Frens

your /biz expectations are too high.

Attached: jan-19.png (884x663, 111K)

In how much time your bot made that?

>written in javascript
so i assume it's a pretty simple strategy then? not to computationally intense. what is it?

Don't hurt me
No more

Attached: 1554753350866.jpg (623x702, 35K)

made me zero dollars, it is a backtest.
One month without leverage.

Yes, very simple and lightweight

Attached: momentum2.png (862x605, 30K)

Ah excellent, some results. Now we can talk.
How did you attain your data? Did you just use something like cryptocompare historicals or did you data gather yourself? I'm currently doing the latter so I can get more useful information (1 min data with orderbooks separated into bins, all of which can be useful depending on the strat).

In your pic related, did you backtest a mean reversion strat or trend strat?
Also I'd check to make sure it works for many different coins. As tempting it is to stroke your own ego with profitable backtests, it's always the painful bad results we should truly seek.
Also, I can't help but notice your risk is very high in that backtest (nearly 70% drawdown), but high reward considering the profit after only 67 trades.
In general high variance is typically not a good idea when using a bot. A strat that loses more often as a result of stop losses is far less risky than an all-or-nothing approach (imo), particularly in crypto. Keep in mind, your win rate doesn't even have to be above 50% to have a profitable strategy, so long as your average profit amount per trade is sufficiently larger than your average loss amount per trade.

I think drawdown is 70 bucks on just over 1000 but then again. 67 trades laughable. No conclusion.

How do you make your setup?
Do you use a third party to keep the algorithm connected to exchange at all times?

>Ah excellent, some results. Now we can talk.

Attached: 1409868180227.png (320x263, 117K)

too ugly
need gui to sell your worthless code to tards

46 trades with a 61% win rate is not a big enough sample size to determine if it's better than random.

I have a strong feeling that a mean reversion strat is not going to work in crypto.

Oh geez, just realized I havent seen an intelligent bot thread on here in awhile.

Epic meme, sir. Well played! *Tips fedora*
(You did get a chuckle out of me)

Honestly, I just use ccxt and have a bunch of threads on 2 servers just going at it for 100 coins. I don't use any third parties. This has failed me a few times (missing data points) but it doesn't stop running when API calls fail. Maybe I'll look into third party stuff. Thanks user

just build a repainting bot that will always show perfect entry and exit and sell it online

Lad we've just been in the crabbiest market ever for months. Crypto's innate high volatility is perfect for short term mean reversion trading? Literally just use a stop loss bro

Data directly from the exchange
Trend following
Yes it works with more than one coin.
Im trying to discover how to generate synthetic data to prevent overfitting.

45m timeframe it works as a swing bot with trades spanning several days.
if you select lower timeframes you get hundreds of trades but the fees start to hurt.
programmed from the ground up.

i tested with larger samples and got similar results.

high: ATH
low: ATL of the previous year. hodlers gonna hodl

I am starting a data science internship at a trading firm next week. I will work on a mean reversion outlier trading strategy using unsupervised clustering, starting with a focus on correlation and cointegration but also with simply the time-series itself. I have been looking into it for a while now and I still have a strong feeling it doesn't work for crypto simply because basically everything is correlated to either ETH or BTC and the swings are too great and the recoveries can be shitty i.e. simply to risky to long (or short)

Good luck in the internship user. I'd still say to look at the backtests for mean reversion in crypto. If everyone in the market thinks mean reversion is a bad idea then by the efficient market hypothesis, it's possible (but not guaranteed) that it's in fact a GOOD idea

if(bloodonthestreets){buy} else {sell}

Thanks fwiend! I completely agree btw. I will try it out for crypto and I hope that it might work but I am not optimistic.

Holy fuck lol that is exactly what I was picturing as I started reading the reply

Attached: 1531175953385.png (676x593, 33K)