After managing to import historical data into the BA Excel template, doing a search on certain race parameters will in most cases return a certain number of races with actual win/lose outcomes. Converting the win/lose ratio to probability then to odds (assuming I have done it correctly) allows a comparison of these "calculated odds" v the Exchange odds on a rank v rank basis.
There are differences between the 2 sets. Do these differences represent "value" bets over the long term is what I am wondering?
Would this calculated probability fit with "Kelly" staking is another thought
Implied v Actual Probability
-
CallumPerry
- Posts: 575
- Joined: Wed Apr 19, 2017 5:12 pm
- Location: Wolverhampton
Could you not pump this back into your historical data and simulate over approx 10K markets (or whatever amount) with flat staking to see firstly, if this has provided you with an edge and then secondly, if it has, compare your equity curve to one from another test where you are using some variation(s) of the Kelly Criteria to see if you can squeeze even more profits out?
The only thing to realistically consider is how much volume you can actually put through the markets but in this situation, it's just a value betting kind of thing you'd only need to get in with a single bet and not out with a closing order?
The only thing to realistically consider is how much volume you can actually put through the markets but in this situation, it's just a value betting kind of thing you'd only need to get in with a single bet and not out with a closing order?
That screenshot has 201 races worth of data compounded to a single probability for race 202. Race 200 to Race 1 will each have a set of probabilities based on Exchange odds v Calc odds so can`t suss out an easy way to do it yet. If the general conscensus was it`s value then might devote a bit of time to it.
-
ShaunWhite
- Posts: 9731
- Joined: Sat Sep 03, 2016 3:42 am
Looking back that's fine but Kelly is about individual stakes.
Which out of the dogs today has a 55% chance priced at 66% and which is a dog priced at 47% when it's a 51% chance. The data says you're a winner backing both, but how do you stake it? People using Kelly usually run a rating system of some sort and if you're going to use technical analysis rather than fundamental analysis I think you're better off sticking to level stakes.
Someone like Archery stands a chance will Kelly because he's essentially creating a ratings system by looking at the underlying data (FA), not just the price data (TA).
Which out of the dogs today has a 55% chance priced at 66% and which is a dog priced at 47% when it's a 51% chance. The data says you're a winner backing both, but how do you stake it? People using Kelly usually run a rating system of some sort and if you're going to use technical analysis rather than fundamental analysis I think you're better off sticking to level stakes.
Someone like Archery stands a chance will Kelly because he's essentially creating a ratings system by looking at the underlying data (FA), not just the price data (TA).
-
ShaunWhite
- Posts: 9731
- Joined: Sat Sep 03, 2016 3:42 am
...how are you deriving that WoM figure?
-
Archery1969
- Posts: 3219
- Joined: Thu Oct 24, 2019 8:25 am
- Location: Newport
Hi,Atho55 wrote: ↑Thu Jun 18, 2020 12:13 pmAfter managing to import historical data into the BA Excel template, doing a search on certain race parameters will in most cases return a certain number of races with actual win/lose outcomes. Converting the win/lose ratio to probability then to odds (assuming I have done it correctly) allows a comparison of these "calculated odds" v the Exchange odds on a rank v rank basis.
There are differences between the 2 sets. Do these differences represent "value" bets over the long term is what I am wondering?
Would this calculated probability fit with "Kelly" staking is another thought
Implied v Actual Probability.jpg
What method did you use to come up with the Calc Odds column ?
Cheers,
Calc odds come from Calc Probability
That is correct Rik due to the fact that we are only selecting the data to match the number of runners in the upcoming race. More runners will gather more % increments
That is correct Rik due to the fact that we are only selecting the data to match the number of runners in the upcoming race. More runners will gather more % increments
You do not have the required permissions to view the files attached to this post.
-
Archery1969
- Posts: 3219
- Joined: Thu Oct 24, 2019 8:25 am
- Location: Newport
Hi,
Not a criticism, so if a selection has never won a race then you don't calculate its probability ?
Cheers,
-
Archery1969
- Posts: 3219
- Joined: Thu Oct 24, 2019 8:25 am
- Location: Newport
-
ShaunWhite
- Posts: 9731
- Joined: Sat Sep 03, 2016 3:42 am
Those values are just straight unmatched values. Started to experiment with taking 5s worth of matched £, multiply it by Back or Lay WOM% to give a £ value per either Back or Lay then divide this £ value by 5s to give a flow value. It`s set up but bolted that many other calcs on the sheet it`s running a bit slow.
Looks like this. Col K is the 5s £ value, Col`s AB & AC are the £/s col`s
Work in progress....
Looks like this. Col K is the 5s £ value, Col`s AB & AC are the £/s col`s
Work in progress....
You do not have the required permissions to view the files attached to this post.