Automation- Is it for the masses ?

[Seymour]

Not that long ago Peter Le created an excellent post entitled 'Building a Strategy', which hooked me from the first few lines, as behind my ongoing attempts at profitable trading, I knew I was more of a strategist than anything else. Jump forward in time and I am now running a value based semi-automated 'system' in practice mode that appears to hold future promise. Or, if I seriously consider the masses of people involved in Betfair software development looking for the same edge, perhaps I am just fooling myself.

In the interests of boosting my own confidence as well as the prospects of other forumites, I would love to hear from anyone who has successfully created or is currently benefiting from either a fully or semi-automated stand-alone strategy using Guardian. The idea of full automation seems like a dream but is it really possible?

Thanks again to Peter Le for the kickstart. Even if this particular strategy fails in time, I have still learnt an enormous amount and further developed my programming skills.

regards
Jon

[PeterLe]

Morning Jon,
I'll dip out of this so that I dont give a biased view and you may get a balanced view by other postings.
I just wanted to add though. Whilst the Practice mode is a god send and great to test initial theories; you will get different results from running it live.
I have run an identical spreadsheet on two accounts simultaneously over the course of a day, one on a live account and one on practice, with different results. So try that
(Make sure you have a safety net ie you could use the excel command "small & Large" in Col 9 which shows your liability/Stake across each runner. Coupled with that you will need to ensure you put a command in that makes sure you dont try laying at high odds etc..If fact put as many safety nets in as possible (you can always remove them if need be)
..and move money to the Aus book leaving a relatively small amount in the main wallet

[EyePeaSea]

Hi Joe,

Not an expert on here by any means, but I'm working on two systems that are completely automated. However, my systems rely on just maths, rather than the skill and intuition that (most/all?) of the successful traders seem to employ and therefore, my systems are easier to automate.

I can only echo what Peter said about safety nets. When I built my trading engine (that I just plug new systems into), I spent 2 weeks putting in checks (maximum liability, preventing duplicate trades etc.) and it's certainly been worth the effort.

Peter - interesting what you say that about getting different results from the Practice mode. At the moment, I'm not making trades (building up confidence in the new system), but always run BA so that I get my data. I'll make sure that I now always run it in live mode!

Ian

[Seymour]

Thanks for the replies Peter and Ian.

I find Bet Angel provides me with a lot of enjoyment and the idea of immersing yourself in a project such as constructing an automated trading bot is very rewarding- if not financial There are times though, when I need a reality check in order to keep my sense of perspective on what I am doing. I do this by searching the net for info on bot construction, triggers, what others are developing etc, etc.. Up until recently I had no idea just how huge the entire Betfair phenomenon (and assoc. businesses) is in Europe, and its quite mind blowing to the uninitiated. This recent survey has left me with some doubts given my feeble capacities compared with many others who are actually paid to study Betfair sports markets and/or software. The concept behind my automation is quite simple which also tends to raise concern, if not alarms in my mind. If its that bloody simple and it appears to work, why isn't every one building bots?

I don't want to know the details of anyone's automation, just whether there really are people out there that pay for the groceries each week, by running an automated bot. I wouldn't have thought this was a very big request, unless of course these individuals are few and far between.

Anyway, thanks for the advice on running practice mode vs. live tests and safeguards gentlemen.

regards
Jon

[Fidib665]

Hi Seymor,
I think, I have built a profitable betting strategy (ie, to be precise, it is a punting strategy since I never green up any profits during the game).
I have been involved in BetFair markets for some 4 years now (soccer only) and recorded some 30k + games by extracting for each game the data relevant to my strategy about every minute. Based on this data sample (sp,e 20 GB of CSV files by now), the strategy (depending on the exact calibration of parameters) in back testing makes about 20% per trade, in particular for lays, not so for backs, where it is more in the 10%s per trade. I split the sample into in-sample and out-of-sample, with similar results for both samples, so I am pretty confident that the strategy indeed works (the total number of bets that my strategy produces over this timeframe is some 20k bets, so I am also quite confident that this is not just a statistical abberation).
I am live-betting since last November with minimum (4 EUR) trades and have made ~900 EUR from it. Certainly not enough to pay my groceries yet, but I consider myself to be in the testing phase still and there is no problem in scaling the amount bet up by a factor of 10 or 100, once I am sufficiently confident on the strategy: The strategy indicates the trades in particular for the huge games (ie, internationals and major leagues) where market depth is sufficient.
In terms of automation, I use the guardian together with a linked Excel sheet, load up the games in the morning (which takes me about 2 minutes per 10 games) and then leave it alone over the day where it fires the bets fully automated into the market and records the market data at the same time for future backtesting exercises. Since it is a punting strategy with relatively low stakes, the operational risk of the strategy is also pretty neglegable (compared to a trading strategy that involves putting huge amounts of money into the market and then removing it after the market has moved by some ticks).
So, while I still do work fulltime, I hope of earning a decent income from my strategy once I fully scale up the amount bet.
Hope that at least partially answers your question.
Many greetings, Fidibus

[GMBing]

I have used the excel link to run fully automated trading system and they do technically work really well - The Problem is a fully automated system needs a definitive set of rules to open a trade and likewise close the trade - And therein lies the problem, is it likely the rules that apply to a 5 runner race with two closely priced favourites would be the same as a 16 runner open handicap. - I now use excel automation to collect, monitor and provide me with the key market indicators I want to see in the market but open and close trades manually base on my analysis of this data - and it works very successfully.

Regards
G

[PeterLe]

Hi Seymor,
I think, I have built a profitable betting strategy (ie, to be precise, it is a punting strategy since I never green up any profits during the game).
.....

Hi F
I havent read on here about anyone automating the football before..thats interesting.
What I tend to do where I know I have an edge is to compound.
What i mean by this is that the stake size is dynamically linked to the P&L in the spreadsheet. Depending upon the strike rate it may only increase the stake by a fraction of one percent, its scary how quickly it grows. In the early days on betfair, there were times I had to switch the screen off as I can see how much was being fired into the market!

Im not sure about the football (as there is plenty of liquidity), but in play on the horses I hit a point of diminishing returns. If you plot a graph over many markets you will be able to tell at which point the stakes were at the optimum.
Regards
Peter

[Seymour]

Thanks for the responses everyone.

Fidibus, it sounds like your getting closer and closer to the Holy Grail of Trading, that's wonderful Although my own focus is on horse racing its good to see someone excelling (sorry about
that ) in such a large volume market and up against traders who live and breath football. Unfortunately my choices are limited to horses in play due to Australian laws. Please keep us posted as to how your strategy develops.

GMBing, yes, I have also found the biggest problem with auto lies in defining rules for opening and exiting trades. At the beginning of this project I sat by watching Guardian over many nights in order to manually correct and cancel bets that could leave me liable. Semi-auto appears to be the safest method, short of writing 20-30 rules in an attempt to cover most possibilities.

Peter,
'there were times I had to switch the screen off as I can see how much was being fired into the market!'

'Im not sure about the football (as there is plenty of liquidity), but in play on the horses I hit a point of diminishing returns. If you plot a graph over many markets you will be able to tell at which point the stakes were at the optimum.'

From what I have read, it seems running automation opens you up to some sort of ethereal 'grey' world where tests cannot always be repeated, practice mode does not mirror reality and larger stakes just aren't possible. As my strategy is based on seeking value on selections I just hope the volatility that gives rise to these values doesn't start to diminish. Leaving aside longer 3m races, I have noticed many Aust. races often seem to lack the odds fluctuation ranges of UK racing.

Jon

[Fidib665]

What i mean by this is that the stake size is dynamically linked to the P&L in the spreadsheet.

Dear Peter, thx for that idea. For the time being, I think, though, that I will stay with fixed size bets due to a) risk diversification and b) reduction of operational risk (the larger the individual bet size, the bigger the risk if something really blows up - in particular on the lay side). However, what I don't do yet but want to include is to link the bet size to the expected value of the bet.
Will however come back to your idea, once I am sufficiently sure on the stability of the system that I set up.
Many greetings, Fidibus

[LeTiss]

I have a bot on horses I'm trying to develop - I think it needs a bit of work though, as in May it made 768 trades & I made a profit of 23p

If you hate horses like me, smallish stakes automation is ideal, as there's no emotion involved. All I need to do is keep analysing what's happened & try to iron out the mistakes

With regards to sport, I have a pre-event bot that fires a £2 bet at 1000 on the fav. This is just to identify a potential opportunity based on what I look for, as there are too many markets to keep checking them manually. I then can race across to see whether I want to get involved or not, so the placing of trades is still done by me. My brain will consider many things that would be impossible to build into a bot

I prefer BA automation to excel. I find Excel becomes unreliable and sticky when there's too many markets on the sheet. BA Guardian is brilliant, as it happily works through the markets with no disruption. If Peter and his team can develop automation further, so it has Excel capabilities, then that would be exciting, or perhaps incorporating bookmaking + dutching into the program

[Seymour]

Hi Le Tiss,
I would become quite excited if I had a bot that made 23p over 768 races. Reaching a point of breaking even can be a glorious thing. It sounds like you haven't got too much tweaking before its profitable.

In regards to BA automation, at one point I found myself wishing for additional rules based on the relativity of other selections, which lead to a flurry of writing 10-15 rules for the other horses, something I would rather not repeat too often.
Yes, I agree it would be great if Guardian had other options.

I get the feeling that there are many who would like to develop automation but find Excel a bit too daunting to undertake. For our purposes, I would like to suggest to anyone considering automation, that it is very possible. There is an excellent on-line free course which is designed for complete beginners, available through the University of Reddit at:

http://excelexposure.com/

regards
Jon

[Zenyatta]

Yes, I'm a fan of automation. It's easy to break-even on the horse racing In-Play. In fact, in just a couple of days after I first started using BetAngel, I had an In-Play bot which went weeks without losing anything.

Unfortunately, I made a disasterous 2-year detour into manual trading pre-off, which I never got to grips with at all. I had to finally put a line through the manual trading, and am looking forward to getting back to automation only.

The main advantage of automation is that it doesn't take much of your time: you can just set-it and forget-it. Whereas, manual trading is a huge time-sink. Another advantage of automation is that it takes the emotion out of it: the automation will implement your system perfectly and doesn't suffer from fear or greed.

The Excel spread-sheet is best for Inplay, because it can put bets on multiple runners simultaneously.
BetAngel Automation rules are best for the Pre-off, because it can handle more complex trading where timing is important. I usually used two copies of Betangel running simultaneously on a VPS.

[EyePeaSea]

The main advantage of automation is that it doesn't take much of your time: you can just set-it and forget-it.

Clearly there are different types of automation

I code all my own automation routines, rather than using BA automation. It takes months of hard work to think of a system, code it and then test it. And then it may not be viable (for a number of reasons) and so you start the cycle all over. My systems run without me during the day, but don't forget all the effort that goes into creating the system in the first place.

Ian

[MichaelLennox]

Guys,

I would say it is not for the masses. It has taken me 2 years to develop a trading robot in excel for tennis that works. I tried almost everything from moving averages, weight of money, Forex training methods like Fibonacci levels. Each failure taught me more and I eventually got there. Technical trading does take need a lot of maths particularly when handling noisy data like tennis odds.

Basically if you want to use automation the safest way is to run it as a failsafe in live trading. This is how the city traders do things. The guys who write these trading systems in the city are Maths grads, most of the disasters you hear are humans putting high risk trades into the market not the trading systems. All the betting companies are the same too, at William Hill's trading floor it’s all young number crunchers.

So if you don't have a flare for numbers don't do it in my opinion as a 100% solution. If you want to use it as a failsafe, get some strong simple rules together and see if it's for you.


Regards

M

[LeTiss]

MichaelLennox

That's an excellent post, and I think you are not far off the mark
This would go some way to explaining the extraordinary success of Peter Webb, who has a financial background, as opposed to a betting one.