I'd be interested, but I don't have an Android phone. A desktop app would be a nice addition, personally. How many coins do you cover and what is the basis of your data?
@titusdecali Currenty we cover most of the BTC and USDT markets in binance. We update the coins every 2-3 weeks.
For our model we use all historical high, close, low prices (since late 2016-early 2017) and volume for each BTC and USDT market for 1m,5m,30m,1h,4h,1d,1w.
Data is gathered from different sources and exchanges, and since late 2017 most of it comes from Binance.
We're planning to add a desktop version in the future, but we will let our users decide what would be the next feature: https://forms.gle/Nj6gEv5Cdj9iAtG96.
@pawel_blazejewski Hi Pawel, you can check our FAQ: https://wrabbit.io/faq.html We use pattern matching algorithms and time series clustering. Our inputs are VWAP, High, Low and Close prices and also propietary indicators.
Wrabbit algorithm is not magic, as it's impossible to give an exact price forecast for any asset. Nevertheless, Wrabbit can give you insights about "how likely" the trend will go up or down.
WRABBIT IS AN ANALYTICS PLATFORM. IT IS NOT INTENDED TO BE TRADING OR INVESTING ADVICE. WE DO NOT RECOMMEND ASSETS TO BUY OR SELL, WE PROVIDE A PLATFORM TO ASSIST YOU IN MAKING YOUR OWN DECISIONS.
@pawel_blazejewski@alan_garcia_granados Sure, but even the trend analysis needs some kind of evaluation as otherwise, there could be just random indicators where prediction is never nowhere near the actual trend predicted. So for example for those calls where the trend is bullish let's say with 80% of confidence, 7 out of 10 times the prediction was correct for a given time interval.
@pawel_blazejewski@alan_garcia_granados Sure, I get it. We have evaluated the performance of the model using the avg percentage change of the prices so we infer the trend from that (positive or negative).
If the avg trend of the prediction matches with the actual trend then we count that as a success.
The accuracy depends on the symbol, but on average we can say that it's about 63-66% percent GIVEN that SPECIFIC METRIC. There are other ways of evaluating the performance of the model (such as the MSE if it's a regression but as I've said, we're not trying to forecast exact prices).
Other methdos include price barriers and a timeout. The model is constantly evolving towards making better predictions.
Because prices/predictions can be very volatile we have included the most relevant timeframes in a single screen, so you can see the big picture as a trader.
Quidli
Wrabbit
Quidli
Wrabbit
Stocksignal
Wrabbit
Stocksignal
Wrabbit
Wrabbit
Wrabbit