Remove this ad
avatar

bogorman

rookie botter

Posts: 23 Member Since: August 13, 2013

Lead

August 24, 2013 12:14:50

Tags : :

At the moment i am building a backtesting system primarily focused on horse racing but would be easy to make it more generic to apply to any markets. If i was to offer this as a paid service to other users would anybody be interested? 
Running these simulations in the cloud are fairly CPU intensive so it would have to be a paid service if only to cover the CPU usage. 

1. I wouldn't want to deal with any data issues so this would need to be a bring your own data service and for better or worse the standard for historic betfair data is the Fracsoft data so you would need to convert your own data to match those files or else provide the farcsoft files. This would be done though an S3 bucket. 
2. We'd need to come up with another standard format for uploading your ratings or other numbers which you would want to use in the simulations. 
3. A really user friendly interface for building different systems and managing the simulations. 

This will probably never happen but i just want to gauge interest in such a system. If there is interest then i will try and code it more as a SAAS service and not just for my own needs. I think most bot developers end up building something to do backtesting but its quiet hard to model accurately bets on an exchange especially for complicated models.

As an example of a strategy builder which i would like to mimic have a look at the now closed down quant blocks.

Potentially if you actually wanted to run the model then the system could generate the code for you to download and also all the infrastructure code. Just a though! 

Quote    Reply   
Remove this ad
Remove this ad
avatar

birchy

Betfair Elite

Posts: 593 Member Since:May 11, 2008

#1 [url]

August 25, 2013 00:42:52

This is probably the wrong place to ask because (as you mention) most programmers will develop their own back testing frameworks. I believe it is actually impossible to simulate a betting market using historical prices because you have to make assumptions about the future. The biggest problem is that it is all too easy to back-fit a strategy that shows a profit, only to discover that you can't use it live because of some unknown futuristic parameter. The classic example of this is using BSP for pre-race bets and then realising you will never know the BSP until AFTER a race has started....unless you can develop an accurate BSP predictor. Also, because the markets evolve so quickly, a pattern that existed last week/month/year may not exist in the future because some punter(s) changed THEIR strategy.

Having said all that, humans are creatures of habit so can be exploited...

www.bespokebots.com

"This time next year Rodney, we'll be millionaires!"

Quote    Reply   
avatar

pic

bot addict

Posts: 63 Member Since:July 28, 2011

#2 [url]

August 25, 2013 08:00:12

welcome bogorman on diybetfairbots,

call it a coincidence ;) , but I am building exactly the same.
It is proceeding very slowly, but proceeding anyway.

I doubt the "market" for such a tool is big enough to create a business but I want to create it for myself and, at this point, I like to create something that could be used also by other people.

If I ever get to a public release of the software, I will write a post in this forum.

Quote    Reply   
avatar

bogorman

rookie botter

Posts: 23 Member Since:August 13, 2013

#3 [url]

August 25, 2013 20:28:37

@birchy
Yea i agree and over doing it on backtesting is not a great idea but some is better than nothing. Yea BSP is a good example of what not to do. I don't think its impossible but obviously its never going to match exactly what you would get from the real exchange but its possible to get close if you have enough data. But obviously its a perfect system and real world betting is not with connections over the internet dropping and whatever else can happen. 

@pic
Cheers. Hopefully one of us can get a system up. 

Quote    Reply   
avatar

guiness

rookie botter

Posts: 39 Member Since:August 28, 2012

#6 [url]

October 30, 2013 08:53:21

Hi guys, just noticed this thread..

Are you still working on this? What language are you coding it in?

I'm interested/have started doing something similar.

I have a background in analysis. Let me know if you are still active with this

P.s. Python has some great libraries for doing this kind of thing and plotting nice charts.

Quote    Reply   
Remove this ad
avatar

guiness

rookie botter

Posts: 39 Member Since:August 28, 2012

#8 [url]

October 30, 2013 12:52:24

Hi,

'one day' is a bit slow mate. One day you will be a millionarre also.. the time is now brother!

p.s. What is your background? What language are you using? What kind of features do you think important to have?
 

Quote    Reply   
avatar

bogorman

rookie botter

Posts: 23 Member Since:August 13, 2013

#9 [url]

October 30, 2013 19:31:52

Hi. From my point of view i am probably not going to build a generic service. I do have a working system but its going to take a lot of work to make it generic so that anyone could make use of it. I will integrate FRAC data into it at some point soon and then it could sort of be a standalone application. 

Technology wise i started writing everything in Ruby and i still do a lot of Ruby but the core of the simulator is now in Java just for speed reasons. Using some Scala too. 

Quote    Reply   
avatar

guiness

rookie botter

Posts: 39 Member Since:August 28, 2012

#10 [url]

October 31, 2013 13:15:22

Ah ok, fair enough. Honestly prob best just keep for your own use given the time it will take..

Ruby interesting.  Just woundering as I'm using Python and was woundering about getting together with someone so save the lonely path of coding on my own.. there are alot of useful libraries Im discovering; numpy, scipy + matplotlib simulate Matlab but better programming environment etc

Good luck with it mate

Quote    Reply   
avatar

bogorman

rookie botter

Posts: 23 Member Since:August 13, 2013

#11 [url]

October 31, 2013 16:57:15

Cheers. Yea python is a good choice for you probably as its for a lot of mature libraries. Ruby is less mature for the number crunching but its got the basics but when it comes to writing web scrapers and stuff like that i usually go for Ruby. 
I think most people on here use Python so you are in good company. 

Quote    Reply   
avatar

pic

bot addict

Posts: 63 Member Since:July 28, 2011

#12 [url]

October 31, 2013 22:38:18

@guiness: not only "one day", but "one day *maybe*"

I have already a testing tool but it is something that works only with my code, which is Ruby code.
I have been using Ruby as my main language since 2006 iirc, last year I decided to learn Scala and so I would like to rewrite my botting framework with it now.
I started with the testing tool and the plan is to write something general enough to be used not only by me but also by other botters.

It is a complete coincidence that I have the same recent history about programming languages than @bogorman.
Instead we had the opportunity to discuss together about the testing tool on other "communities" before we meet also here on diybetfairbots.

Quote    Reply   
Remove this ad
Add Reply

Quick Reply

bbcode help