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

[Grinny]

Grinny, if you are looking for settings that are applicable to moving forward in your trading strategy, I would suggest that crypto is evolving, and so are it's trends.  6 months ago is not an accurate picture of where the currency is now, or where it will be tomorrow.  I would look closer at the numbers when evaluated over the last two months to formulate your strategy going forward.

Sounds legit. But who knows if the trend that was established the last 2 months is the most successfull one for the future?
There are so many unknowns (mtgox, diff, ltc, fed. restrictions, silk road etc. - you name them) in the btc calculation... perhaps the last two month were the most untypical regarding the price trend - who knows?

In general: Isn't EMA about long term trading under the same conditions?
And furthermore: the given tools only allow 'us' (butterbot customers) to look into the past.
So when Pablo speaks of internal back testing, i assume he uses the same methods we use? --> 'run test'
Otherwise: what's the point about the back testing feature at all?

[fible1]

Grinny, if you are looking for settings that are applicable to moving forward in your trading strategy, I would suggest that crypto is evolving, and so are it's trends.  6 months ago is not an accurate picture of where the currency is now, or where it will be tomorrow.  I would look closer at the numbers when evaluated over the last two months to formulate your strategy going forward.

Sounds legit. But who knows if the trend that was established the last 2 months is the most successfull one for the future?
There are so many unknowns (mtgox, diff, ltc, fed. restrictions, silk road etc. - you name them) in the btc calculation... perhaps the last two month were the most untypical regarding the price trend - who knows?

In general: Isn't EMA about long term trading under the same conditions?
And furthermore: the given tools only allow 'us' (butterbot customers) to look into the past.
So when Pablo speaks of internal back testing, i assume he uses the same methods we use? --> 'run test'
Otherwise: what's the point about the back testing feature at all?

Ok, so bottom line, I made a mistake; I forgot to include the new EMA values with the differential starter settings. The values have been modded on the post and here they are again; please accept my apologies for this.

Values:


We recommend the following starter settings for Mt. Gox:

Short EMA: 10 Long EMA: 21
Buy 1: 0 Buy 2: 0.25
Sell 1: 0 Sell 2: 0.25
2 hour time window.

The following settings are for Bitstamp and BTC-e:

Short EMA: 20 Long EMA: 42
Buy 1: 0 Buy 2: 0.15
Sell 1: 0 Sell 2: 0.1
30 minute time window.

Please let me know if you have any questions.

Pablo.

[mboehler]

How do you set different settings for the different exchanges?  I thought the settings you put in affected all of the exchanges at once.

[fible1]

How do you set different settings for the different exchanges?  I thought the settings you put in affected all of the exchanges at once.

Hi M ,
  You are correct; although the next full release will include independent settings for each exchange, the current one does not, however the recommended settings are still valid if you trade by groupingthe low or high fee exchanges and ignoring the others.

I hope that answers your question .

Please do not hesitate to let me know if you need anything.

Pablo.

[Grinny]

I found this interesting post from oda.krell over at Goomboo's Journal Thread:
https://bitcointalk.org/index.php?topic=60501.msg3106359#msg3106359

[...]
EMA20+10 is dead. At least with hourly interval size.
One of the best *hourly* parameters I found during my search: EMA28+18, treshold -0.4/0.4
Good *daily* parameters: EMA20+6, EMA16+4.
[...]

This i.e. "EMA20+10" means: Short EMA 20 / Long EMA 10
And the Threshold "-0.4/0.4" means: Buy: 0/0.4 Sell: 0/0.4
Right?

What do you think of his settings?

[fible1]

I found this interesting post from oda.krell over at Goomboo's Journal Thread:
https://bitcointalk.org/index.php?topic=60501.msg3106359#msg3106359

[...]
EMA20+10 is dead. At least with hourly interval size.
One of the best *hourly* parameters I found during my search: EMA28+18, treshold -0.4/0.4
Good *daily* parameters: EMA20+6, EMA16+4.
[...]

This i.e. "EMA20+10" means: Short EMA 20 / Long EMA 10
And the Threshold "-0.4/0.4" means: Buy: 0/0.4 Sell: 0/0.4
Right?

What do you think of his settings?

Hmmm,
   It's certainly an interesting post and his back testing results he posts for the short/mid term are pretty good. He's making an assumption for all of this however that I tend to disagree with. He is discounting long term back testing for several reasons. I disagree with that as his reasoning seems to be that the older data represents pretty unique market conditions and that only more recent data has predictive value.

I don't agree because EMA as a long term strategy necessarily has to be back tested over the long term, the 3 month frame he is using is a functional but bare minimum window. The longer you back test for the more likely your results will have a predictive value as it takes hundreds of aggregate trades to cover unique trades that radically deviate from the mean. In the short term your results are most likely the result of mean deviators, be they positive or negative.

To deal with this he is manually discounting mean deviators (April 10 deal he mentions); you can't do this artificially, pick and chose your time frames as these sorts of deviate situations will arise, you have to look at them in aggregate to get reliable results (which he disagrees with). I am not discounting what he's saying, I just think he's using bare minimum windows of time which won't necessarily reflect accurate predictive results (and he's also kind of picking them around events). Still it's a good analysis.

This is one of those things that you have to make your own mind up on. I feel that a lot of what he says has merit and I think that it's (mostly) a reasonable argument. I'm not sure that I would personally use his logic in my personal trading however. I repeat, its not that the logic is blatantly wrong, it's just that its debatable and I personally fall on the other side of that debate.

[NginUS]

Hi,

I have a suggestion for the ButterBot developers that I feel would be a great benefit if implemented. It's automatic dynamic optimization, and here's how I envision it working.

The bot would be constantly backtesting in the background, in 1 month increments moving backwards by 1 week at a time, taking snapshots of the pricing curve & testing every possible options setting against that curve to determine the most profitable settings for that given snapshot, or preset of market condition. Then it would move back a week, & do it all over again- over & over.

Eventually the bot would have determined the most profitable settings to use in any given market condition, and could then keep on the lookout for a week's worth of pattern change, and switch the settings to match the most profitable preset for that week's market conditions as if they were going to continue all month.

Viola! Self optimizing bot.

What do people think?

-NginUS

[PuertoLibre]

I keep getting this error.



All my MTGox Api settings are correct. I recently reinstalled the plugin this morning.

[PuertoLibre]

Bug when hovering over icon:



Any open spots for a more recent beta test?  

[fible1]

I keep getting this error.



All my MTGox Api settings are correct. I recently reinstalled the plugin this morning.

Hi PuertoLibre ,
    The main reason you would get this error is that the API keys are entered improperly of or that you have not authorized the API keys on the Mt. Gox exchange. lease double check both these situations and drop me an email at Pablo@Butter-Bot.com; I will be happy to help you debug.



Pablo.

[fible1]

Bug when hovering over icon:



Any open spots for a more recent beta test?  

Noted, thank you, we will look into this ASAP.

There should be a Beta coming up in a few weeks, if I recall correctly you are on the Beta testing priority list, just drop me an email when you see the Beta Testing announcement and I will set you up with the latest version .

Pablo.

[fible1]

Hi,

I have a suggestion for the ButterBot developers that I feel would be a great benefit if implemented. It's automatic dynamic optimization, and here's how I envision it working.

The bot would be constantly backtesting in the background, in 1 month increments moving backwards by 1 week at a time, taking snapshots of the pricing curve & testing every possible options setting against that curve to determine the most profitable settings for that given snapshot, or preset of market condition. Then it would move back a week, & do it all over again- over & over.

Eventually the bot would have determined the most profitable settings to use in any given market condition, and could then keep on the lookout for a week's worth of pattern change, and switch the settings to match the most profitable preset for that week's market conditions as if they were going to continue all month.

Viola! Self optimizing bot.

What do people think?

-NginUS

NginUS,
   It's an interesting suggestion; we will be looking into it in detail, than you!

Please let me know if you need anything at all .

Pablo.

[PuertoLibre]

@ Pablo

It looks like the backtesting feature doesn't respect the "Trailing Stop Loss" value in the B-Bots settings. Which means that various configurations being tested will work differently in [actual] practice.



The results turn out fudged...

[fible1]

@ Pablo

It looks like the backtesting feature doesn't respect the "Trailing Stop Loss" value in the B-Bots settings. Which means that various configurations being tested will work differently in [actual] practice.



The results turn out fudged...

Hi PuertoLibre,
   TSL halts the bot when it is triggered; since there is no telling at what point you will reset the bot (because once TSL is triggered the bot must be manually reset), back testing can not account for this.

I am going to repeat something I have said often: DO NOT use TSL unless you are an expert trader with a thorough understanding of EMA and how our bot works. TSL is easy to understand in principle but highly complex when coupled with an EMA logic bot.

We originally included TSL as a feature because it was requested by a small group of expert traders for their own testing purposes, it is a very hard feature to use and I would recommend all but the most expert of traders to stay away from it to avoid losses.

I am replying to your email now .

Please do not hesitate to let me know if you need anything.

Pablo.

[bruter]

@ Pablo

Would you give us an example of "Trailing Stop Loss" feature use.

Thanks

[fible1]

@ Pablo

Would you give us an example of "Trailing Stop Loss" feature use.

Thanks

Hi Bruter ,
   The most simple way to use TSL would be with a very high value (in the high teens to mid twenties) to get out of a flash crash; again this carries a lot of risk because since the bot will shut off you could be out of luck if the price recovers before you have a chance to reset it.

TSL can be used as a stop loss feature directly with lower values if you set very high threshold values (that aren't likely to be crossed).

Again, it's a feature who's use we as a company do not advocate (as expressed in the user manual); there are models where it useful but you would have to be monitoring the bot constantly which defeats the purpose of having a bot.

I hope that answers your question, please let me know if you need anything at all .

Pablo.

[PuertoLibre]

@ Pablo

Would you give us an example of "Trailing Stop Loss" feature use.

Thanks

Hi Bruter ,
   The most simple way to use TSL would be with a very high value (in the high teens to mid twenties) to get out of a flash crash; again this carries a lot of risk because since the bot will shut off you could be out of luck if the price recovers before you have a chance to reset it.

TSL can be used as a stop loss feature directly with lower values if you set very high threshold values (that aren't likely to be crossed).

Again, it's a feature who's use we as a company do not advocate (as expressed in the user manual); there are models where it useful but you would have to be monitoring the bot constantly which defeats the purpose of having a bot.

I hope that answers your question, please let me know if you need anything at all .

Pablo.

Ah, I was automating your recent version of your bot and that is why I noticed that it didn't apply TSL. I am actually making a watchdog app that keeps an eye on the bot. (which means I can try out different TSL's on a minute/hourly/daily basis without being there by following some simple rules.)

I noticed quite a few times across a broad spectrum of settings that even good EMA settings turn out pretty bad in the back testing when there is a "flash crash".

----------------------------

These are my thoughts:

I realize this is an EMA bot. Simple rules, simple executions. As you can see in my pic above.....

IF you use 1 hour settings (and above) AND there is a flash crash within 30 minutes... (meaning, it goes from where it was to the new bottom (lets say a 25% drop in price)...the bot reacts to it after the next update by initiating a sell order. Correct?

As you can see above, it reacts very suddenly after the fact by selling at a major loss. (lets say 25% loss)

-------------------------

I assume from what you said that the TSL feature "turns off" the bots trading ability before it executes the "sell off" and thereby incurring the loss, correct?

If so I can see why the anonymous customers asked for this feature.

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

Why doesn't the bot check the 1 minute interval data and use that as the TSL trigger rather than whatever settings you choose? (lets say in this case "1 hour" ticks)

I guess what I am asking is there a way to differentiate the "TSL tick" settings apart from the "trading tick" settings?

Edit: Nevermind, it is a dumb idea!

[fible1]

@ Pablo

Would you give us an example of "Trailing Stop Loss" feature use.

Thanks

Hi Bruter ,
   The most simple way to use TSL would be with a very high value (in the high teens to mid twenties) to get out of a flash crash; again this carries a lot of risk because since the bot will shut off you could be out of luck if the price recovers before you have a chance to reset it.

TSL can be used as a stop loss feature directly with lower values if you set very high threshold values (that aren't likely to be crossed).

Again, it's a feature who's use we as a company do not advocate (as expressed in the user manual); there are models where it useful but you would have to be monitoring the bot constantly which defeats the purpose of having a bot.

I hope that answers your question, please let me know if you need anything at all .

Pablo.

Ah, I was automating your recent version of your bot and that is why I noticed that it didn't apply TSL. I am actually making a watchdog app that keeps an eye on the bot. (which means I can try out different TSL's on a minute/hourly/daily basis without being there by following some simple rules.)

I noticed quite a few times across a broad spectrum of settings that even good EMA settings turn out pretty bad in the back testing when there is a "flash crash".

----------------------------

These are my thoughts:

I realize this is an EMA bot. Simple rules, simple executions. As you can see in my pic above.....

IF you use 1 hour settings (and above) AND there is a flash crash within 30 minutes... (meaning, it goes from where it was to the new bottom (lets say a 25% drop in price)...the bot reacts to it after the next update by initiating a sell order. Correct?

As you can see above, it reacts very suddenly after the fact by selling at a major loss. (lets say 25% loss)

-------------------------

I assume from what you said that the TSL feature "turns off" the bots trading ability before it executes the "sell off" and thereby incurring the loss, correct?

If so I can see why the anonymous customers asked for this feature.

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

Why doesn't the bot check the 1 minute interval data and use that as the TSL trigger rather than whatever settings you choose? (lets say in this case "1 hour" ticks)

I guess what I am asking is there a way to differentiate the "TSL tick" settings apart from the "trading tick" settings?

Edit: Nevermind, it is a dumb idea!

Hey ,
  TSL actually liquidates all your BTC for USD and then shuts off so it's a bit different.

You can use it for panic settings but I think it's best to just let EMA protect you in case of a flash crash because it can be more reactive with the lower time frames and lower time frames are as effective as the higher one in low fee exchanges (as posted previously); I think that's the best solution .

Please let me know if you need anything .

Pablo.

[seleme]

Hi guys, haven't checked this for a while... is there any news about btc-e altcoins feature?

And maybe it is a time to check mcxnow too, their popularitiy is growing by day.

Hi Seleme ,
   We are working on hosting right now, after that is released we will most likely roll out Altcoins pretty quickly .

Please let me know if you need anything .

Pablo.

That's great man

[NginUS]

Pablo,

What's going to be new in Butterbot version 3? And when is it due to ship?

-NginUS