I want to share my experience with this bot.
I had it runing for a month aprox ( 2015-02-16 till 2015-03-14).
I had it's tick interval set to one hour.
What I noticed is that it could have made more profits if it was more sensible to price movments.
What is it that triggers a buy or sell order? Is it posible to adjust this matter to make it more sensible. I understand that this would carry more risk.
Please visit this thread for more information.
The Honey Badger is a "Low Frequency Bot" it subdivides the long chart into managable segements.
First dividing into:
Green Dragon (ma30> ma60 > ma90)
Capitulation (not red or green but after green)
Red Dragon (ma30> ma60 > ma90)
Cat Bounce (not red or green but after red)
The logic then divides each segment into 3 parts:
What to do (buy or sell) when entering the new mode
When to buy the mode
When to sell the mode
Its up to you to then fork the strategy and make "short trade decisions" in each select mode (or enhance the mode selection function).
On its own, the algo trades 80X for all the BTCe Bitcoin data we currently have on hand. As you've seen, the algo also scales down to 1h candles very well.
On my personal / commercial version I've boosted that 80X gain to 2000X+ in the same period by creating additional short period buy/sell logic in each of the 12 subclasses. The only thing that stop you from doing the same is creativity and persistence.
@sai please merge with my support thread, thanks!
"Its up to you to then fork the strategy and make "short trade decisions" in each select mode (or enhance the mode selection function)."
How would this be possible? Which parameters should I manipulate?
In my personal/marketplace script I've added several hundred lines of additional logic statements to each of the 4 main trading modes. I have not (at this time) added any additional indicators except sma3.
You'll have to experiment. Its unlikely you'll make too much improvement to the freeware just by changing constants.
I broke the long chart into 4 major modes for you. Each of those I broke into 3 subclass. Each of those subclasses can be traded more effectively with further logic refinement.
How can I copy what you have done to make 2000X?
Look at the logic for each subclass decision. Work on them one at a time, and find better reasons to buy/sell than the default logic propositions.
"How is HB describing the turning point of this event sublcass?"
"How could I better describe the subclass in terms of MA slope, convergence, divergence, and crossover?"
Consider the the timing of change; "is less than" vs "becomes less than" How long ago did it happen?
One subclass at a time... many backtests into the future... you can slowly refine the script to make better decisions. The enterprise version has no other indicators besides 12h moving average, although it does make calls to 12h High Low and Close, as well as 12hMA3 to track the MA2/MA3 cross.
I built HB framework with 12h SMA as its only indicator because it is the fastest calculation, so I can backtest over the longest periods, which gives a broader dataset to refine logic. SMA also scales perfectly from one candle aggregation to the next, whereas many other indicators (SAR for example) built in 12h timescale need to be retuned to trade in 1h.
I have also enabled communication between subclasses. I numbered each decision and I use each decision to engage "state machines" which are various types of linear and non-linear timers, or integers. An example, "how many ticks have I been in capitulation mode? What is the value of a sine, logarithmic, etc. function of that time period?"
Then, each decision class has to check with with a selection of state machines as well as moving average geometry before making a trade decision (or changing a state). The interaction of the moving averages, their derivatives; the state machines, their derivatives; and the logic tree creates an emergent situation where each decision made has a small story of interrelated and emergent "thought" behind it.