Open GoogleCodeExporter opened 8 years ago
Ow.. Seem to get it.. Is the idea not to flood the mtgox with orders that are
too far away from the current price, but to wait for the 0.1% or less of a
price difference between current price and the target order and just then make
the actual order?
If so, maybe a more appropriate default for a price difference would be a bit
more than the 0.1%? So the order is already in the mtgox queue when the price
drops far enough so the order is fulfilled immediately?
Original comment by purge...@gmail.com
on 2 Jan 2013 at 10:49
It's funny that you should bring this up, I literally just checked in a change
to increase the window to 5%
The reason for gating the order is just as you suggested, to prevent orders
which are too far from the current price.
However, there is another need for this gate. bcbookie scans the bids from
highest to lowest priority, when an 'active bid' (which just means it meets
some sort of filtering logic) is found it breaks from the scan. Without some
sort of gate you could end up with the case where a higher priority bid may be
15% away from the current price would block a lower priority bid which may be
only 1% from the current price.
The $-1 target was for debugging. There are a couple of different cases where
bid_maker sets the bid to zero. The -1 was so I could differentiate between
them. You can ignore the warning.
Also, you no longer need to run report_gen.
report_gen is most probably going away (along with wc_server). report_gen takes
to much processing time to generate charts that 99.99% of the time no one will
look at. bid_maker is single focused on generating the bids. To cover the
reporting gap, I'm thinking about how to implement charting on request through
the node server (which is replacing wc_server).
Thanks for the feedback and Happy New Year!
Original comment by brian.mo...@gmail.com
on 2 Jan 2013 at 11:12
Although I marked the original issue as fixed, there is still room for
improvement. Suggestions are welcome :)
Original comment by brian.mo...@gmail.com
on 2 Jan 2013 at 11:17
Changed status to accepted.
Original comment by brian.mo...@gmail.com
on 2 Jan 2013 at 11:18
How long should it take on average before the system starts trading? I've been
running for a few weeks now, but no trade has been initiated by the system yet.
Bcbookie is running. I did restart on a new gene pool a few times, but the old
ones have been running a few days, using several computers (about 22
connections on average). Any hints?
Original comment by tim.st...@gmail.com
on 12 Feb 2013 at 11:55
Depends a lot upon your 'active' gene configs and calculated genes. If none of
them suggest a trade in the range up to 10% lower [AFAIK] than the current
price, than the bcbookie will just do nothing wait for said conditions to
occur. You can try to add a gene_def that will trade with minimal markup
[~1.2-1.3%, dependent on your mtgox commision rate], so it will do trades more
often [check the report_gen output for a overview of the frequency and the
spread of the trades]. If the trades are plentyfull, but tightly grouped around
the few spots along the timeline, you can try to tweak the gene_def parameters
stbf [check bct_alt.py for a rough description] and buy_wait [number of minutes
that config will wait between trades].
Are you running the bid_maker? AFAIK, he's the one that scans BOBs and marks
the appropriate ones as potential orders for bcbookie.
Brian, feel free to correct me about all the stuff I got wrong =]
Original comment by purge...@gmail.com
on 12 Feb 2013 at 12:13
10% looks a bit more like 5% ;)
Original comment by purge...@gmail.com
on 14 Feb 2013 at 9:35
Original issue reported on code.google.com by
purge...@gmail.com
on 2 Jan 2013 at 10:16