Question for the mathmatically inclined

[Daytripper]

There are 3 horses in a race on a track thats rated a slow 6.

Horse A has had 1 start for 1 win in the wet Horse B had had 5 starts
for 2 wins and 2 placings in the wet Horse C has had 10 starts for 2
wins and 3 placings in the wet

Forgetting the class of field that they competed against in their wet
track appearances. This is a pure maths/numbers question.

Obviously Horse B has the better wet track credentials.
How do I convert that into a number or rating.

I could simply divide the number of placings into the number of starts
but that would mean Horse A = 100 Horse B = 80 Horse C = 50

How can I make the rating that takes into account the number of starts
that they have had.

In the example above my manual rating would be :
Horse A = 50
Horse B = 90
Horce C = 60

Any ideas ?

 

[Home Alone]

So you want to discount the 1 start for one win scenario?

I think you are looking for a mathimatical formula based on your assumption/preferences.

Can I ask what you manual rating would be for

Horse A - 2 starts 2 wins
Horse B - 10 starts 4 wins 4 placings
Horse C - 20 starts 4 wins 6 placings
(ie double the start for the same percentage wins).

 

[Daytripper]

So you want to discount the 1 start for one win scenario?

I think you are looking for a mathimatical formula based on your assumption/preferences.

Can I ask what you manual rating would be for

Horse A - 2 starts 2 wins
Horse B - 10 starts 4 wins 4 placings
Horse C - 20 starts 4 wins 6 placings
(ie double the start for the same percentage wins).

I don't want to discount it completely, just that I want to give some weighting to those horses who have performed consistently on wet tracks.

Your scenario above, my manual rating would be :
Horse A : 75
Horse B : 80
Horse C : 65

 

[sethlad]

I guess there's 2 parts to any equation:

1.How to calculate the ratings &
2.How to turn them into prices.

Everyone who uses ratings seems to have their own theories,got a mate who swears by neurals...Im yet to be convinced.
It amazes me the number of punters who buy ratings but have no idea how to use them or frame a market with them.

So many use the old "add the ratings & then /100"
IMO thats next to useless,heres a simple theory Ive used.

Give each horse a score,based on its position in relation to your top rater.
Eg:1 behind top rater=90 pts, 3 behind = 70pts
(Obviously an example only,figures need to be based on results/history etc)

This can be a bit trial & error but like anything,once you have your system,it works itself out.
Then simply add the scores,divide by the individual horse score & you have a 100% market.Adjust slightly for average bookie market.

From memory you work for a bookie so this will hardly be news.
Like anything,its all about your ratings being accurate in the 1st place.

seth

 

[Home Alone]

just that I want to give some weighting to those horses who have performed consistently on wet tracks.

That is the discounting I meant.

Not going to be exact, but I think you are looking at something along the lines of

Z * (% win + (%wins + %placings)) = rating

where z is a factor determined by your historical data.
ie 1 start = .25
2 starts = .375
5 starts = .75
10 starts = .86

A mathematical formula to determine Z could also be done where the basis would be somthing along the lines of :-

1
________________

1 + (x + 3)

It not exact, but the factor increases and approaches 1 with the increase in starts.

 

[Daytripper]

I guess there's 2 parts to any equation:

1.How to calculate the ratings &
2.How to turn them into prices.

Everyone who uses ratings seems to have their own theories,got a mate who swears by neurals...Im yet to be convinced.
It amazes me the number of punters who buy ratings but have no idea how to use them or frame a market with them.

So many use the old "add the ratings & then /100"
IMO thats next to useless,heres a simple theory Ive used.

Give each horse a score,based on its position in relation to your top rater.
Eg:1 behind top rater=90 pts, 3 behind = 70pts
(Obviously an example only,figures need to be based on results/history etc)

This can be a bit trial & error but like anything,once you have your system,it works itself out.
Then simply add the scores,divide by the individual horse score & you have a 100% market.Adjust slightly for average bookie market.

From memory you work for a bookie so this will hardly be news.
Like anything,its all about your ratings being accurate in the 1st place.

seth

I use the neurals too and it probably makes up about 65% of my overall rating.
However the neurals are hopeless when it comes to assessing wet track form which leads to my present predicament.

The last few seasons I've simply skipped any meeting with tracks rated slow or worse but it seems a fairly draconian measure when I'm sure there is a solution out there.

Of course, once I have the final score I then convert into markets - which as you have shown is relatively simple (I basically use the same maths as you).

 

[Daytripper]

That is the discounting I meant.

Not going to be exact, but I think you are looking at something along the lines of

Z * (% win + (%wins + %placings)) = rating

where z is a factor determined by your historical data.
ie 1 start = .25
2 starts = .375
5 starts = .75
10 starts = .86

A mathematical formula to determine Z could also be done where the basis would be somthing along the lines of :-

1
________________

1 + (x + 3)

It not exact, but the factor increases and approaches 1 with the increase in starts.

I like it .

Will give it a trial and see how it goes.

Thanks a lot for that.

 

[sethlad]

[QUOTE=Daytripper;7419456]I use the neurals too and it probably makes up about 65% of my overall rating.
However the neurals are hopeless when it comes to assessing wet track form which leads to my present predicament.


I heard that neurals are crap in the wet but why is that?
Isnt there a WT neural that supposedly takes that into account?

seth

 

[Yaaablett]

[QUOTE=Daytripper;7419456]I use the neurals too and it probably makes up about 65% of my overall rating.
However the neurals are hopeless when it comes to assessing wet track form which leads to my present predicament.


I heard that neurals are crap in the wet but why is that?
Isnt there a WT neural that supposedly takes that into account?

seth

What are neurals?

 

[Daytripper]

[QUOTE=Daytripper;7419456]I use the neurals too and it probably makes up about 65% of my overall rating.
However the neurals are hopeless when it comes to assessing wet track form which leads to my present predicament.


I heard that neurals are crap in the wet but why is that?
Isnt there a WT neural that supposedly takes that into account?

seth

No idea why - the programming must be screwed.

I ignore that WT neural as I do the TIME neural.

 

[sethlad]

What are neurals?

Its a maths algorithm that takes into account the traditional factors,eg weight,distance,trainer etc & produces a rating figure for each horse.
Available on the R & S site

seth

 

[Daytripper]

Its a maths algorithm that takes into account the traditional factors,eg weight,distance,trainer etc & produces a rating figure for each horse.
Available on the R & S site

seth

Its taken me nearly 5 years to get my own neural settings to where I want them to be at and producing consistent figures.

Even so there is a hell of a lot they don't include.
For example as mentioned the WT and TIME neurals are not correct, the neurals don't take into account weights and the barrier position neural throws up suspicious figures at times (I ignore that one also).

First up and 2nd up form is also ignored so horses that are first up tend to get a terrible rating.

For the unintiated, don't expect to go in there and have it producing winner after winner because it won't.

I still spend another 30 minutes or so per race adjusting the ratings to include factors such as the above.
And its still nowhere near perfect !!!!

That said, it can be a more than useful tool.

 

[Home Alone]

A mathematical formula to determine Z could also be done where the basis would be somthing along the lines of :-

1
________________

1 + (x + 3)

It not exact, but the factor increases and approaches 1 with the increase in starts.

And before someone picks me up, that formula doesn't approach 1 as the starts increase. It approaces zero. Doh. So don't use that


The formula probably should use logorithms in some shape or form with an upper factor limit of 1 .

 

[kolchak]

Rate a starter's race, on the ratings of other starters Then look for recent races and consistancy. I like to take prize money into account when looking at results...Can't actually come up with a rating figure, just a table of starters races for a race, which are ranked, then look for things like weight and barriers and prize money and dates. If I see a starter which has a heap of races ranked up the top, I usually place a big bet on it, if I can't see any reasons why it's races are not good enough.

 

[Daytripper]

Here's another one.

Just trying to work on a formula to calculate how many points a jockey should earn in a 1-2-3 jockey challenge event depending on the odds given.

This is what I've come up with.
There are 6 possible points to be earned from any one race (3 First + 2 Second + 1 Third).
My formula is to get the percentage of the horse quoted/rated at and multiply that into the 6 possible points. So if a horse was $4.00 (25%) the likely average jockey points earning would be 1.5 (6 * 25%). This is however flawed as a horse at $2.00 (50%) would get a 3 average possible jockey points - ie it assumes that the horse would win every time. To magnify it even further, a horse at $1.50, would earn 4 average possible points which is impossible as the most you can earn in any one race is 3.

So to rectify, I came up with an arbitrary figure of 0.66.
ie: Formula now is (6 points * horse %) * 66%

A horse at $2.00 would now have an average earning figure of 2 which seems a bit more reasonable.

However am a bit miffed that I had to come up with an arbitrary figure of 66% instead of making the figure more mathmatically certain.

Any ideas, anyone.

 

[KardiniaPark]

I dont understand why horse A isnt the best horse in the wet? Its undefeated in the wet. Maybe it lost its first five races on a good track and had one race in a group 1 on the wet and won by 12 lengths.

 

[Daytripper]

I dont understand why horse A isnt the best horse in the wet? Its undefeated in the wet. Maybe it lost its first five races on a good track and had one race in a group 1 on the wet and won by 12 lengths.

Refer my post #15 please.

 

[Lensen]

Here's another one.

Just trying to work on a formula to calculate how many points a jockey should earn in a 1-2-3 jockey challenge event depending on the odds given.

This is what I've come up with.
There are 6 possible points to be earned from any one race (3 First + 2 Second + 1 Third).
My formula is to get the percentage of the horse quoted/rated at and multiply that into the 6 possible points. So if a horse was $4.00 (25%) the likely average jockey points earning would be 1.5 (6 * 25%). This is however flawed as a horse at $2.00 (50%) would get a 3 average possible jockey points - ie it assumes that the horse would win every time. To magnify it even further, a horse at $1.50, would earn 4 average possible points which is impossible as the most you can earn in any one race is 3.

So to rectify, I came up with an arbitrary figure of 0.66.
ie: Formula now is (6 points * horse %) * 66%

A horse at $2.00 would now have an average earning figure of 2 which seems a bit more reasonable.

However am a bit miffed that I had to come up with an arbitrary figure of 66% instead of making the figure more mathmatically certain.

Any ideas, anyone.

The horse paying $2 does not equate to getting 50% of the available points - the most they can get is 3 for the win. SO if a horse is paying $2, then it's a 50% chance to get the 3 votes - so that's 1.5.

You'd then have to work out what % he is of getting second and third as well.

 

[Daytripper]

The horse paying $2 does not equate to getting 50% of the available points - the most they can get is 3 for the win. SO if a horse is paying $2, then it's a 50% chance to get the 3 votes - so that's 1.5.

You'd then have to work out what % he is of getting second and third as well.

Thats actually a good way of looking at it. I'll have to put my mind to calculating the % chances of finishing 2nd and 3rd respectively unless someone beats me to it.

Would you then do it this way ?
=AVGE (points1st+points2nd+points3rd)

 

[stressless]

Here's another one.

Just trying to work on a formula to calculate how many points a jockey should earn in a 1-2-3 jockey challenge event depending on the odds given.

This is what I've come up with.
There are 6 possible points to be earned from any one race (3 First + 2 Second + 1 Third).
My formula is to get the percentage of the horse quoted/rated at and multiply that into the 6 possible points. So if a horse was $4.00 (25%) the likely average jockey points earning would be 1.5 (6 * 25%). This is however flawed as a horse at $2.00 (50%) would get a 3 average possible jockey points - ie it assumes that the horse would win every time. To magnify it even further, a horse at $1.50, would earn 4 average possible points which is impossible as the most you can earn in any one race is 3.

So to rectify, I came up with an arbitrary figure of 0.66.
ie: Formula now is (6 points * horse %) * 66%

A horse at $2.00 would now have an average earning figure of 2 which seems a bit more reasonable.

However am a bit miffed that I had to come up with an arbitrary figure of 66% instead of making the figure more mathmatically certain.

Any ideas, anyone.

FOOK my dog ,never thought having a punt could be so complicated. Danger mouse you gota long road ahead of ya

 

[hess]

Most ratings will have trouble dealing with few starts in the category you're assessing, the best way to do what you're wanting is to use simple averages and stdev but when there's less then 10 data points these can be flawed.

The way I'd go about the problem is to use a points system, I'd go with 5 points for 1st, 3 for second and 1 for third but thats a personal preference as I like my ratings to be more biased towards horses with a good win strike rate. Then just average your totals and if you're particularly keen divide it by the stdev. Of course you'll need to scale your results to an appropriate size for your overall ratings.

eg.
Horse A: 10 starts (5-2-2)
- total is 33 from 10 starts an average of 3.3 per run. stdev of 2 so 1.65 if you divide by stdev.
Horse B: 20 starts (2-0-7)
- total is 17 average of 0.85 at a stdev of 1.5 so 0.56.
Horse C: 10 starts (0-0-1)
- average of 0.1 stdev of 0.32. 0.31 if dividing.

Dividing by stdev is mainly used to differentiate between results with similar means so you need to be careful to interpret the data appropriatley. Obviously the difference between horse B and C is more than using the normalised figures would suggest.

Any approach will have trouble dealing with horses with fewer than 5 runs and trying to rationalise the form of a horse with only 1 start at a particular category is inherently flawed if you're going to use a statistical approach.

 

[hess]

To discount ratings for having less runs I'd multiply by a factor that reflects your confidence in the rating (on a scale between 0 and 1). So when a horse has had just 1 run you have very low confidence and would use a smaller factor.

The method used to determine that factor is a matter of personal preference, you could use an arbitary table of the factor for how many data points there is and then if you use excel use the lookup function.
Or you could try and come up with a formula that determines the factor.

 

[shotgun_blues]

Just my 2 cents on this subject.. take it or leave it.

My top two raters Saturday Metro return -6% profit on local tote.
Huge results considering tab takeout.

I have never looked into track condition in my life...
reason...
You just can't factor results on a horses form over 1,2 or 3 starts
of its career on a certain track condition when so many variables come into play
over those 1,2 or 3 starts or whatever it may be.

In fact, the wet trackers that have run well in the past on wet tracks will
just continue to be overbet by the general public looking for an easy collect on their next bet.

All my results are consistent across the board regardless of track condition so when I last
updated my ratings formulas I decided against factoring results in certain conditions.

You'll have much more success concentrating at overall runs at that track, fitness, win and places %ages,
API's, weight changes, & Beaten Lengths, and class (basic prizemoney algs).

We ain't looking for winners we are looking for value and profits over the
long journey. This you can't rate...after you've done your ratings best thing is to do is record
what else is highly rated round the traps. These ones you won't make a buck out of over the long journey,
or if you are lucky enough to it will be marginal.

 

[hess]

Agree completely with what you're saying Shotgun_blues.
Difficulty with rating for wet tracks is that there's not enough data to establish the horse's ability to handle the condition - a horse could be 2 starts for no placings but only finished 0.2 lengths away once and got held up in the other run... you'd want at least 10 runs before being confident you've accounted for variance.

I don't factor in wet track form, I rate the race as if it were a good track and then cross-check my top raters against their history on wet tracks and am just looking to see if one is substantially better or any are completely useless.