Update On Progress V5

jason ย ยทย  2019 Jan 11

Update on progress V5

V5 has been running on my production servers since a couple days including an updated V4 to 4.13 adding slightly better crash protection, slightly sharper exits on higher volatility. I also added market orders if prices run away from us or after a certain time.

V5 made it's first production test trade with a loss of -1.19% as it entered after the pump and sold about 19 hours later as prices kept creeping down. V5 is designed to get out if nothing much is happening to prevent crashes like in the chart. CoinBase V5 example

Why did it enter after the pump?

First of all we use lagging indicators so this is normal for our strategies. We can only catch pumps if our setups already triggered a buy. In which we then are lucky but nothing we expected or predicted. Besides that, V4 and V5 both uses hourly candles. This pump, very unnatural btw, happened within a couple minutes. So this means that we will only see this pump basically when it's all ready and done.

Because of the volatility it will trigger our setup indicators for a position. If our setup is triggered during natural or over a more sustained growth across several hours/days, this is perfect and gives us a more solid base of growth. The crash happened within a couple minutes too, which would have rekt V4 with a 6-7% loss. This is the weakness of our current systems but V5 handles this type of situations better as it doesn't stay in trades for too long.

Can we solve this?

I looked at the raw minute candles data and there's definitely patterns emerging that can protect our portfolio better. A solution for this is to create a supportive strategy running on 1 minute candles. I can do real-time in a complete rewrite on the strategy execution system but 1-minute candles will do for now. To recognize what's going on it would compare historical hourly/minute data from previous days/weeks, filter out anomalies, like the pump and dumps happening now, and compare with current minute volatility and trade frequency. This would be the first step of semi-predicting market movements. It would be designed to get out of crashes before it happening, but it would work the same way for detecting potential pumps. This will need to be tested of course but it's a first step to combine our lagging indicators with predictive indicators for potential big market movements.

I can also potentially make the strategy "remember" events like this. A high risk event, as pumps within minutes are very very likely orchestrated/manipulated by few individual(s), it's very likely it will dump the same amount again if the market doesn't move the way the manipulators want to. I'll have to figure out how to solve this but probably keep the strategy in a high risk state.

At the same time we can compare minute data from minutes before and asses risk for a potential incoming pump or dump. This would be very similar to how our upcoming Binance Pump monitor strategy for alts would work.

I used to run a social media analytics company with a friend of mine for the past 5 years. It reminds me of the analyses we did there. We checked which days performed better on reaching your audience on Facebook and at which hours. We filtered these posts to only include organic posts so that the results would not be influenced and tainted by for example promoted posts. The entire page including text analysis were are all automated. It would analyse your companies Facebook page and throw out a 100-page pdf analysis of your performance and recommendations. Here are some examples: LS example 1 LS example 2

Anyway, V5 is running but I need more time to prepare the dashboard and all information to support this. I'm preparing very detailed pages with backtest results and performance it's going to be awesome ๐Ÿ‘Œ. This is one of the biggest updates as of today yet. Expect early next week for all the changes to be live and more info for our pricing models.

Once live, I will start working on the supportive strategy so that we can limit our risk from crashes like this week. This is very important to me as this is one of the biggest weaknesses of our strategies. And would keep me awake never knowing a crash like this would get us rekt. Because of this I recommend V5 even more over V4 for now, until we have this supportive strategy running. V4 will benefit more from this supportive strategy than V5. The detailed information pages will also include strength and weaknesses of each strategy.

Support for V4 and V5 for Binance will be launched hopefully somewhere next week so that anyone can use our strategies if they have no access to CoinBase. But I'm probably too optimistic about my release timeframes lol.

Development for the Binance Pump monitor will be delayed into February as my focus is now on making V4 and V5 less risky. However, since the supportive strategy will use some of the same methods to determine anomalies, we can re-use parts of this for the Pump monitor.

Both strategies are designed so that I can feel confident and safe when I have my money managed by our bot, that I will hopefully have the same-ish or more when I check a couple months later. If I have a couple $ to spare, I want to put it to work instead of idling on my back account for 0.3% interest, but at the same time, not wanting to risk a 100% hodl exposure and be at the mercy of the market. Consider supporting me once we go live with V5, so that I can keep working on making it better.

If you have any ideas to tackle markets/decrease risk, always feel free to share them with me and we can think on how we can apply this to our strategy. It's a long process to make an algo that actually works long term, but one by one will we tackle all these situations that are thrown at us ๐Ÿ’ฏ.

Jason

Latest posts