Running Bots (Automated betting)

[adrianmfoster]

I have written an automated betting program using ba and excel which runs well and wins consistently, winning about 4 out of 5 bets, every day for about 7 days. Its then like someone turns a light out and it suddenly stops winning. I have tried the program about 6 times and each time its the same, about 7 days of winning followed by a sudden turn round.

I thought the first few times the winning was perhaps lucky, though the theory is sound, or computer speed, but it follows the same pattern every time.

Anyone any ideas or conspiracy theories?

[2020vision]

I have written an automated betting program using ba and excel which runs well and wins consistently, winning about 4 out of 5 bets, every day for about 7 days. Its then like someone turns a light out and it suddenly stops winning. I have tried the program about 6 times and each time its the same, about 7 days of winning followed by a sudden turn round.

I thought the first few times the winning was perhaps lucky, though the theory is sound, or computer speed, but it follows the same pattern every time.

Anyone any ideas or conspiracy theories?

Hi Adrian

Sounds promising but looks like you have spotted a trend. Have you tried
stopping after your winning sequence and then starting it up again after a
few days of non-use? Just curious? Also how many events or markets
per day or hour does it run? Are you sure it's not down to your triggers?
Do you ever switch off your computer between sessions?
Perhaps someone has found a way to "provoke" it into losing? I doubt it will
be down to your computer speed. Just throwing a few ideas at you because
they may trigger (no pun intended) a thought that cracks it for you?

In the meantime let us know how you progress - Michael

P.S. I am just trying to learn this stuff myself so please bear with me if I sound like a novice!

[mak]

hi
i don't know a lot for sure,but if this 7th day is Saturday probably it will be the speed of the api.

[adrianmfoster]

I run on every race every day, turn off my computer each evening and on again the following morning, and it just doesn't stop winning for the one day, it loses every day from then on.

I think the only way forward is to run it for a week then work out how long I need to stop before it will run successfully again, but this does suggest the bot is being watched from above.

[Spike]

It's almost certainly just normal variance and a bit of coincidence, clustering is normal in randomly distributed results. Certainly no-one is "watching from above," that kind of thinking is ends in believing in Leprechauns. Assuming that you're in an active market and you're not putting in massive stakes it's virtually impossible for anyone else to "see" your activity amongst the maelstrom.

It is conceivable that some days of the week are better than others, but that would be easy to prove.

Finally- if what you're seeing has a real underlying structure then you can easily overcome it (or at least prove it) by running for 3 days, stopping and starting again, that way you will never loose right?

The simplest and most likely explanation is that your strategy just isn't striking value bets. If I were you I'd work with that hypothesis.

[adrianmfoster]

what do you mean there are no leprechauns, next you'll be telling me there's no santa claus!

[danjuma]

Funny you have noticed this adrianmfoster, same thing tends to happen to me. Normally I spend hours developing a strategy and writing a spreadsheet for it, and test it with small stakes and it's profitable during the trial. As, soon as I think it's ok and increase my stakes slightly, it start's losing. I might be paranoid, but at the back of my mind, I am starting to think whether BF has ways of monitoring these things and then implement ways of affecting the strike rate! Just a thought.

[osknows]

One thought - If you're developing strategies based on historic data you may be experiencing overfitting of your models. It's easy to come up with a set of rules that perfectly fit a set of data but as soon as you use this to predict future events/prices etc the model becomes less accurate. The common way around this is to split your data into Training, Test and Validation sets and test across each other. If the models developed using the training data remain accurate using the test and validation data then the model should be more robust than using all data combined.

[jokerjoe]

Also as you increase stake size you increasingly affect market bahaviour. Not easy to get around but if you've got the time and enough flexibility on price you could drip-feed into a position.

[doris_day]

How many markets has your bot played in ? As others have said, you're likely to be looking at normal variance as your numbers are probably quite small at the moment. You really need a few thousand trials before you can make generalisations.
As others have also said, look at the days of the week, time of the day and type of event. They all will have different patterns of liquidity and therefore will respond differently to your bot. There's nothing like collecting all the data you can and analysing it once the numbers get big.
I've been there, done that and can assure you that once you've got enough data, you'll probably be able to see the patterns.

[adrianmfoster]

To reply to the points made.

I am an ex computer programmer/systems analyst of over 15 years experience and can assure you I tested my system thoroughly before I started as I am also a tight git!

My 'system' is profoundly simple, so simple I am sure others out there pursue the same tactic, but simple is often best. In theory, the system can not lose, but obviously with speed factors etc, it can. On each of six occassions I have run the system, it wins consistly, 80% of all bets placed, but only for the first 7(ish) days. It then proceeds to turn around and completely reverse the win percentage winning only about 20% of bets placed. It does not matter if it is a saturday or wednesday. It has done this each time I have run the system which I believe is too much of a coincidence, so leprechauns or not, something, which I cannot explain but only guess at, effects my operation.

My main reason for this post is to see if this is a common occurance or notas an aid to diagnosing the problem. Obviously if it happens to alot of people, epecially placing predictable bets, the question of bet intercepion is a possibility, if it is not common then it is not.

[knot]

lol. 7 days

[danjuma]

To reply to the points made.

I am an ex computer programmer/systems analyst of over 15 years experience and can assure you I tested my system thoroughly before I started as I am also a tight git!

My 'system' is profoundly simple, so simple I am sure others out there pursue the same tactic, but simple is often best. In theory, the system can not lose, but obviously with speed factors etc, it can. On each of six occassions I have run the system, it wins consistly, 80% of all bets placed, but only for the first 7(ish) days. It then proceeds to turn around and completely reverse the win percentage winning only about 20% of bets placed. It does not matter if it is a saturday or wednesday. It has done this each time I have run the system which I believe is too much of a coincidence, so leprechauns or not, something, which I cannot explain but only guess at, effects my operation.

My main reason for this post is to see if this is a common occurance or notas an aid to diagnosing the problem. Obviously if it happens to alot of people, epecially placing predictable bets, the question of bet intercepion is a possibility, if it is not common then it is not.

I am sort of inclined to agree with you. Who knows for certain that the big exchange don't monitor these trades/bets and manoeuver it in such a way to ultimately benefit them. I read somewhere that they have their own employees trading/betting the markets.

[Captain Sensible]

A lot depends on your strategy but it can get easy to spot bots coming into play on the markets especially if you're trading them manually them day in day out. I know of a few bots out there, there entry points and entry times and will take advantage if conditions permit me to.

[adrianmfoster]

Captain sensible

If you don't mind me asking do you spot triggered bets manually or do you have an automated bot spotter