• Hi Guest Just in case you were not aware I wanted to highlight that you can now get a free 7 day trial of Horseracebase here.
    We have a lot of members who are existing users of Horseracebase so help is always available if needed, as well as dedicated section of the fourm here.
    Best Wishes
    AR

Types of system

decastro said:
that jumps up and down like a yoyo. Tiny adjustments in my system can have huge effects on the profit. I mostly ignore the profit. If I had the choice between A) predicting the winner in a race and having a badly wrong rankings for the rest of the horses in the race and B) getting the winner wrong but having almost correct rankings for every single horse, then I choose B.
Click to expand...

Understood. Let me try another tack. Have you compared the efficiency of your rankings for your 500 races against your bigger sample and if so how do they compare.
 
ArkRoyal said:
Have you compared the efficiency of your rankings for your 500 races against your bigger sample and if so how do they compare.
Click to expand...

I didn't do that - and because it takes so many hours to do the bigger test, I'm not about to do that in a hurry. When I did the big test I *was* interested in the profit, so much so that I neglected to test any other aspect of its performance!
 
decastro said:
I didn't do that - and because it takes so many hours to do the bigger test, I'm not about to do that in a hurry. When I did the big test I *was* interested in the profit, so much so that I neglected to test any other aspect of its performance!
Click to expand...

As you can probably guess I am 'concerned' that you are creating a heavily backfitted system. A 500 race sample is very small and with all the sample data being from the summer you are ignoring a lot of seasonal variations that have a massive impact on race results.
 
ArkRoyal said:
As you can probably guess I am 'concerned' that you are creating a heavily backfitted system.
Click to expand...

I'm assuming that by backfitted you mean overfitted. I am acutely aware of the phenomena of overfitting and am armed with a multitude of techniques to either test for it or, preferably, avoid it.

ArkRoyal said:
A 500 race sample is very small and with all the sample data being from the summer you are ignoring a lot of seasonal variations that have a massive impact on race results.
Click to expand...

500 is very small if all you are considering is "win or not win", but I am extracting much more information from each race than that.
 
decastro said:
I am acutely aware of the phenomena of overfitting and am armed with a multitude of techniques to either test for it or, preferably, avoid it.
Click to expand...
Best advice I can give is to try to ignore your expertise in statistics and AI and think about improving your lack of racing knowledge.
 
decastro said:
500 is very small if all you are considering is "win or not win", but I am extracting much more information from each race than that.
Click to expand...

If it's profitability your after you'll be better off with winners. Winners tend to pay better than losers even if they're in the correct finishing order

What machine learning algorithm are you using?
 
Horseplayer said:
What machine learning algorithm are you using?
Click to expand...
I'm not actually using anything you could really call AI... I'd say it was just statistics - combined with some complex data processing, including some computationally expensive iterative processes (that I'd rather keep to myself).
 
Also a knowledge with odds with BF sp not only the 5% but the bigger prices if trying get much on can bring the price in also.

Maybe if you put up your base selection ideas people could soon see if they are viable or not to be honest i do think many play from our ideas because in most cases to much waiting and work involved to run them right.

But Good Luck!
 
I just had it confirmed from proform that the commission is not subtracted from the (exported) odds I am using. :oops:

Mind you, exchange odds appear significantly more generous than SP odds, so maybe that could be my saviour.
 
Last edited:
decastro decastro Please do not take this the wrong way, just wanted to give a heads up on our ethos and etiquette.

Forums are a great place to learn from and you should not be surprised by the fact that those who interact and contribute to the discussions get much more back from the forum than those who just want to take without putting anything back.

I'm not saying that you are expected to 'tell all' but posts such as the following will not win you any friends.

decastro said:
The details are top secret
Click to expand...

decastro said:
including some computationally expensive iterative processes (that I'd rather keep to myself).
Click to expand...
 
So this brings me back to my original post in this thread. How could I ever prove to anyone that my (type A) system is profitable? If it was a type B system then I just post up my predicted winners in advance and everyone can look at the SP odds. But if I have a system with a set of probabilities for each horse and the betting depends on finding value on the betfair exchange, then what do I do? The exchange odds bounce up and down all the time, how do I prove what odds my system got? Do betfair offer a proofing service?
 
decastro said:
Do betfair offer a proofing service?
Click to expand...
No, but they do provide some details on the maximum and minimum odds matched but not do not tell you when these odds were matched or for how much. I believe that Fracsoft offered BF prices with a timeline and amounts available at a cost per race.
 
You must log in or register to reply here.
Back
Top