(Closed) Butter Bot!: Premier Bitstamp, and BTC-E EMA Trading Platform (Closed)

[madpoet]

The bot's not meant to trade all time frames.  Pablo's said this repeatedly.  You don't like what it does, oh well. Don't buy it.  Go write your own bot since you obviously are much smarter than all of us 

[1713434999]

1713434999

[1713434999]

1713434999

[1713434999]

1713434999

[1713434999]

1713434999

[shapemaker]

Huh?  You want me to prove that selling at a random time isn't better than what the bot does?  That doesn't even remotely make sense  Done arguing, the math to me is pretty easy.  And since I'm up a few hundred bucks over the last few days, I'm good.

No, no, that part was a supposition, what I want you to prove is your statement that the BOT actually wields profit like you said. So, yeah, just post an example for MtGox that works in all time frame and I'll send you 1BTC.
Pretty simple.

Here's one that yields more BTC over 2 years (the max time frame). Buy in was at 2.515 dollars each, so 100/2.515 = 39.76 BTC for $100.

Holding that 39.76 BTC for 2 years, then selling for $800 yields $31809. The bot manages $34287 in the same time, which also obviously is more in BTC (42.85 > 39.76). QED.

Feel free to send my BTC to the address in my sig.

Feel free to read my question that is quite correctly worded.
Post some settings that make BTC profit in ALL the backtest timeframes of Gox and you get your 1BTC.

If you fiddle with the past numbers you can always find one case where it works, thats easy and... doesn't mean absolutely nothing.

I demonstrated a working setting for the longest available timeframe, which yields more in fiat and BTC than buy and hold. This alone demonstrates your argument is on shaky ground.

However, the same settings also work for 1 year ($7827 > $6394). Feel free to do the rest yourself, it's clear by now you're gonna just nitpick your way out of paying.

Nitpick... please learn how to read: "So, yeah, just post an example for MtGox that works in all time frame and I'll send you 1BTC."

There is no shaky ground here, and you know very well the setting you posted doesn't work for the other time frames... you are the one nitpicking what you choose to post in order to avoid the argument I expressed and that you realized can't actually refute it with any numbers.

If you expect the same setting to work for both 1 week and 2 year timeframes, you don't understand EMA. Looking at 1 and 2 year time frames, the bot would have made more fiat _and_ BTC than buy and hold. 6 month profit is clearly less (but still profit), and 3 month profit is very close to b&h. Smaller timeframes don't make sense because EMA _is_ a long time strategy, not daytrading.

With that background, looks like you were just begging the question with flawed premises to begin with. EMA is a long time strategy, and at 1+ year timeframe beats b&h with a clear margin.

How about now you did something constructive? Suggest ways in which to improve the bot (in your opinion) so that this isn't just useless time wasting from your part.

[shapemaker]

honestly, you find a bot its working if you get in 2 LONG years a profit of 7% ? and thas just becuase you found some configuration that worked in the past,
that means nothing as price rised a lot,

jbssm is quite right in my opinion,

bot just need a trading engine,

7% more than b&h, which this question was all about.

I don't refute the argument that the bot could be improved. I'm sure the devs are open to suggestions. Go ahead, I'm sure you have much to share.

[Stingery64]

let me illustrate what bot does most of the time to lose your bitcoins,

1º sell "high" because there is down tren
2º wait the trend to be so strong that yelds you a profit, however almost never is that strong so it cut loses finding the up trend again,


the two events that triggers a Sell and a Buy, are always is thoses positions: "point 1 higher in the graph than point 2"
however the situation is not profitable, bot sold "high" then bough back "high" again
its always doing same thing,  it just would make money when there are really big and long down trends.

any way, Pablo is giving great support, but i also loosed lot of bitcoins with the bot

https://i.imgur.com/QQG7EwT.png

[fible1]

Huh?  You want me to prove that selling at a random time isn't better than what the bot does?  That doesn't even remotely make sense  Done arguing, the math to me is pretty easy.  And since I'm up a few hundred bucks over the last few days, I'm good.

No, no, that part was a supposition, what I want you to prove is your statement that the BOT actually wields profit like you said. So, yeah, just post an example for MtGox that works in all time frame and I'll send you 1BTC.
Pretty simple.

Here's one that yields more BTC over 2 years (the max time frame). Buy in was at 2.515 dollars each, so 100/2.515 = 39.76 BTC for $100.

Holding that 39.76 BTC for 2 years, then selling for $800 yields $31809. The bot manages $34287 in the same time, which also obviously is more in BTC (42.85 > 39.76). QED.

Feel free to send my BTC to the address in my sig.

Feel free to read my question that is quite correctly worded.
Post some settings that make BTC profit in ALL timeframes of Gox and you get your 1BTC.

If you fiddle with the past numbers you can always find one case where it works, thats easy and... doesn't mean absolutely nothing.

I think we've been quite clear that EMA is a long term strategy; so asking for settings that work in all time frames doesn't make sense. In the short term trades are pretty random (be they profit or loss), this has to do with the fact that trades in any strategy have to be looked at in aggregate (Think of the Law of Large Numbers). One trade or two trades or even ten trades don't mean much in any strategy.

You can see this in the back testing engine, if you back test 1 week, your results will change day to day from profit to loss, etc. If you back test 6 months to a year you will see pretty steady results (be they profit or loss they will remain so with the same settings over the long term).

I don't claim there are magic settings that work in all time frames, as a matter of fact, I have been pretty adamant that EMA trades, like all strategies, need to be looked at in aggregate over the long term. Do you want a real in depth discussion of this point and of EMA as applied to Bitcoin in general?

Please see here: https://bitcointalk.org/index.php?topic=60501.0

Butter is heavily based on Gomboo's principles, there you find an educated discussion on the mathematics and principles of EMA trading beyond what we can provide here.

I hope that helps :=).

Pablo.

That's not the issue. The issue is that since you don't provide custom dates to test the strategy then the only option is to ask for a setting that works for different dates in you backtesting.
If you did, it would be obvious that you couldn't find one setting that worked for the 1st 6 months (I hope 6 months are long term enough for you) of an year and for the last 6 months of the year. The settings would be different in those 2 periods of 6 months and that defeats the purpose of the bot.

Those setting don't exist and I know it because like some user suggested I downloaded the data from the bitcoins charts site and I was testing it, and guess what... what every trader in the financial market knows... there are no EMA (SMA, WMA) crossing strategies that improve your shares capitalization in different time periods.

So please don't tell us this long term strategy buzzword as an excuse, cause you are just using it to mistaken the people.

Hey ,
  It's not a "buzzword", it is a long term strategy, it's also a trend strategy, not a High Frequency Trader (HFT). The fact that you don't like the word doesn't change that EMA will work best over the long term for the multitude of reasons described.  

I frankly don't see your issue with time frames as we have provided them, they back test your periods to the day, so if you back test for 6 months today, it will back test 6 months to the day, if you back test 1 month from now, it will back test 1 month to the day so there really is no way to game this as we can't control what markets happen in those time frames.

Now I like your idea of including custom time frames, that's something we could definitely look at for the next version, I mean that, there really is no issue.

I agree that the strategy can be improved, we are working on a new engine, because all strategies can be improved; you are however ignoring several people who posted that are making and have made a profit with EMA; it's not fair to pick and choose what data you want to accept in your argument and what you will ignore.

I agree that the strategy does not guarantee profits for everyone at all times, if you find such a strategy (which doesn't exist), you will be posting from your LearJet; what I can tell you is that: EMA will tend to generate more profit than not OVER TIME. Goomboo has made this argument quite successfully over 50+ pages, so it's not like we invented.

I again encourage you to review Goomboo's thread, he is the absolute authority on this issue and I think there is a lot you could gain from reading it. I mean absolutely no offense, I think that if you took the time to read about the strategy, maybe peruse Goomboo's thread and post there as well you would gain some insight on something that other people are telling you has worked for them and something that has been discussed ad nauseum in one of the most respected trading threads on this forum (Goombo's thread).

EMA as Applied to Bitcoin (Goomboo): https://bitcointalk.org/index.php?topic=60501.0



Pablo.

[shapemaker]

the two events that triggers a Sell and a Buy, are always is thoses positions: "point 1 higher in the graph than point 2"
however the situation is not profitable, bot sold "high" then bough back "high" again
its always doing same thing,  it just would make money when there are really big and long down trends.

That is because your settings are too aggressive. It is not a daytrading bot. The whole idea behind EMA based trading is to smooth over such quick disturbances, and only react on clear up- and downtrends.

[Stingery64]

the two events that triggers a Sell and a Buy, are always is thoses positions: "point 1 higher in the graph than point 2"
however the situation is not profitable, bot sold "high" then bough back "high" again
its always doing same thing,  it just would make money when there are really big and long down trends.

That is because your settings are too aggressive. It is not a daytrading bot. The whole idea behind EMA based trading is to smooth over such quick disturbances, and only react on clear up- and downtrends.

my friend, you are wrong,
is just an illustration, you can see the test by yourself in the bot, 60-70% of the trades return a lower % if the original sell.
thats becuase it sell high, and buy high...

its easy, it just lack a simple engine to buy low when the chance is there,

needs diversification and buy trigger rules.

its.
so.
easy.
-

[shapemaker]

the two events that triggers a Sell and a Buy, are always is thoses positions: "point 1 higher in the graph than point 2"
however the situation is not profitable, bot sold "high" then bough back "high" again
its always doing same thing,  it just would make money when there are really big and long down trends.

That is because your settings are too aggressive. It is not a daytrading bot. The whole idea behind EMA based trading is to smooth over such quick disturbances, and only react on clear up- and downtrends.

my friend, you are wrong,
is just an illustration, you can see the test by yourself in the bot, 60-70% of the trades return a lower % if the original sell.
thats becuase it sell high, and buy high...

If your settings make the bot react to such a small and insignificant dip (time wise), your settings are wrong, not me.

its easy, it just lack a simple engine to buy low when the chance is there,

needs diversification and buy trigger rules.

its.
so.
easy.

Looking forward to your bot (or your suggestions even how to code that feature!).

[jbssm]

Hey ,
  It's not a "buzzword", it is a long term strategy, it's also a trend strategy, not a High Frequency Trader (HFT). The fact that you don't like the word doesn't change that EMA will work best over the long term for the multitude of reasons described.  

I frankly don't see your issue with time frames as we have provided them, they back test your periods to the day, so if you back test for 6 months today, it will back test 6 months to the day, if you back test 1 month from now, it will back test 1 month to the day so there really is no way to game this as we can't control what markets happen in those time frames.

Now I like your idea of including custom time frames, that's something we could definitely look at for the next version, I mean that, there really is no issue.

I agree that the strategy can be improved, we are working on a new engine, because all strategies can be improved; you are however ignoring several people who posted that are making and have made a profit with EMA; it's not fair to pick and choose what data you want to accept in your argument and what you will ignore.

I agree that the strategy does not guarantee profits for everyone at all times, if you find such a strategy (which doesn't exist), you will be posting from your LearJet; what I can tell you is that: EMA will tend to generate more profit than not OVER TIME. Goomboo has made this argument quite successfully over 50+ pages, so it's not like we invented.

I again encourage you to review Goomboo's thread, he is the absolute authority on this issue and I think there is a lot you could gain from reading it. I mean absolutely no offense, I think that if you took the time to read about the strategy, maybe peruse Goomboo's thread and post there as well you would gain some insight on something that other people are telling you has worked for them and something that has been discussed ad nauseum in one of the most respected trading threads on this forum (Goombo's thread).

EMA as Applied to Bitcoin (Goomboo): https://bitcointalk.org/index.php?topic=60501.0



Pablo.

Sometimes I feel I'm speaking Chinese  

I know it's a long time strategy. But like I said, just try this. Try the bot for 3 or 4 different 6 months periods, they can even be overlapping in some part they just cant be exactly the same.
Now, the setting your find for the bot that will give profit in one, will not give profit in another, they will make you loose BTC in other time frames! Equally LONG time frames!

What you are doing all over this thread is cherry picking some settings that work for a very specific time frame. It's doesn't mater if it's a long or short time frame, you are cherry picking in every example you give.

I read Goombo's before I came to this one a long time ago before I was even a member of this forum. And the thread is also clear if you look at the numbers. You are loosing BTC by using the EMA crossing strategy, but yes, you are only making money because the market is bullish
Same goes for the analysis of EMA strategies you can find about other markets. They give you profit in fiat, only! What they manage is to avoid major losses. So, those people invest in lot's of different products, and since the overall tendency of the market overal several equities is to go up... they cut the losses with EMA and make profit from the ones that actually go up.
This is what's happening with this bot.

[Casimir1904]

Hey Guys ,
   I think it's important to understand that Butter makes USD in an uptrend and BTC in a downtrend. That’s just the nature of the EMA strategy. We had some discussions on this early on but I think it's worth mentioning again that the thread has grown, I will probably put this in a couple of places of prominence to make it clearer.

Pablo.

Feature request:
1. Option to choice between usd-btc or btc-usd trading
2. negative sell/buy threasholds so its possible to guess for comming up or downtrends ( risky part and not for beginners for sure maybe something like advanced options?)
3. Backtesting with showing BTC value.
4. Longterm and short term trend detection to automaticly adjust settings or semi-automated based on data of last day,last week, last month.

With feature 1 everyone can decide for his self how to trade, more btc or more usd ( as long btc is in uptrend in longrun option 1 makes more btc and usd ).

[Stingery64]

just a test to see how many times the EMA lines triggers a buy to buy cheaper than you sold at:

-------------------------
12 buy trades in 2 month

10 of them it bough higher than it sold.
no sense strategy.

As some smart guys already notice, bot make you lose your bitcoins slowly

thats a 83% fail ratio of trading.

https://i.imgur.com/uL1au4w.png

i am waiting impatiently to see the back testing of the proposed changes

[shapemaker]

just a test to see how many times the EMA lines triggers a buy to buy cheaper than you sold at:

You still don't understand the idea behind the bot, do you? If it detects it exited at a bad position, it will buy you back with as little loss as possible. Also, the bot isn't affected by emotions, so it will do the right decision even if you couldn't/wouldn't.

Even if the bot makes 8/10 "bad" decisions, the bad decisions are small. The profitable good trades will hopefully cover the losses. Besides, no trading strategy is foolproof. If you had one, you'd already be a billionaire.

12 buy trades in 2 month

10 of them it bough higher than it sold.
no sense strategy.

As some smart guys already notice, bot make you lose your bitcoins slowly

thats a 83% fail ratio of trading.

You don't even let us see what settings you used. Even so, the bot did better than buy and hold, so it wouldn't have "lost" you anything.

100/122,25 = 0.81799 BTC

0.81799 * $696 = $569

Bot's ending balance = $655. How's that a loss, or fail? I agree, the bot could have done better, but you used a short timeframe.

edit: stupid mistake

[Stingery64]

i used a short time frame so it dont take 300 minuts to checkhow many fails there are on 200 trades over 2 years.

i don't expect to have 0 fails, i expect it to try get profit if there was a down trend, and it don't.
if not, then cut loses buying back is what should be done,


its better than buy and hold? well... due the way it trades, responsability relays into how big market drops are.

you start with 100$  at 122,888 = 0.813 BTC
you end up with 0.900 btc


i guess you are right, you won 0.087 btc in two months,


if you buy and hold, at the end of those two month you wil have 0.813 btc to buy at 722.367  = 587,284371$
if you use the bot you have 0.9 btc to buy at 722.367 = 650,1303$
 

587$ to 650$ = 10% win

now those are real numbers.
quite far from the "593.21%" the bot says due the way its calculating the profit (price rises)

bot should show % win based on how many btcs it won, and not in the price rise of bitcoin, as it return the fake feel that thnks to the bot
you won "593%"

in 12 trades, 10 are loses and 2 are win, thanks to those two wins, bot was able to recovery the loses,
but if there was no nice scenario were price drop enough to get a big enough win that recover the loses, then it would keep eating your bitcoins slowly,

a strategy to win something at each drop is the key to make this bot a real winner and not a wasteful of bitcoins software.

sure you can check the setup i put for the test, is the one that return 13 000% "profit" in one year,
https://i.imgur.com/9KfsOTy.png

[shapemaker]

587$ to 650$ = 10% win

quite far from the "593.21%" the bot says due the way its calculating the profit (price rises)

Regardless of the strategy, the numbers are real. You started with $100, and ended with $655. The profit for that indeed is 555%. Now, that is "only" 10% better than buy and hold, but show me any other investment that will return 5x profits in 2 months.

For that matter, 10% profit in 2 months is by itself very good, for that translates to 60% profit a year which in stock or forex markets is amazingly high. And that is in addition to whatever profit there is in BTC price rise.

[Stingery64]

sorry but no matter what you say, if the bot sells high and buy even higher almost always, then the bot is dumb, its a fact, numbers don't lie.
the profit is misleading based on the marke price rise.

This extra 10% was just a lucky configuration based on past results, we talk about the way bot operates. Market itself will mutiply by 5x, so you dont need to risk into loses using the bot if you just look at this way.

let's see if with the proposed changes Pablo can backtest as soon as possisble and we can take a real profit using it when a drop takes place, as i can make better decisions manually.

[shapemaker]

sorry but no matter what you say, if the bot sells high and buy even higher almost always, then the bot is dumb, its a fact, numbers don't lie.
the profit is misleading based on the marke price rise.

I will gladly take 10% profit over buy and hold over 2 months.

This extra 10% was just a lucky configuration based on past results, we talk about the way bot operates. Market itself will mutiply by 5x, so you dont need to risk into loses using the bot if you just look at this way.

Market does what market does, even if it's crashing to 1/5 of the current size. You can't rely on BTC price increasing all the time. As soon as BTC price starts to tank, buy and hold won't look that hot anymore, no?

let's see if with the proposed changes Pablo can backtest as soon as possisble and we can take a real profit using it when a drop takes place, as i can make better decisions manually.

That I agree with. There definitely are many possible places one could squeeze additional 1-2% profit in daily fluctuations that don't evolve into full trend reversals. As long as the bot can react quickly enough, that HFT aspect could well be worth the while.

However, there must be an option to set a limit to how much fiat/BTC is allowed to be risked in these HFT trades, separately from the current "reserve" setting. I'm pretty sure you wouldn't want the bot accidentally wiping the week's profits in one miscalculated HFT cycle.

[Stingery64]

well its not a 1-2% extra profit, its going to be MUCH more, and its neither an HFT strategy, is just as easy as check if bitcoin price already droped to an enought % to worth a buy and take profit, instead of just follow the trend back, because doing it, then 90% of time make the bot "sell high - buy higher",

[vesperwillow]

I see many things being repeated over and over in this thread. Many of them are suggestions. There's also a lot of the same conjecture from many pages ago, namely "the bot sucks i made a loss" etc.

EMA isn't perfect, but it does work. It's going to need time to get through the trade losses. I honestly don't suggest letting it dabble with mtgox right now. The market is simply too volatile right now, especially mtgox. I shut off my mtgox trader, but have made profit with btce the past few days. Yes, I have made BTC with btce this past weekend.

Some folks don't like it, don't yell at the developers or other people: make one yourself, or simply suggest enhancements to this one. Why all the fuss? Don't trade with what you're not willing to lose.

While many of the recent suggestions for improvement are really great suggestions (some of them myself and others have suggested awhile ago), for the time being folks need to work within the means of the bot. Keep the suggestions flowing.

[manort]

Hi folks,

Interesting product. I have been trying it out for the last few days and I'm afraid that I can't say I'm impressed, either by the performance of the bot or some of the arguments on the board.

I would give this a second look if:
-I could configure the profit window to show BTC profit/loss.
-I could filter out the rise of bitcoin as a measure of the bot's success. I want to see how successful this thing is over-and-above a buy-and-hold strategy.
-recommended and configurable settings for high and low volatility markets. It would be great if the bot could self-detect the change in market conditions and adapt.

A bot that loses me BTC while increasing my dollar holdings is only selling off BTC and calling it profit. I could do that without a bot.

Pablo, you have been very active on the boards and I appreciate that. I would be very willing to pay for a bot like this if I could see that it works with a reasonably high likelihood of success in increasing both my BTC and dollar holdings. In that case, I would pay because the thing would be paying for itself.

There may be some arcane set of settings that will achieve this, but to be honest, one reason to have a bot is that it can do it for me. If I have to work that hard, it's probably better that I just go back to staring at the charts.

I'll probably keep playing with this for another few days and then I am planning to post a review on my website. Please contact me if you have any relevant information.

Cheers,

BTCWarrior

[vesperwillow]

Something I haven't seen people talk about, I think I will remind everyone about: Keep an eye on the Minimum Profit column of backtesting.

Find settings which minimize this, it will keep your BTC in your pocket. Ideal settings give you things lower than half a percent, for example.