This part also worries me.
Someone asked me an apt question when I read back this portion of your reply. I was asked,
how does the bot know that 144 points in the past was a relevant (or even a good) starting point for the future [decisions]? I couldn't help but shrug as they asked me that because it is a very good question. If the Bot doesn't use the highest recent point as the entry point (but ? apparently) just the 144th entry point from "now", it can apparently make....pretty bad decisions.
The point that came to mind is asking if there is any way to add a feature where:
--A) We can see what that entry point actually is at any given time.
and
--B) Change the length of that entry point to what we want. (72 previous points instead?)
3. Trailing Stop Loss takes the highest value Bitcoin has achieved since you turned on the bot and will SELL all your BTC if the value goes down from the maximum by the percentage you have specified. So it really was following it's programming when the value changed in your favor, by design, it will only intervene when there is a loss. This is one of those features that is very sensitive to set and that you have to keep an eye on though.
First question, does this field accept a negative value?
Second question, Does this rely on the mystery 144th point in the past? (or when I actually turned on the bot?)
Perhaps a bot with internally tweakable settings might allow even more customization.
You are thinking about EMA in the wrong way, in a perfect world you would actually want all data.
The way EMA works is that the previous interval is more important than the the interval that came before it, that is because you are only interested in trends that are currently happening / might happen and not trends that happened in the past. If the bot would not look at any historical data it would not be able to calculate if there is a trend for after a while (gathering data from the start which becomes old data after a while).
If 2 people (A and B) would run a bot that doesn't look at any past data and A is running the bot for months and B just boots it up, only A is able to determine if we are currently in a trend and B wouldn't because you can't just look at the current price to say wether the price is going up or down (you need to compare with past data).
The beauty about EMA is that the relevance of each interval is exponential decreased. This means that
all data points are reflected in the current EMA calculation, however they are getting exponentially getting less relevant. The number 144 is was in the free bot this bot is based on, and is a nice combination between having enough past data (having ALL past history would only help you a tiny bit in better calculating the EMAs, since old stuff of exponentially less important) and not spamming the Mt. Gox API to hard / having to wait to long for the bot to start.
There was a discussion in the old thread (for the free chrome browser bot this bot is based on) on what this number should be and it was changed from 20 to 144. In most other bots this is configurable, but you should never change this number except when you know that lowering this number will make your EMA less accurate.