Algo trading strategies which actually work?

>>3485731
whatever

>>3485741
Thank you for your insight but it's not very helpful.

its not a meme, its just not easy, if it was you wouldn't be posting on 4chan

>>3485731
Price action

>>3485798
Don't look for reversals, look for strong trends and buy/sell if you are at least 80% sure it's gonna hold for a while. Use trailing stops with distance depending on the stability of the trend and don't trade during the news.
Scale your lot size accordingly depending on your capital.
Trade only tue-thu and only ranging markets.
That's all you have to do.
If you won't be able to convert all of this into an algo bot you're retarded and shouldn't be trading.

>>3485731
bots are not very good when it comes to manipulated prices, scams and ponzi schemes..

>>3485865
this.

bots are about trend follow or arbitrage, dats it, try finding some tradingview trendfollow strat and test it on every coin/timescale and find if it workout

>>3485865
>Bonus round
-Don't use take profit
-adx with Bollinger bands and daily support/resistance is the best
-if you manage to get >60% w/l consistently you're basically rich

>>3485865
Honestly I did pretty much all that. The biggest problem I have right now is to determine where should a trade stop reliably without killing eventual profits. It's easy to control losses when you're already in the black, but all methods I tested tend to fuck up on fake breakouts or sudden candle direction changes (which later got corrected, so it made me lose where I should have won).

I lost around 2% in the last few months, which probably means that means that my bots are not inherently 'terrible', but that's still a loss.

>>3485913
That's interesting. Where does daily support/resistance come into use? Should I have bot only make trades when candles are breaking through/bouncing off them?

>>3485964
>Should I have bot only make trades when candles are breaking through/bouncing off them?
That's what I try to do, but the hardest part is to detect a support or resistance. I had a working method in the previous bull-market, but in a bearish it turns out to be bad!

File: 1496790957085.jpg (925KB, 3000x4080px) Image search: [Google]

925KB, 3000x4080px

>>3485904
problem with trend following is low win rate, sometimes even lower than 30% which is means a strategy is a doomed one. all odds are against you and your back test is based on luck if that low. it means the strat have been hitting the long runs enough to be profitable, but that low rate will result sooner or later disasterously long losing streaks, 10-30 trades in a row losing etc. it's then all about luck there is no edge in the strategy. some inexperienced people might say the big wins make up for the frequent losses, that is complete false

preferably over 70% win rate can be a winning

>>3485990
I sort of used a support/resistance system based on a really basic method of just 'scanning' the chart starting at higher highs and stopping at lower lows (with a lookback of 24 ticks) and marking a line when there was a high accumulation of rejection points at an 1-2pip of height area. It was actually pretty reliable, but didn't help much with the issues I've mentioned earlier.

Now to think of it, if you're talking about setting resistance/support lines based on 1D timeframe and then using them in intra-day trading. I believe you could use the same method but on days with a look back of 1 week or something? I guess it could actually more reliably show you entry points, but can't tell for sure.

>>3486095
>Problem with trend following is low win rate
Wat
>Hurr durr I'm a trend reversal guy let's just bet on this peak and see if it will go back
Vs
>Wow this thing is really starting to go up/down I really should jump on it

Which one do you think has the higher win rate

File: 1480462872017_IMG_20161126_172909.jpg (267KB, 1920x1920px) Image search: [Google]

267KB, 1920x1920px

>>3486225
dont even try to argue you have clearly no idea what you are saying. and you dont "bet" anything when you're trading

>>3486225
Well to be honest he's not wrong. For some time I ran Gekko that was just following trend. And after I looked on what it did it appeared that when it was winning it was pretty much based on luck.

Peaks usually at least predict that something is going to happen.

Maybe entering a trade by following trends and then leaving it at peaks would be somewhat reliable strategy, hmm.

Look up "Scalping".

Go to a 5min chart for bitcoin. Apply bollinger bands with 20 sma.

You'll see what all the bots and traders are doing.

Patterns of the past simply dont predict the future. If a whale decides to take a 10 million position or dump 20 million on the market, there is literally no way for you to know that except him telling you.
And therefore TA sucks balls. Those professing it might have gotten lucky in the past and there is some surviorship bias going on.

>>3486282
You don't need to worry about where and how to leave a trade, just fucking use trailing stop.
>Pro:
You can follow a trend indefinitely if it keeps going up, thus being able to close trades in the 500+ pips range if you're lucky
>Cons
After trend reverses you will hit your stop loss not at the peak but a bit below thus not making the maximum amount of money, boo hoo

Put 5% of your money on black. If the balls hits red or zero, double your bet on black for the next spin. Continue doing this until it hits black, and then start over.

>>3486443
Problem with trailing stops it that there are too many cases where trade goes slightly into the green and then it turns around to never go back. Lots of pros also fall for that, unfortunately, judging by the traders I was following on some platforms.

So while trailing stop is probably OK when you're extremely certain of your bots performance, I never got to a level where a simple trailing stop loss would be good enough.

>>3486424
>Look up "Scalping".
>Go to a 5min chart for bitcoin. Apply bollinger bands with 20 sma.
I've used scalping strategy based on bollinger, stoch and long T3 to detect trend changes (usually when candles went through it), but it had around 40% success rate.

>>3485865
What's wrong with using reversals?
I've been daytrading for a couple of weeks now and it seems to work pretty good. Only times i really got burned was because i didn't set a stop loss and some coin crashed based on news.
Just looking at 5m candles and RSI i've done okay considering how new i am to this

>>3486535
Read about 'divergence' and have a panic attack.

>>3486442
Of course it's not right all the time, but even being right 51% of the time makes you a profit. Well maybe not with fees, but you get the idea

>>3486552
Just did, what's supposed to be to scary about it?

File: Neo_SmartContract_Competition.png (154KB, 748x458px) Image search: [Google]

154KB, 748x458px

>>3485731 (OP)
I don't think the market in the last weeks could let you draw good conclusions

>>3486424
I wrote a bot who does this (execept my bot moreover continuously looks at the last few hours and adjusts the band parameters to the values that would have resulted in the highest gains, i.e. adopting to the market) but desu.
I've become a bit disillusioned, however, as well. Say you got 10 BTC and let the bot use 10% of that to buy Litecoin in a dip. It goes up by 3% and you sell. Let's say that took 2 days.
Then you locked up a few k capital for two days in a risky market to make under hundet dollars. And you can only take 10% of your capital a few times and you have to know on which markets to play.
If I put some BTC into OMG or MTL or whatever and wait for the next news, this will probably make a similar return.
I've made a lot of money betting on the right horses in this sphere, but only a few bucks with the bot.
But that's just my experience.

>>3486610
Maybe you could use your strategy at CFD brokers and play with high leverage? That is of course if you're certain that your strategy works well, because higher profits also mean higher loses.

How did you approach the 'adjustment' you mentioned? Personally I would just make Keltner's channels based on the last whole day to see how things moves (and then maybe buy or sell on cross?), but since you didn't mention them I assume you had a better idea.

I think maybe I'll need to revise this whole scalping business if it's actually possible to make it work.

>>3486618
Well, it pretty much means that "RSI either works or doesn't work, but when it's not working we're still pretending it does instead of just facing the facts".

>>3486713
Of course all of these indicators are not perfect.
Doesn't explain why trading reversals is bad in general

>>3486713
Heh, divergences are signals on their own.

>>3486885
When put together with other indicators, I agree. But just by looking at RSI you're going to get burned pretty fast.

>>3486713
I just have a simulation-flag in the routine that, when true, doens't query price data from a market but from a recorded list. The algo runs on the market and then periodically takes a tiny break to run itself on the recorded previous prices and passes the optimal paramerters back. I'll put it back on once we're in the bullmarket again.
But you know there are some statistics that 90% of traders would make more money from just holding, and I think this applies with crypto in particular.

>>3486524
That's not a problem of trailing stops, that's a problem of stop loss set too close to the open price.

>>3487026
How so? That's not really the case usually, around 5% of entries my strategies made would lose 100% if they could, you can't wait forever for the trend to reverse. I mean, you could, but that's somewhat unrewarding.

>>3485731
>murdering your potential profits.
anon thats just greed pure an simple

if youre already making 5% dont think that you should have made 10%

>>3487247
That's not the case I'm talking about, it's about it being at a huge loss despite everything... which then inverts. Instead of a big win you get a lose if you stop. But if you leave the stops too far, hoping for reversal, and it doesn't, then you have a huge loss which is hard to recuperate with winnings.

>>3487091
You can't think of any creative ways to prepare for the scenario you've given?

>>3487778
There is no way to prepare for this scenario, except for maybe having extremely tight stop losses and expecting that winnings will outnumber them.

In the long run 1H-1D trading based on technical analysis doesn't work, at least none of the methods I tried worked in the long term.

>>3487968
NEVER use extremely tight stop losses. As I said, base them depending on support/resistance lines or fib lines, whatever you prefer.