Idea For Trading Model / Bot

[insurance2suit.com]

Hi,

I'm thinking about writing a bot for Betfair and was wondering what kind of things might work..

I've got an idea for a trading model which would look something like the following for Horse Racing:

>> Place early bets in the queue a few ticks either side of the traded price.
>> Wait for the price to 'steadily' move towards my bets and wait for them to be taken.
>> Once taken immediately place a bet on the other side to 'scalp' the tick difference.
>> Whilst the open bet is waiting to be matched the model inputs a load of recorded data (fill rate, prices, weighted averages, volume etc.) into an SVM to determine if the price is more likely to stay/move in a favorable direction; or more likely to move against me.
>> If it's the former then keep the bet open, however if not then scratch the bet and continue the cycle until another bet is matched..

This calculation is run every 200ms whilst waiting for the next req/response from Betfair and would probably only really work in the 5 mins or so before the start of the race.

Obviously this is massively simplistic but was just wondering if this is the sort of thing that could work?

Thanks

David

[Franklin1]

not a terrible idea.

Issues I see are:
1 - Do prices move steadily?
2 - SVM is binary - not a problem if prices move steadily, but if not how far/fast they move will be an issue.

[insurance2suit.com]

1) Yes 'steadily' is definitely the keyword there and I'd need to protect against bets ripping through me and the surrounding money is genuine, along with safe guards that if I did get on the wrong side of a trend I could get out quickly.

2) This is why I envisage initially it could only be used in a small time frame ( < 5 mins) before the off in races where there's decent liquidity.

I imagine there will be lots of issues to work out, mainly around liquidity and volatility, but never done anything like this before so wouldn't mind getting an idea if such an idea, in principle, could work; or if it's just a lost cause and all the bots out there at the moment are just way ahead of the game!

[Franklin1]

have you tried doing it manually? Might be a good way to start.

[JayBee]

Hi,

I'm thinking about writing a bot for Betfair and was wondering what kind of things might work..

I've got an idea for a trading model which would look something like the following for Horse Racing:

>> Place early bets in the queue a few ticks either side of the traded price.
>> Wait for the price to 'steadily' move towards my bets and wait for them to be taken.
>> Once taken immediately place a bet on the other side to 'scalp' the tick difference.
>> Whilst the open bet is waiting to be matched the model inputs a load of recorded data (fill rate, prices, weighted averages, volume etc.) into an SVM to determine if the price is more likely to stay/move in a favorable direction; or more likely to move against me.
>> If it's the former then keep the bet open, however if not then scratch the bet and continue the cycle until another bet is matched..

This calculation is run every 200ms whilst waiting for the next req/response from Betfair and would probably only really work in the 5 mins or so before the start of the race.

Obviously this is massively simplistic but was just wondering if this is the sort of thing that could work?

Thanks

David

You are going to trade and then optimise a model to see if you did the right thing.

Interesting.

[edt123]

interesting

have you used this method of data analyiss to predict movement with any success?
I am looking at this are as a way to opfimise myodds rather thanscapling so I'd be interested in how you get on or what exerieces you have.
My experiences were with the tock market and I found the charts of very little practical use most of the time for the simple reason that nearly all the moveenbt you saw was coming form other traders blowing smoke too and precious little was really getting traded a lot of the time.