VmaPredict2.js (new)
#61
Well I don't understand the framework Gekko is build upon. I was thinking, if possible perhaps through CSS tricks to overlay a line graph with transparant background over the backrest image
That could keep most code as is but through an overlay add something new. (If it doesn't interfere with the framework ( I'm not deeply into Gekko internals and my js is just a few weeks coding. I'm more into c# c++ python not into web development. )

Those candle sticks look great though!.
Console testing is possible web would be nicer.
[....Resistance is futile...]
  Reply
#62
(10-12-2019, 06:33 PM)PGTART Wrote: Well I don't understand the framework Gekko is build upon. I was thinking, if possible perhaps through CSS tricks to overlay a line graph with transparant background over the backrest image
That could keep most code as is but through an overlay add something new. (If it doesn't interfere with the framework ( I'm not deeply into Gekko internals and my js is just a few weeks coding. I'm more into c# c++ python not into web development. )

Those candle sticks look great though!.
Console testing is possible web would be nicer.

Gekko coding isn't my specialty either,
but I was able to test your strategy with this modification of the green gekko.
https://github.com/mark-sch/gekko
One of the ways to do what you are proposing would be to submit the tests to a database.
Then we could make any web interface.
  Reply
#63
Hm yes that's an option how should I log a custom variable to a database from a node js Gekko strategy.?

That might be great though I might then be able to use python (Jupiter notebook) to analyse the data that would help me a lot.
[....Resistance is futile...]
  Reply
#64
hello,

with node js we can use sql lite; gekko uses it as Database engine.

look at this examples :
https://gielberkers.com/how-to-connect-t...in-nodejs/

it's easy to open a db (stored in a file) and insert into a table.

Dieguz
  Reply
#65
Thanks for the URL, which i had more time for this.. (then first i would write a tensorflow trader, and think about an extra nodeJS just to create graphics side by side of the same gekko database (if sqllite can handle that, it be then just two tabs)
....

This is maybe funny dont know for sure, but using the latest version and its TOML
I decided to check hte math on LTC instead of BTC at first not that much trades.
Most where pretty safe and the final result not so great.
Then i only altered DownHill and set it to 10 (since the LTC is a lot cheaper as the BTC and so 700 made no sense).
A lot more Downhill exits happened, and therefore more trading opportunities.

start time 2018-08-08 03:02:00
end time 2019-08-07 17:01:00
timespan a year
start price 67.30215
end price 88.99618
market 32.23378%
amount of trades 135
simulated profit 63.13972% (slightly more then twice the market)

Although that makes me wonder too, should i add an additional downhill fraction so both can work ?.
The reasoning would then be , get out, by trying to predict minum fixed loose or use fractional loose.

OR below xxxx use fractional above xxxx use a minimum. ... hmm (trying to be less risky later).
[....Resistance is futile...]
  Reply
#66
yeah, i think that downhill logic is the way to optimize the strategy; an efficient downhill creates opportunities of magica or utopica buys.
in my simulations i found that downhill logic is crucial and it's necessary to set that value according to the currency

for example if ltc is 48 eur, downhill could be 10 or a little more; watching the graph of the currency in the last period can help.

also highshot, daystop must be customized for the specific currency;

if i understand correctly you want to create another downhill with fraction; in some cases downhill1 is activated, in other cases downhill2 is activated.

another interesting idea may be to adjust downhill using some logic; in some period of time dowhhill may be set to x, if the market is good downhill may be decreased, if market is poor, increased; so the downhill will become dynamic; but i don't know if could work.
but downhill is the way to go.
  Reply
#67
Yeah it's weird downhill was intended as a alternative predicting loss stop safety but it seams one needs to loose money to make more of it. Maybe it should be made smarter yes checking momentum or statistical deviations.
Maybe act different on bear bull and sideway.
The rsi part was made to detect flat markets but is not so good yet..

Another thing I've noted is that 6 hour graphs on biance have a pretty good market 'guts' sense but slow. Maybe downhill ratio based upon 6h trend.

Well lots to think about..
[....Resistance is futile...]
  Reply
#68
Well i tried some alternative coding, like a % of maLong smaller then or larger than that, but so far no luck.
It currently works best as is it has to do with the (quite regular and common) price falls i think in combination with the other rules.
I've not yet found any improvement more often it got worse

ea R295 if (this.Toml.maLong *0.1 < this.Toml.DownHill )this.Toml.Downhill = this.Toml.maLong*0.1; else this.Toml.DownHill = this.settings.DownHillStop;
[....Resistance is futile...]
  Reply


Forum Jump:


Users browsing this thread: