Topic: (Closed) Butter Bot!: Premier Bitstamp, and BTC-E EMA Trading Platform (Closed) - page 100. (Read 274784 times)

Hey Guys ,
   I think we will be ready to make feature announcements for V2.1 and I think you guys will be happily surprised with what we have come up with .

In the meantime, because I've gotten a bunch of PM's about this, here is the full trade history for 1 year back testing:

http://pastebin.com/iKv4M4wW

Notice how the first couple of months are erratic until the math has had time to settle.

Whatever anyone needs, please just let me know .

Pablo.

You Read my mind..  +1 to this idea..

.

Thanks Pablo. 

I agree very much with this feature request, we have a lot on our plate right now but I will discuss this with the dev team. This way we can all settle our thoughts on this, I think it's better to be safe than sorry. Don't expect it too soon though, we are prioritizing new features right now, especially back testing.

Pablo.

Hi Kuroth,

You're not the only one thinking this.  I've also been watching the trades on the hour and came to a similar opinion that perhaps there was manipulation to trigger the bots. It doesn't have to only be about trade fees, it could be someone wanting to unload at a price and manipulates the BTC price up by a few cents to get their coins sold.  But I dismissed my theory as just paranoia as well.

Perhaps in the future versions of the Bot, we can set our own times. For example, instead of 1 hour time-frames, we can do 1 hour and 1 minute, or 23 hours instead of 24 hours.  This way, there won't be 50,000 people with Bots that can be triggered on set specific times.  The ability to customize time-frames would provide greater diversity and make it difficult for manipulators to time things. Or maybe I'm just too paranoid and suspicious of everything.

Related to this point: It is well known that there are a group of very wealthy early Bitcoin adopters who play with Mt. Gox by moving their large reserves. It's actually something a lot of people try; I heard from a marketer not too long ago who was launching a bogus site and he claimed he would be moving the Bitcoin market somewhere between the 12-16th August (don't remember the exact date). Mt. Gox doesn't need to do it, enough people with a lot of time on their hands are trying it already. The good news is that the market is maturing and less gullible than it used to be so these dumps and schemes are generating increasingly lower swings.

Pablo.

I tend to be a paranoid person..   

I think a lot of it was brought on from back in my foolish days when I use to play a lot of on-line poker..   I swear there are 1,000's ways those poker sites could cheat their customers if they wanted to and I believe to this day(Although I never had proof) that some of them did cheat..


So I guess when I look at Mt Gox and I noticed very strange buy bids that seem to get placed right around the top of the hour that look like this

0.1   0.1126   103.89040


Who in the world would place a Buy bid to buy just 0.1126 of  BTC?  Makes no sense..   Makes my wild mind wonder if someone working for Mt Gox is placing these at the top of the hour to try and trigger Trade bots to buy or sell thus generating a Trade Fee...   

Ok I know I am paranoid..  Just makes you wonder sometimes how we could be played.. 

Hi PuertoLibre ,
  Restarts affect the bot in the sense that they represent downtime. The bot will get all it's data points from the server every time you turn it on, so that's not an issue.

There is no easy way to determine the impact of downtime, I can tell you that it puts a serious dent in bot profitability. I understand we can't all run servers, but we should aim for as much uptime as possible. The more downtime the more backtesting goes out the window. Just FYI, I've been running the bot nonstop on my macbook air since we launched the software so it is doable (I use a free app to stop the laptop from going to sleep, that's all you need). And just to repeat myself, if downtime is bad, intervention explodes a hydrogen bomb on back testing.

Please let me know if I can be of any further service .

Pablo.

Ah, you stole my idea! Hehe, anyway, considering how the bot works it sounds like it needs a 24x7 environment.

Any idea how badly restarts will affect B-Bot? Is there a way to measure the damage from such an interruption?

(The reason I ask is because B-Bot is a plugin inside a browser....so 24x7 uptime even in chrome is hard to achieve. I restart my chrome once a month on average.)

Is there a way to determine just how bad the result is with "reasonable" levels of interruption?

Vs

A server environment (that short of a tornado or an Admin hanging by the cables...really wouldn't interrupt the process.

I'm sorry I posted my answers all in a row today, and I'm sorry I'm spamming the thread again but I just want to tell you guys how happy I personally am with the discussions we have been having. Amongst ourselves we disagree on certain points but it's so encouraging to me to see you guys arguing your points with each other. I hope we can continue to have these discussions on a regular basis and to grow them into a collaborative effort .

Pablo.

We are looking at slippage, the solution to this, as Kuroth points out, is mathematically complex; this is something we will likely work on after the next release.

Pablo.

I think high frequency trading would be an interesting new strategy module when we get that off the ground. Part of the reason we will introduce strategy modules is to help people who want to make shorter term profit. We're open to suggestions as to what specific stratgies you guys would like to see .

Pablo.

Hey ,
   The back testing assumes that the bot has run both without interruption and without intervention. I think (and this is just me thinking out loud, no plans on this yet) that it would make sense to have the bot eventually run on our servers and have people mod settings and use tools from the extension.

Just to repeat, drops in connection are not good because the bot misses opportunities, intervention in trading is critically bad to how the bot will perform over the long term; it basically annuls the back testing.



Let me know if you need anything .

Pablo.

Well its right here  https://bitfunder.com/

And it works like a stock market..   

you dont really need a bot for that, just check the charts once a day

But here again all these Wild %'s of profits was due to the Price of BTC being at the $15 per BTC  at the first of the Year..  Not saying we should throw that out of backtesting but I am not sure we will ever see upswing like that again anytime soon..   Just about ANY method during that time frame would have made a Large Profit..   


I am trying to be more realistic with my Backtesting..

Yes Goomboo went to 1d as well..   This has been requested and Pablo says they are looking into it.. If not 1d would at least like to see 4h and 8h and 12h options..

https://cryptotrader.org/backtests/wGfXaAKif8RFme5Bg

According to the ema strategy of 0.25 and 1 day changes, the profit would be close to 400% since Jan 1, 2013. (10/21 ema)

2013-01-02 18:00 Simulation started. Balance: 1000 USD
2013-01-02 18:00 BUY 74.887 BTC at 13.28
2013-04-16 19:00 SELL 74.887 BTC at 68.36
2013-04-20 19:00 BUY 39.986 BTC at 126.62
2013-05-03 19:00 SELL 39.986 BTC at 98.10
2013-05-19 19:00 BUY 31.670 BTC at 122.50
2013-06-06 19:00 SELL 31.670 BTC at 118.97
2013-07-17 19:00 BUY 37.832 BTC at 98.50
2013-08-02 19:00 Simulation completed. Balance: 37.832 BTC (3953.46 USD)

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

Same settings, only with an EMA of 14/21 and you get 800%

https://cryptotrader.org/backtests/EhQvz6SatHr4HPdeN

2013-01-02 18:00 Simulation started. Balance: 1000 USD
2013-01-02 18:00 BUY 74.887 BTC at 13.28
2013-05-06 19:00 SELL 74.887 BTC at 112.25
2013-05-21 19:00 BUY 67.653 BTC at 122.89
2013-06-07 19:00 SELL 67.653 BTC at 111.00
2013-07-26 19:00 BUY 77.349 BTC at 96.02
2013-08-02 19:00 Simulation completed. Balance: 77.349 BTC (8083.08 USD)

Someone previously asked me why. I think I forgot to respond.

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

https://cryptotrader.org/backtests/QviZdcwv6DNhCjjAL

2013-01-02 18:00 Simulation started. Balance: 1000 USD
2013-01-02 18:00 BUY 74.887 BTC at 13.28
2013-05-05 19:00 SELL 74.887 BTC at 115.98
2013-05-19 19:00 BUY 70.124 BTC at 122.50
2013-06-07 19:00 SELL 70.124 BTC at 111.00
2013-07-25 19:00 BUY 79.405 BTC at 96.95
2013-08-02 19:00 Simulation completed. Balance: 79.405 BTC (8297.91 USD)

14/21 ema

triggers at -0.01 in both buy/sell

830% results.

Now someone else pointed out that the Bot dont take into account Liquidity and resulting slippage..  Not sure the Bot could ever do that as it seems it would be pretty complex to code for..   

But keep in mind at least according to what Goomboo stated, he was NOT using a Bot..  But was dong this EMA method manually..  Thus he could take into account more things like Liquidity and resulting slippage and the Trade Fees etc..