Topic: (Closed) Butter Bot!: Premier Bitstamp, and BTC-E EMA Trading Platform (Closed) - page 25. (Read 274781 times)

Hey Pablo...

Any idea on when we will see custom date ranges for the backtest engine?

Cryptotrader now wants to charge $8/mo for backtesting capabilities ONLY...

As much as I love butter-bot the backtesting is worthless without custom date ranges.

Try Short: 20 Long: 42
Freq: 30 Min
Buy 1: -50
Sell 1: -50
Buy 2: 0.10
Sell 2: 0.15

And go from there...

Sup guys, I'm a newbie here. Could anyone please tell me some config's for BTC-E (EMA) at the moment?

Thanks guys, that clears it up.

I think that is quite confusing.
If i am backtesting at 9:30 and it gives a X gain result, and then i backtest again at 9:31 i get another one.

I think the backtest should always be based on the 0:00 + the time period (hourly/2h/6h/etc) + the tick offset
not on timenow() + the time period (hourly/2h/6h/etc) + the tick offset



Let's imagine i try to create a comparative table of backtests to choose the best one i prefer

If I am backtesting on Hourly with 0 tick offset, i think is confusing to consider that the backtest is the timenow() + 0 offset because the timenow() is changing all the time so i cannot rely on the data of my own table after 5 minutes have passed.

I think it could be more clear always basing the backtests on the 0:00 hours (like 1:00, 2:00, 3:00, etc for hourly) and i you want to have your trades to execute at 0:45 you can configure the offset to +45 minutes. Like that you will be able to create a comparative table and get some conclusion.

Does it make any sense?

Cheers,

What are the recommended settings?

(I forgot what it was)

Well, I'm not exactly sure, but I'd speculate that the thing that creates the difference is the price of BTC/USD as the live price is always fluctuating, so may have changed in that 30-60 mins that you were tweaking the settings. If say the price was $830 and when you first backtested, it's probably not going to be the exact same price when you backtested it slightly later, it might be $825 or $840 or whatever and I'd suspect that this difference would be responsible for the difference in the backtesting results.

Just a thought.

Firstly, I love this and think its great.

However...

I backtested the recommended settings for BTC-e for 1 week....all fine.  Tidy little profit.

I then fucked around with the settings for next half an hour to an hour, and upon returning to the original settings the backtest results were different.

Why is this?  My numbers are exactly the same.

Did anyone compare the performance of the different bots that are available? Especially because there seem to be some free ones... I recognized the following ones:

CryptoTrader.org https://bitcointalk.org/index.php?topic=222696.0;all
Haasonline Simple Trade Bot https://bitcointalksearch.org/topic/haasonline-simple-trade-bot-for-btce-and-bitstampmain-topic-211979
HAL10K https://bitcointalk.org/index.php?topic=391630.0;all
Easybot (open source)  https://bitcointalk.org/index.php?topic=354871
Gekko - a javascript trading bot for nodejs https://bitcointalk.org/index.php?topic=209149.0;all

Butter Bot says they had 3464% up since two years  But what is the performance of these Bots if you compare it to a simple hold strategy?

Or are all bots roughly the same and it all depends on your fine tuning? If that is true, then which has the best default setting, best usability?

Do all use the same EMA strategy?

If you're on BTC-e at least, and of course depending on your settings, you may get a 'sell order is imminent' email soon by the looks of it.

Watch I jinxed it by mentioning it, lol.

Thanks a lot for your answer. It does make sense and I can see the logic now. I thought it was increasing BTC. Seems good to me. I guess trial will become long term.

Good idea! So that others can benefit from this. I'm in.

We should also start a contest,

To see who can fine tune butter the best from week to week, month to month. (Back Testing and Forward Testing)

We can start off from the default settings until we max out in terms of optimized opportunity.

Might be fun?

With all the talk about trial software.....

Pablo should probably turn the trial software into a simulation.

Unlimited time, but your account won't actually trade in real life. As long as you keep it open it will trade imaginary numbers but nothing actually goes out to the exchange API's. Just simple math.

If a person tried the simulation then after some unknown period of time, they would be able to fine tune the settings and see the ongoing results of their changes. If it convinces them, they buy the software license, if not, no harm no foul.

Pablo should also implement profiles (I believe it's on the board already) so that users can openly exchange settings either out in the open or between each other. Giving newbies to the ButterBot simulator(formerly trial software) some "settings" which they can see results within a reasonable period of time.

It will also help everyone figure out what kinds of changes makes sense and why it works better than another setting.

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

One "problem" with the backtesting feature is that it doesn't really give you a clue as to why one setting works spectacularly while other settings work downright terrible.

People turn to BitCoinWisdom because it gives them a visual representation of the last 900 update cycles. It lets people do the tweaking and only in one dimension.

It would be nice if there was eventually a fully supported website by the butter bot team that shows users (interactively) what different settings do when they are changed. Letting the casual simulator or licensed user of Butter bot focus in on what could be improved on.

Allowing you to:
Change the tick settings with vertical highlights,
Change EMA's all four EMA pairs with highlights that spots where crossovers occured in the past based on your chosen settings,
Change the EMA's pairs and see how many spikes and drops were taken advantage of in a given period of time.
Change the periods of update to see what is more advantageous.


In other words interactive, visual and dynamic so people can figure out what the numbers actually mean and what changes with each setting.

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

Pablo's team should also (geeze one too many ideas) focus in on a little addon to find the heart beats of the market. Both the large ones and the smaller ones.

Not incredibly, if you look over the graphs (RTBTC/BitCoinWisdom etc) you can find the timing of various bots that do large trades at pretty precise intervals. Allowing you to time your own buys and sells (or butter bot) with that in mind.

no

Hello,
 
  I have kinda been looking into trying this, but I was wondering.... I see that you can set reserves for held BTC and USD what I would like to know is if that can be set as a percentage? IE can I do 30% instead of $500?

If your funds are held in BTC on the exchanges and the price of BTC/USD is rising, the bot doesn't trade as it measures the value of your account holdings in USD and, with the BTC/USD price rising, you're deemed to be making money. If the price of BTC/USD begins to fall and according to your settings the bot determines that the trend price of BTC is falling, it will sell the BTC back into USD as, again, it determines the value of your account holdings from a USD perspective.

As yesterday the price of BTC/USD was largely upward, it's quite normal that no trades would have been made with your bot because the funds held in your account at commencement were in BTC and valuing your account from a USD perspective, the value of your BTC was increasing. Had your starting funds been USD, it's likely that the bot would have bought BTC sensing the upward trend in the price of BTC.

Hope this helps.

Hi,

Is it normal nothing happens at all after 1 day? The bot seems to be setup properly as it can see my 0.2 BTC on BTC-e properly but then nothing. I used the recommended settings for BTC-e. It seems like a good thing after reading the thread but am I missing something. I use the trial version as I want to see what happens.

The first time I tried to run backtesting on the new version, it failed, outputting just the. Top row of the table where the data goes. So I checked all over looking for something that could be causing the issue, but found nothing wrong. So I tried it again & it's been working ever since.