profitable because taxes will kill you. better to stick long term

>>25363013
>3000% avg trade
>4 bars per trade
how can you take yourself serious anon

>>25363224
yes.
go cope with your 1 minute algos.

i am this guym the ip changed:
>>25360184

>>25363289

>>25363289
I'm still me, why are you pretending my results are yours?

>>25363359
i think hes just trolling in general

>>25363359
sorry. I meant this >>25359105
similar screenshots.

>>25363401
post your chart, looks like heikin ashi bullshit results tbqh

>>25363461
sure man.

heikin ashi kekresults usually give a much higher percent profitability rate.
more like 80 or even 90%.
this is good old candlesticks.

>>25363550
sweet. many trades tho, what happens when you add commissions and slippage?

>>25363595
you do realise this is the weekly chart right?

new years is actually a good time to discuss while normies are not on 4chan.

>>25363156
Enlighten me, retard.

>>25363727
Here, I will spoonfeed you: the point is not how many times you win or lose, but by how much.

>>25363823
risk reward ratio
you win 4 out of 10 times.
average loss is 5%
average win is 45%
???
profit.

technically traditional backtests are not enough when too small in sample size also.
for this you need to do monte carlo backtests as well.

>>25352888
fee structure makes it impossible to use basic indicators.
moving average by definition will be few lags behind the point of trade.

what i would call static indicators (moving averages, rsi, stochrsi, aroons etc.) will not yield any profits, and any novel methods (i.e machine learning, bayesian statistics, any filtering and signal smoothing) generate too many anomalies when used lag by lag.

personally been at it for few months now and despite shifting more than 200k a month, i've lost $200 from my original $500 trial investment.
it could be that we are at the most volatile times with the most volatile assets of them all, so perhaps im onto something, but i cat tell, being in the red so far

>>25364763
>personally been at it for few months now
What has been your strategy?

I don't quite understand your point about the fee structure - Binance's fee is like 0.1%, which if I set my sell signal at 1% profit, doesn't seem like that much.

>>25363823
No shit?! I was referring the profit/loss, not amount of trades.
>> 25362181

>>25352888
not that hard to rig up a neural network that will predict price up/down at tomorrow's opening/closing. I'm at about a 59% hit rate, but its a bitch to automate.

>>25364823
>What has been your strategy?
still working on it and wont share it until i make it myself

>I don't quite understand your point about the fee structure
see attached. fee and slippage is everything. if it wasnt for fees, simple 10MA would make you 4x money compared to buy&hold

>>25354417
That's some good autism, and I mean that in the nicest way. I'm certain there are profitable bots out there though, I think the key is more in letting the bot loose on the right thing at the right time rather than the super micro details of how it works.

Like the best bot in the world isn't going to make money on a dumping shitcoin. If you notice a general somewhat long term upward trend on eth for example you should be able to set the bot onto it and let it ride the swings up for days.

A bot that watches everything and gives you cues for manual trades would probably be profitable as well.

>>25354417
>Eventually all your manual effort will be the same as if you never had a bot and you were trading manually
this so much. after 12 months of coding and back testing and data grooming.....I'm better than bot. Really a wakeup call. Fucking confidence building though

>>25365156
Also I got into autistic swing trading for a couple weeks and learned alot watching the bots fight each other. I was watching GRT bounce off .40 I think and swung a grand between like .32 and .39 multiple times. That's like 20% in a couple minutes each time.

I would watch the bots blatantly placing up buy and sell walls to corral in the price. At that point you realize a bot programmed to basic stock market rules can never work. These markets are so unregulated and volatile that it can almost be counterintuitive sometimes, ie a guy would place a $5 million buy wall not because he wanted to buy at that price but rather so he could box the price in a certain range. You really have to consider the power of whales in these wild west markets when swinging.

>>25365228
Agreed. I run an algo daily to process all the new info automatically. Then I place orders as necessary.
Trading less frequently has brought me bigger gains

>>25365323
Normal market rules really dont apply to crypto. Its more like the chinese stock market, wild west as you say. Theres a lot more literature about the workings of chinese market than of crypto, and a lot of it is applicable. Considering most of the whales are chinese and probably cut their teeth with chink stocks.

>>25365812
What's the difference between western stock markets and chinese? I imagine their markets must be heavily regulated (aka rigged) to allow for ideal growth etc.