They already ran out of people with cash and zero common sense....?

They need to check under the rocks. There is always one or two in there somewhere.

Instability is good!

@ Pablo

First, are you okay? You've been pretty silent for quite some time.

Second, do you have any idea how many samples your bot takes when building the first EMA value?

According to the math here:

http://stockcharts.com/help/doku.php?id=chart_school:technical_indicators:moving_averages

To build an EMA function, the first value has to be an SMA value. I know the number of periods the EMA functions sample after the first. I just don't know how many samples are used in butter-bot for the first SMA value. My numbers come close but not quite on the money.

======================

As an aside, does the backtesting feature take into account the available volume at the time the trade occurs or does it just assume the buy and sell would have been made reguardless of the size of the order?

I ask because while programming a B-bot backtesting feature I came to realize that buying or selling 1 BTC is not the same as 100 BTC at each trade. Eventually you would not meet enough supply/demand to fulfill an order despite the EMA's threshold values being crossed.

Which might mean the backtesting feature is "blindly assuming" any value can be sold or bought at a specific time frame regardless or true demand/supply at that given moment. Which has obvious implications on the accuracy of backtesting and/or the validity of results.

Anything you can help me with casually is much appreciated!

You know, with MTGOX recently lowering the fee structure to 0% for four days, you'd think it would be very easy money.

But no matter what settings I try on Butter Bot, I can't seem to be making a positive return...

Thats right, SloK.

Give them 1980's styled heatsinks when your customers ask for modern, efficient heat dissipation techniques. Even when they are shown assemblies and were shown copper heatsinks and quiet efficient operation.

We need more people like Slok. The type of person who is there when the demonstrations are shown with 2010 technology. Then, is perfectly happy to see disappointed customers get 1980's styled (PC) tech and will even heckle and ban them for complaining. When BFL runs out of processed aluminum alloy they will probably just stick a rock on each chip and call it "revolutionary" heat dissipation.

We always need someone to pee in the coffee pot when it comes to touted expectations. You sweet, kind and totally lovable BFL whore...

You'll ban anything with any common sense.

BFL probably hasn't made any of their customers rich.

Most of their customers will probably be poor even if they started out rich.

Did you try their number in Costa Rica?

43,000

Quick take it! They still care about you! 

I am probably getting rich from the FTC\PPC rally.

LTC requires an abundance of constant attention...many have been burned across the alt coins in the past. (including FTC)

Make sure to send your console logs to Pablo so he can check it out.

I have never seen that type of error.
It has several features to disable the engine from trading.

At the engine level in general and at the level of each independent exchange.

It's probably because you are using a long period of time for updates.

I am glad someone pointed this out as this crossed my mind when I read the last page.

If you switch your charts to 1 week, it will show a near vertical climb when it reached 900, sharp climbs like that are a bad indication on sustainability. The stochastic indicators for that period also showed the market was way overextended in the positive. It was kind of obvious what was going to happen.

1 day indicators also show that the market should (for a short time) stabilize soon. Probably around 350 from eyeballing it.

We should reach 400 by 4pm, maybe 8pm.

Probably because different settings are required to make what I mentioned above a profitable scenario.

One settings is good for uptrends, others are much better for downtrends. The bot is currently too simple to make effective decisions based on more than one trend. It will do "okay" in all scenarios. Fine tuned settings for one particular type of market will make it work very good in one direction, then switch, and play a different card.

The current B-Bot requires a lot of refinement and many parameters to do what people really want it to do. Till it gets it, we will just have to babysit it every day to assess what the current scenario is playing out as. And adjusting as appropriate.

For example, I knew when it peaked at 900 that B-Bot wasn't going to react very well to a flash crash on most current users settings. So I turned it off when I noticed the market was rising too fast in 1 day and did manual trading while expecting the inevitable retracement.

In about 15 minutes it crashed 300 dollars, sure, it recovered for a number of hours, but if the timing was off, B-Bot would have cost people quite a bit of their earnings. There is definitely lots of room for improvement.

In either case, you should be encouraging B-Bot use and adoption. In the end, the more bots there are, the more volatile the markets will become.

You can increase your BTC holding, you just have to know how to set up the bot and what to expect from normal market activity.

If you had 20k at 130 USD, then at 900 USD you should have made (in fiat) about 70k to 90k (with this bot). When it comes back down to roughly about the same level as you started, your bot will then buy BTC and increase your future earnings as it swings up again. Of course this may take months and a very volatile market. The next time (assuming the same level of volatility) you should have alot more BTC than you started with. With each long term increase and decrease you should be holding more USD and BTC. If it only goes in one direction then you'd be losing BTC over the long term.

Though honestly, if you are looking for short term gains you'd have to look at a BOT that can look at different kinds of indicators and more than 4 thresholds. Hopefully B-Bot development goes down the same path as Haasonline.

(Disclaimer: EMA is probably not the most efficient route to compare to a buy and hold strategy, it is considered a "lagging indicator".)

@ Interested folks who don't quite understand how Butter Bot does things:



If you look at the above graph, this is how EMA values are read by the bot at 1 hour or 2 hour intervals.

If you look at the first arrow in orange, it shows that EMA (and butterbot) works by measuring deviations from two different EMA values and spits out a difference measurement. From this it generates a column called "DIFF". A negative (down trend) in red means that the deviation is downward, while a positive one in green is in a up trend.

If you look carefully, you'll notice that even though the deviation is growing, butter averages it out over whatever the update period is.

You can have a very large price drop and butter will show a proportion of change based on it's update frequency. A gradual price drop over a protracted period might not set off your threshold values! <--- important point

Depending on your settings you will find that butter (can and does) report a "DIFF" of -0.2% at two VERY different price points even as the market continues to drop. It is a deviation measurement as the numbers it is using are (EMA values) that change over time.

Sharp drops [depending on your update frequency]  in a very short period of time register as a larger "DIFF". While long term drops register as a very small DIFF over a larger period of updates.



=============================

A potential mistake you might make when using the bot is the following:

If you mistake the DIFF measurements as being based on price...you'd be setting up the bot in the wrong way.

So if you set a threshold value (for either buy or sell) of a very large value, [1%+] then you are instructing the bot to only react when the DIFF value it measures has a very sudden change in market prices based on the two EMA values you have chosen. (NOT based on the price)

Your thresholds determine when the bot reacts. Larger values means the bot reacts at sudden changes between updates. The smaller thresholds describe smaller and more gradual corrections [up or down].

Butter bot treats all threshold values as positive numbers. <---- IMPORTANT POINT

Placing any negative number in the threshold does NOT mean the bot will look for price drops based on price. It will default to any DIFF larger than 0.

In other words, if you use -0.2 as a buy threshold, Butter bot treats the threshold value as "any DIFF number greater than 0". So you should use positive numbers instead in any of the 4 threshold settings which will not leave you confused as to what you wanted the bot to look for.

==========================

If you are the type of person who is looking for sharp drops, you should change your update period to a shorter value. This will tell butter to update more frequently and make measurements often.

If you are the type of person who is looking for trends over many hours and days (or weeks) you should use a longer update interval. (1 hour or higher). Butter will then ignore any sharp drops and rises and focus on the long term EMA deviations.

Basically, a longer update period will cause butter to smooth out sudden changes in favor of long term changes.

Shorter update periods will allow butter to react to sudden moment-to-moment large deviations (drops and rises) with extreme sensitivity. But beware because it may pay attention to noise rather than trends.


============================

I would suggest (and only a suggestion) that if you are in a down market (as we are right now) you focus on trends. If we are in a uptrend market, then you might want to use a fast update period and high DIFF setting.