Quote from: PuertoLibre on July 19, 2013, 09:54:50 AM

Quote from: fible1 on July 16, 2013, 02:22:23 PM

Quote from: PuertoLibre on July 18, 2013, 01:09:29 AM

I have been learning how the bot works by watching it make decisions. I think there is room for improvement with just a few extra features.

Is there any chance that the tables that show up showing the Time, Price, Diff could be outputted in some other form? CSV or just as a regular HTML web page that gets re-generated with ongoing measurements of diff?

The reason I ask is because the bot gives valuable information that can make it extremely trivial to make a profit at any setting. The issue is the chrome plugin seems to generate a valid web page with "all the details" but the page itself is not directly accessible.

I am a fan of automation and would very much love to get my hands on the data spit out by the B-Bot in "popup.html". I have been trading manually alongside the B-Bot in a different account and that has netted me about 10 to 15 BTC in the last two days. But the B-Bot has actually lost .3 BTC in a test account (originally with 4.5 BTC) in the same time frame.

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

The reason I ask is because I noticed that the B-Bot pays close attention to the differences between the two EMA lines to determine the crossover. Both going up and down. The shape of the curve and the steepness can be determined from simply looking at the difference values. What I haven't seen yet is any form of the bot tracking what the original entry point was. (Maybe it does, but it doesn't give any indication of what that value might be)

I assumed the "Stop Trading Loss" feature was implemented as a way to trigger some kind of percentage where the bot would execute a trade. The only problem is it does not appear to work. (I am unsure)

I place very low settings on the bot to see if it would trigger trades even with a drop of 1%+ (my stop loss setting) and it seems to follow the curve even when it drops more than 1% (in my favor).

So there seems to be a very critical feature missing from B-Bot v2 beta. It follows EMA fine but it loses plenty of trading opportunities. (as evidence on my other account where I gained nearly 15 BTC by follow the EMA lines manually)

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

If I could pull out that table data as a popup.html that keeps tabs on the diff data it would be a godsend. Is there any way to pull it open in any (full featured) browser?

Hi PuertoLibre

,
Let me take your questions one at a time:

1. Whatever results you have seen since you got your bot a few days ago, be they profit or loss with B-Bot are likely the product of market randomness. This is the discussion we were having with Kuroth; profitability on an EMA bot, generally any kind of bot, can only be judged in longer time frames as to account for said randomness (think: "The Law of Large Numbers") and let trends emerge and the math do it's job.

I am happy you have seen profit by trading crossover, but again, unless you have been doing this consistently for some time, profits there are also likely the product of market randomness, the true measure of an algorithm is it's profitability over the long term.

I am not saying your strategy is bad, trading crossovers only is an established strategy, but rather that your results are not reflective of the strategy itself. This is part of the reason why professional back testing models generally go back several weeks if not months or years.

I am giving your bot plenty of time to see how it does things. I have been reading through Gomboo's discussion and there are very good insights in there. (So far I am only on page 6)

Today the bot lost another .2 BTC in an originally 4.5 BTC account. The current remaining values stand at 4.070 BTC remaining.

In a larger account I do manual trading with I gained 1.7 BTC by doing manual trading today. This was only because I was busy with hardware upgrades and was only able to do 1 trade today.

-------------------------
Side note:

I have completed automating part of (65%) of my strategy as I carefully have analyzed what I actually do and have tried to make all the information come together so all the valuable information is presented to me in one place. The last component on my program is using information made available by B-Bot to assist me in making final decisions. (None of my automation routines can actually see the EMA curves directly, so that is why I am looking at B-Bot to fill in the missing piece of information.)

The goal is to give my eyes a bit of rest since a machine can at first collect information in one place and make intelligent decisions without paying attention to dozens of windows. Ultimately to allow a computer to follow the same rules I use my brain for to do the trading for me.

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

What I find disconcerting is that B-Bot has so far missed so many opportunities (both at 15min scales as well as 1 min scales) that it worries me.

Quote from: fible1 on July 16, 2013, 02:22:23 PM

Our bot is based on Gomboo's discussion of EMA as applied to Bitcoin and all these issues are discussed there, recommended read

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

2. The bot calculates 144 data points to determine EMA values. This is 144 points in whatever time frame you select, if your time frame is 1 hour it will go back 144 hours and the first hour will be your entry point.

This part also worries me.

Someone asked me an apt question when I read back this portion of your reply. I was asked, how does the bot know that 144 points in the past was a relevant (or even a good) starting point for the future [decisions]?

I couldn't help but shrug as they asked me that because it is a very good question. If the Bot doesn't use the highest recent point as the entry point (but ? apparently) just the 144th entry point from "now", it can apparently make....pretty bad decisions.

The point that came to mind is asking if there is any way to add a feature where:

--A) We can see what that entry point actually is at any given time.

and

--B) Change the length of that entry point to what we want. (72 previous points instead?)

Quote from: fible1 on July 16, 2013, 02:22:23 PM

3. Trailing Stop Loss takes the highest value Bitcoin has achieved since you turned on the bot and will SELL all your BTC if the value goes down from the maximum by the percentage you have specified. So it really was following it's programming when the value changed in your favor, by design, it will only intervene when there is a loss. This is one of those features that is very sensitive to set and that you have to keep an eye on though.

First question, does this field accept a negative value?

Second question, Does this rely on the mystery 144th point in the past? (or when I actually turned on the bot?)

Perhaps a bot with internally tweakable settings might allow even more customization.

Hi Puerto Libre

,
I think "WhyDifficult" did a good job of explaining the logic behind the 144 data points so I will leave it at that.

I think there is a misunderstanding about how EMA logic works, allow me to explain:

EMA is trend logic. The bots function is to get you out of a crash as early as possible and into a rise as early as possible. It is not a day trading bot and the math behind it does not suggest you will make a profit out of every single trade. The math suggests that if you run the bot for a decent space of time the profit trades will amount to more money than the loss trades and thus you will profit. I explained in a previous post that this has to do with the leveling of market randomness well expressed by the logic of the "Law of Large Numbers." If you advance on Gomboo's thread you will see that although loss trades will likely outnumber profit trades, profit trades should account for more money over time.

I mentioned before but I will say it again, 15 minute time scales are not the best, 1 minute is certainly no better. As a trend bot, the longer the space of time you trade at the more likely trends identified will be real. If you are trading at 1 minute intervals, you are really using bot as a high frequency trader which it is not intended to be. Those shorter time frames exist because people requested them for specialized testing of models; Gomboo always suggested trading at one hour and now he has brought it up to one day. One of the key reasons you are seeing losses, I suspect, is that the fees associated with one minute trading are eating into any money you could be making. I encourage you to use nothing shorter than 30 minutes unless you have a thorough understanding of EMA and a specific goal in mind with a good basis behind it, not just to see more trades.

A lot of the discussion we are having here comes from a misunderstanding of EMA logic and confusion of trend logic from day trading logic.

I am here to help with whatever you need so just drop me an email and I will be happy to clarify any points you need help with. I suggest you consider the points we have been discussing and reevaluate your understanding of Trend/EMA logic from day trading/high frequency logic. I would really encourage you to bring your trading window up to at least 30 minutes or your fees will continue to eat into your profits.

Please let me know if I may be of service any other way.

Best,
Pablo.