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

[seanrarey]

Your mileage will vary depending on your tick-offset, but here is the 1hr v 30 minutes for the last 3 months:


There is a significant penalty for anything under 60 minutes in all the simulations I run.

When I go back manually to see why, it comes full circle to that all important momentum timing.

Sean

[seanrarey]

here's 145%:

Ah... you are using double gates... so in affect, we are doing the same thing, just different ways.  With double gates, your poll time is 30 min, but trades are at 1hr (has to pass through 2 gates @ 30min each).  I am running a single gate @ 1hr.

End result, same.

[velacreations]

looking at my tests, if I go to 1 hour, I actually have a higher percentage profit, because the bot bought earlier (at $175), but also sold earlier ($308)

at 30m, he bought at $195, sold at $318.

so, what you want is the 1 hour buying for you and the 30 minute selling for you.

[velacreations]

Ah... you are using double gates... so in affect, we are doing the same thing, just different ways.  With double gates, your poll time is 30 min, but trades are at 1hr (has to pass through 2 gates @ 30min each).  I am running a single gate @ 1hr.

End result, same.

well, yes and no.  the end result is that your settings sold at $288, whereas mine sold at $318. that's almost 10% difference.

[seanrarey]

Ah... you are using double gates... so in affect, we are doing the same thing, just different ways.  With double gates, your poll time is 30 min, but trades are at 1hr (has to pass through 2 gates @ 30min each).  I am running a single gate @ 1hr.

End result, same.

well, yes and no.  the end result is that your settings sold at $288, whereas mine sold at $318. that's almost 10% difference.

Oh, sir, be so very careful not to get wrapped up in that number.  Our settings could be identical, with tick-offset 5 minutes apart and have vastly different buy/sell numbers.

It is a simple matter to go back and "pick" the moments it was optimum to buy and sell.  It is not possible to predict the next optimum moment.

Thus, it is important to establish a good solid strategy, and stick with it, regardless of the luck factor of exactly when the order falls.

Refer to my suggestions on experimenting with tick-offset.

Ours is a perfect example.  We are both trading the same EMA, both trading at 1hr interval.  The gate setting is different (mine is faster), but the bulk of the difference in trade point was the tick-offset; i.e. luck.

[seanrarey]

Ah... you are using double gates... so in affect, we are doing the same thing, just different ways.  With double gates, your poll time is 30 min, but trades are at 1hr (has to pass through 2 gates @ 30min each).  I am running a single gate @ 1hr.

End result, same.

well, yes and no.  the end result is that your settings sold at $288, whereas mine sold at $318. that's almost 10% difference.

Further examination of the charts uncovers another reason you had a better sell point then I did.  You sold slower... later... with your slower gate setting (0.09 v 0.01).

Nice work.

[uck]

I know weve been talking about new features we'd like to see, and i for one love the backtest engine. Howabout a forwardtest engine?

Seriously, It would be nice if the backtest engine had a little more info like number of losing vs winning trades, so we could also try to see if we could find settings that at least in theory minimized the number of trades where we just turned around and bought it back slightly higher.

[seanrarey]

I know weve been talking about new features we'd like to see, and i for one love the backtest engine. Howabout a forwardtest engine?

Seriously, It would be nice if the backtest engine had a little more info like number of losing vs winning trades, so we could also try to see if we could find settings that at least in theory minimized the number of trades where we just turned around and bought it back slightly higher.

But, it does.  You can scroll down and see every trade, win & lose, & by how much.

[uck]

I meant if it automatically summarized it at the top like it does the min profit...

[seanrarey]

I meant if it automatically summarized it at the top like it does the min profit...

I see.  I guess my question is, why would you care about that statistic?  Who cares if it makes 99 losing trades & 1 wining trade... if you get the best return?

To expand... 60% wining trades that gets you 10% return on your money is not better then 20% winning trades that gets you 40% back on your money.

The testing gives you the important data... return over time.  I am trying to figure out of what use the percentage of wining/losing trades would be in forming your strategy?

[ghostshirt]

I like the 1hr poll times, but to slow the bot back down I use the 20/42 EMA.  EMA is not about "agile", it is about "timing".

The 20/42 EMA results in a greater spread and makes it slower to cross.  This eliminates a lot of trade noise.  I then use very fast gate timing (-0.2/0.01) so that when it does cross, it trades.  The crosses are much sharper with 20/42 settings as opposed to the more traditional 10/21.

I like 20/42, but with 30m.  It seems to perform better, especially with big/quick drops, like this weekend.  Your 1 hr setting sold at $288, whereas the 30m sold at $320.

I think going back to 1hr might be better for times with less volatility.

Thank you both for sharing your settings.
I tried 20/42 with both 30 min and 1 hour settings and 1 hour yields almost two times profit in 3 months and 6 months backtesting.

[fible1]

Hey Guys ,
   I've been following the conversation, couple of things:

1. For those of you having issues with Mt. Gox, this is a limited bug that triggers under specific circumstances; we will release an update to take care of this plus some other minor work ups tonight.

2: The graph issue is indeed an issue with BitcoinCharts.com: BitcoinCharts doesn't have the best API and it sometimes crashes in moments of high volatility, we will see about how to mitigate this.

This post below worries me a lot:

Does anyone have any great settings that work on 1 minute frequency? (and is willing to share?)

=======[Discount Code for Butter (10% off)]=========

https://butter-bot.com/?r=f5a17f5b

Use my referal code cause' im sooo cute!  



You'll get an extra month free, it's all on me!

OMG... wrong type of system for 1 minute settings.  Do the tests... you will run out of money fast.  

Seriously never recommend anything under 60 minutes with EMA.

1 minute settings, backtested for 1 month.

Profit:77.47%Max. Profit:77.47%Min. Profit:-0.60%Trades:32

I have backtested it further up to 6 months but the backtesting server seems to be sporadic at the moment.

There was another setting that got up to 127% with trades going back much further. I can't figure out what the settings were exactly as I kept trying new ones.

Please, please, please avoid trading at intervals below 30 minutes. Even though 1 may back test well for the past month, like someone already said, BTC has been doing nothing but climbing for the past month. EMA is a trend strategy and trends are more easily identified over longer stretches of time, at 1 minute, you are really just trading on noise and more likely than not, you will loose your shirt.

I know it is fun to watch the bot trade more often like a High Frequency Trader (HFT) would, but the reason we don't have an HFT yet is because so far we haven't found one that is productive, please do not use Butter as an HFT.

I am here for whatever you guys may need .

Pablo.

[velacreations]

Oh, sir, be so very careful not to get wrapped up in that number.  Our settings could be identical, with tick-offset 5 minutes apart and have vastly different buy/sell numbers.

yes, I agree, though going back through 1 hour vs 30m, 1 hour tends to buy lower, and 30m tends to sell higher, all other things being equal, going back 6 months or more.

You are right about the tick off-set.

Further examination of the charts uncovers another reason you had a better sell point then I did.  You sold slower... later... with your slower gate setting (0.09 v 0.01).

yes, it may be a good thing to avoid bull traps.

BUT, it's not just you, it's also my settings with 1 hr.  At 1hr, it gets in lower, but sells lower, too.

How do we combine them into a great strategy?

[uck]

what if the Bot dynamically slowed down and sped up depending on conditions it was sensing? maybe volume? maybe some other volatility indicator?

[velacreations]

Thank you both for sharing your settings.
I tried 20/42 with both 30 min and 1 hour settings and 1 hour yields almost two times profit in 3 months and 6 months backtesting.

what are your other settings?  30m outperforms 1hr for me.

Here's 1 hr:


Here's 30m:


there's 50% difference.

[velacreations]

what if the Bot dynamically slowed down and sped up depending on conditions it was sensing? maybe volume? maybe some other volatility indicator?

what would be really awesome is if we could store settings profiles, so then you could switch strategies, depending on what the market is doing.  Like right now, the bot is looking for a place to buy in, so I want it set with the 1 hour settings and maybe some other tweaks.  But, last night, the 30m settings did better.

So, I have 2 slightly different strategies, each is good at only 1/2 of the equation (one is better for buying, other better for selling).  Combining them together, I would see a significant bump in performance.

[fible1]

***************
Update
***************

Friends,
  We just released an update labeled V2.4 to deal with some of the issues arising from yesterday's release.

Improvements include:

* Graphs now show override settings

* Fixes for trading issues on Mt. Gox

* Bot now starts as disabled when first installed: Note that the bot will be running and NOT disabled if you install it on top of a previous version that is currently running.

* You can turn trading at exchanges ON and OFF independently


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

Pablo.

[PuertoLibre]

Does anyone have any great settings that work on 1 minute frequency? (and is willing to share?)

=======[Discount Code for Butter (10% off)]=========

https://butter-bot.com/?r=f5a17f5b

Use my referal code cause' im sooo cute!  



You'll get an extra month free, it's all on me!

OMG... wrong type of system for 1 minute settings.  Do the tests... you will run out of money fast.  

Seriously never recommend anything under 60 minutes with EMA.

1 minute settings, backtested for 1 month.

Profit:77.47%Max. Profit:77.47%Min. Profit:-0.60%Trades:32

I have backtested it further up to 6 months but the backtesting server seems to be sporadic at the moment.

There was another setting that got up to 127% with trades going back much further. I can't figure out what the settings were exactly as I kept trying new ones.

Careful with the 30-day backtest at this point, as the market has done very little but go up.  This is going to make it a bit deceptive with some settings.  As the market movement normalizes, shorter time frames can eat you alive with fees and short-trades.

I made 137% during the same period using 1hr settings (hopefully the picture posts)...

Try the below settings and see if you guys/gals can refine it as much as you did with the other longer period settings.



With v2.4 the bot should no longer be broken on these settings.

(Note: This uses reverse EMA settings. This is targeting spikes and drops of about $15 to $20 at 1 minute interval rather than long term trends on a larger interval.)
(Note1: Make sure to heed Pablo's warning above when backtesting and refining the settings. See if you can make it better.)

[Basaj]

Hey Guys,
Just installed and setup my bot,added the API keys etc, and have hit a bit of a roadblock. I can see the icon in chrome and the numbers on the bottom are fluctuating so it is "live" but when I click the icon I have a warning triangle. It says that the API is connected but none of the trading platforms are. I have only set it up with BTC-E so far but beside btc-e it says disconnected. Am I doing something wrong or is it simply because the bot has not attempted a trade yet?

Also, if you dont mind, does anyone want to post some decent settings for the bot engine on btc-e? I have been messing around with them but would be curious to know what other people use.

Thanks a million in advance!

[PuertoLibre]

Feature request: Add the exchange btcchina.com