MTGox Automatic Trading Bot Tracking To Gauge Usefulness As A Security

[MtGoxTraderBot]

Taking a page from the volatility bot (beat me to the name darn you), I have been testing this for a couple months on my own java bot for mtgox. But wanted to document the findings so far, I began with about 45 btc, only made about 50btc so far (5btc profit). But based on my findings so far I would estimate I could turn around a decent profit over time. I have not been able to figure an algorithm that would be able to work in the EMA and calculating spreads to make a profit so I have come to the conclusion that it requires a person watching at all times just in case a sudden dip or rally in pricing comes around.  These bots do perform better under volatility as the other thread pointed out. I have some slightly different views on how the trades should be performed though and until I get my nominal trade fee down another tier I dont think I will make big profits on it. But so far got it down to the .46%, I am looking into setting up a security on glbse also in imitation. Now that I have my trade fee down some though I will keep toying with it for another week in hopes that I wont need to open up a security. I have visited the forums for information many many times and gotten lots of great information, so hoping this will help other bot builders in their future endeavors. I will not be using multiple markets, my goal is not to arbitrage in any way, I only plan to use the market to move the btc back and forth without any sort of emotion involved since I wont freak out when the pricing goes up or down and FUD panic sell/buy in desperation.

And to quote the other thread:

Bot is very secure, never makes a loss, makes money on volatility. The bigger the market volatility, the bigger the returns. But according to my calculations ROI is pretty low. On a high volatile market (like the one this week and last week), it would yield up to 0.6-0.8% per week (I already substracted 10% management fees from that), but usually less than 0.5%. Profits should improve if there's bigger trade volume, currently bot pays 0.6% fee per trade as the trade volume is less than 100 BTC. For trade volume > 100 BTC, MtGox charges only 0.55% etc.
Bot is working on EUR/BTC market and has orders in the range between 3 EUR/BTC to 10 EUR/BTC (current price is around 7 EUR/BTC). If the price is swinging in this range bot is making money. If the price will go upwards of 10 EUR/BTC that shouldn't be a problem because all the profit is always on the BTC side (bot is always buying first if there's available EUR balance) and each order is valued at the EUR price, so I can move the range upwards without losing money. If price will go downwards of 3 EUR/BTC there will be a decision to make if to move the range downwards and lose some profit or close down the operation and return the money.
MtGox account will of course be protected by 2-factor authentication and withdrawals through API will be blocked.
I will setup this as a company with voting rights. I plan to issue 1000 2000 shares at 0.1 BTC, total 100 200 BTC, for starters.

Everything I have read up on his thread is correct ROI is low, but I think it can be increased by drawing lower profits on the trades until the fee is dropped down. This will be my goal, to break even on each trade until I have my fee down to less than 0.4%.



Bots funds: Going to start with 50Btc to have a nice round number to begin with.

Bot MtGox USD                                                     
Approximate value of funds: 50btc                                                      
Current Trade Fee:  0.46%

Bots performance:

Code:

Week   Start 	     End 	Profit      Fee (10%)	      Dividend per share	         ROI/week (So far)
------------------------------------------------------------------------------------------------------------------------------------
1     9/16/12      9/22/12     4.3      N/A(glbse isnt open yet so fee/share wont be here)        ~8.6% 
1     9/23/12      9/29/12     2.7      N/A(glbse isnt open yet so fee/share wont be here)        ~5.4% 
 

This week has been going not quite so good, the bot got stuck on the wrong side of the trend and had to wait for the market to rise back up to continue trades.

[MtGoxTraderBot]

If  you happen to have developed a bot in the past and can think of algorithm that would possibly assist with filtering mtgox's API. This seems to be the usual issue as I cannot seem to get the API to recognize individual identifiers on each trade so I am having to resort to using the trade amount, which is difficult since mtgox rounds off sometimes. Any ideas on that?

[GeoRW]

watching 

[markm]

It seems to me that such bots might work better on markets that do not have a percentage based fee?

Or does it actually prefer trades so small that a fixed fee per trade (say maybe about half a cent to a nickel in USD or CAD terms) would be larger than the typical percentage-based fees are?

I have not yet actually determined exactly what the fees will be when I do turn on the usage-tokens system on the Digitalis Open Transactions server because until I do turn it on I will not know exactly how many API calls, and thus how many usage tokens, typical scripts will tend to burn through in the course of normal operation. Also it is known that the "test GUI" java client makes many more API calls than it actually technically needs to (its programmer wasn't having to pay for each API call so wasn't really thinking about minimising the number of calls it makes).

So far wild ass guesses though have been that we would probably sell somewhere in the range of 100 to 500 "usage tokens" for a bitNicKeL (NKL), and as you can see by consulting http://galaxies.mygamesonline.org/digitalisassets.html bitNicKeLs have usually been about $0.50 to $0.60 or so each (though still climbing, it seems...)

-MarkM-

[MtGoxTraderBot]

One thing is for sure, your better off losing a bunch of btc testing the API on your own, mtgox never understands the questions I ask about the API. I get back things like, we dont understand. Huh? Whats an API? Why do I work in mtgox r&d? But I think I have it figured out now. Basically the bot needs to only do 1 order at a time, if there are any open orders showing then it wont continue trading. That is why the EMA wont really work (for now) sometimes I have to change the settings for it to take a small loss to get back in the mix. But the new bots on the market make it difficult since they are trading back and forth against each other .01 at a time so it looks like the trades are coming and going at the speed of light. I am attempting to get an algo going that will take this into account after x amount of time and take a loss of up to x amount and buy in or sell to get to that range instead.

[MtGoxTraderBot]

The fee is not to tough to manage, just have a set amount you want to trade at depending on the fee. But the swings in the market with the new bots out there are making it tough since they are doing .01 at a time the bot is reading a never ending line of trades.

[MtGoxTraderBot]

I will make a version to check it against depth and test it out to see how it goes.

[markm]

The fee is not to tough to manage, just have a set amount you want to trade at depending on the fee. But the swings in the market with the new bots out there are making it tough since they are doing .01 at a time the bot is reading a never ending line of trades.

Maybe that too then is another way in which trading on an Open Transactions server might be easier, though this depends on how the specific asset contracts happen to be set up at the server.

On my server, for example, all assets are integers, for many reasons. This means that all prices must be integers, which in turn means that the number of effective digits of granularity you have in prices is determined by the scale of the market you choose to trade on.

A server offering a USD asset would probably use cents, but since this is just a for example hypothetical let us assume a server that has integer dollars and integer bitcoins.

On the scale one market you could offer 9 dollars per bitcoin, ten dollars per bitcoin, eleven dollars per bitcoin etc.

On the scale ten market you could offer 101 dollars per ten bitcoins, 102 dollars per ten bitcoins, etc.

On the scale one hundred market you could offer 1001 dollars per 100 bitcoins, 1002 dollars per 100 bitcoins etc.

On the scale one thousand market you could offer 10001 dollars per 1000 bitcoins, 10002 dollars per 1000 bitcoins etc.

See how the number of effective decimals of price you get is increased by using a larger scale market?

It is my hope that this will help lead to reasonable trades instead of a proliferation of dust-spam tiny little trades...

-MarkM-

[jasinlee]

That would be nice, my eyes hurt after watching 10 minutes of the spam flying around.

[MtGoxTraderBot]

1 Week complete of trading, good profits overall of 4.3 BTC, I am going to remove the 4.3 from the account and start this week back up with 50 again to keep my math going easier for the coming weeks tracking.

[🏰 TradeFortress 🏰]

1 Week complete of trading, good profits overall of 4.3 BTC, I am going to remove the 4.3 from the account and start this week back up with 50 again to keep my math going easier for the coming weeks tracking.

Put it in GLBSE or Cryptostocks

[MtGoxTraderBot]

Well profits appear to be ok with the bot, but it looks like we are on a slow downward spiral now so this should be interesting to see if it can keep up. The lower the overall average price, the easier it is to maneuver the trades.

[MtGoxTraderBot]

Update posted. The market had a nasty dip but bot has recovered now and back into the swing of things. I have a guinea pig investor lined up who I am going to test larger amounts with next week and if prices drop this could get really interesting. Working out the kinks on that part now.