Odds Calculator ??

[it2neal]

Does anyone know of a program that will convert a series of numerical values such as Bris Prime Power figures into odds ?

[csmith]

CJ has one on his site

pacefigures.com


click on tools and there it is

[it2neal]

Thanks for your reply. Either this is not what I need or I don't know how to make it work. I just want to take a series of Bris PP numbers, i.e. 121.8, 120.7 119.9, 117.5, 115.8, and convert them to relative odds. Don't want to assign length values or any other variables.

[cj]

It can do what you want, just don't use the extra stuff.

I used only column C (Figure 1) and set it to 100 for 100%.

I then entered the numbers you listed and it gave me the following:


fair odds bet odds
1.5 2.2
2.6 3.5
3.5 4.7
12.2 15.4
31.3 39.3


If more people took the time to learn how to use this tool with whatever figures they happen to like, they would save themselves A LOT of bad bets.

[it2neal]

Hmmm.. I put those figures in Column C ( Figure 1 ) and set it to 100. Hit Control M and it runs through 1000 sims but gives no result.

[cj]

Here is how I had it setup:

http://www.pacefigures.com/images/sim.jpg

[mrharness]

cj

Any chance you will make that into a stand-alone .exe program? Would be nice for anyone that does not use Excel.

[cj]

I'd be more likely to make it a web based application when I get some free time.

[it2neal]

Thanks. I didn't have a length value in because the numbers have nothing to do with beaten lengths. Also, if I put in an additional value under 100, it doesn't give me odds for that number. I'm simply looking for the odds relationship between the numbers. Still doesn't look like this program will do that.

[mrharness]

I'd be more likely to make it a web based application when I get some free time.

:ThmbUp:

[cj]

Thanks. I didn't have a length value in because the numbers have nothing to do with beaten lengths. Also, if I put in an additional value under 100, it doesn't give me odds for that number. I'm simply looking for the odds relationship between the numbers. Still doesn't look like this program will do that.

Aren't the results I posted what you are looking for? Anyway, I was just trying to help, good luck.

[it2neal]

The numbers I am using are Bris Prime Power figures. Sometimes there are figures under 100. This program doesn't seem to recognize anything under 100. Anyway, thanks for the effort. I really appreciate it.

[cj]

It does, if you play around with it I'm sure you can figure it out and get it to do what you want.

[gallahadion]

Does anyone know of a program that will convert a series of numerical values such as Bris Prime Power figures into odds ?

I've made lines for myself before and without the help of a computer program. Its a real eye opening experience, you actually have to examine yourself to find out what you really believe about horse racing. It can be very intensive, but if I was going to make a line I wouldn't rely on a computer and someone elses generated numbers, I'd use my own brain and own reasoning. Just my .02

[it2neal]

I have been playing around with the program as you suggested. I noticed that if you put in the same value say 10 times, you get ten different odds. Shouldn't the odds be exactly the same ? Do you know the reason for this anomaly ?

[cj]

They won't be exactly the same because it is using a random number generator. It should usually be very, very close.

[it2neal]

On a 10 horse field, the odds range from about 9.9 to 12.9, a 3 point spread. This doesn't seem to be acceptable. There is obviously one exact figure for all 10. A program that produces that exact figure is what I'm looking for. Wouldn't happen to know of one, would you ?

[chickenhead]

There is obviously one exact figure for all 10.

No there isn't. In the first place, any number you get is based on a whole bunch of assumptions. Secondly, when you start to get up into the higher odds, a small change in percentage gives you a bigger gap in the odds. Part of the reason most people who use an odds line set an odds limit of how high a horse can be on their fair value line for a bet.

[chickenhead]

btw..worth pointing out, the difference between 9.9 and 12.9 is 2% so far as win% goes. Getting better than 2% resolution is the least of your worries.

[Jingle]

The Program you are looking for is Figline. It was written by Gordon Pine of Netcapper. You can enter any figs you want ie. speed, class, Prime Power, etc. and it will generate an odds line. You also have the option of requiring an edge ie. 10%, 25%, etc. His email address is info@netcapper.com or Voice Message 310-397-9586 if he didn't change it. I recall it cost me $30.

Good handicapping.

[chickenhead]

The Program you are looking for is Figline. It was written by Gordon Pine of Netcapper. You can enter any figs you want ie. speed, class, Prime Power, etc. and it will generate an odds line. You also have the option of requiring an edge ie. 10%, 25%, etc. His email address is info@netcapper.com or Voice Message 310-397-9586 if he didn't change it. I recall it cost me $30.

Good handicapping.

Which is exactly what the program he was provided did, for free.

[formula_2002]

Here is how I had it setup:

http://www.pacefigures.com/images/sim.jpg

do you have any verification data for that ?

[Tom Barrister]

If you want odds based on the BRIS Prime Power numbers, use BRIS's Powerline. You'll have almost exactly the same thing. You can get that for free by signing up with BRISBet, although it only covers tracks that BRISBet handles.

[cj]

do you have any verification data for that ?

Yep, my bank account.

I used to have a much simpler version that is exactly like Figline, and yes, it would give you the same exact odds for the same figures. I'll see if I can find it floating around here and try to post it.

[Dr Win]

this one can help

[Murph]

Would someone please post agian, the formula to convert percentage
to odds? Thanks, in advance for your help. Searches on 'percent to odds'
bring up too many results as does any search includng odds or
oddsline.

Murph

[cj]

Percentage to Odds:

1 / Percentage - 1

So, 20% would be 1 / .2 - 1 = 4

Odds to Percentage:

1 / (Odds + 1)

So, 4 to 1 would be 1 / (5 + 1) = .2 or 20%

[cj]

this one can help

Yep, that is the one I made that I couldn't find. It will do exactly as the original poster requested.

[JoeLong]

Dick Mitchell in his seminal book "Thoroughbred Handicapping as an Investment" explained starting on page 105 how to create a Handicappers Odds Line from a Power Rating. You should get a copy all of the math is included.

[cj]

Dick Mitchell in his seminal book "Thoroughbred Handicapping as an Investment" explained starting on page 105 how to create a Handicappers Odds Line from a Power Rating. You should get a copy all of the math is included.

I'll have to find that one. I'm pretty sure I've read most of his stuff.

The math isn't very difficult. Anyone can customize the following steps to fit their preferred numbers:

Decide on a cutoff for contenders. If the top rating is 100, do you want to assign a line to all horses? Usually, picking a cutoff is best, such as any horse within 5 points.
Decide on a base rating. That is the rating you will give to the lowest rated contender. Using the method I'm explaining, between 2 and 5 points works best.
Determine the amount to subtract from each rating. This is a fixed amount and is calculated by subtracting the lowest contender rating from each contender and adding the base chosen above.
Add each horse's new rating to get a total.
Determine if you will assign a percentage to non-contenders. If you do, divide the total in the step above by the remaining percentage. For example, you decide non-contenders should get 20%, this leaves 80% for the contenders. So, divide the total by .8 (80%) to get the "new" total.
Divide each horse's new rating into the total to get the expected win percentage.
Convert the percentage to odds.
Here is an example of the above, a 6 horse field.

1-100
2-91
3-98
4-92
5-75
6-95

I use a cutoff of 7 points. So, in this example, the high rating is 100, so to be a contender the horse must be 93 or higher, leaving the 1, 3, and 6.

I use a base rating of 2.

I then subtract the lowest rated contender's rating from each remaining contender and add 2. So, the 1 is 100 - 95 +2 = 7. The 3 is 98 - 95 + 2 = 5. The 6 is 95 -95 + 2 = 2.

I add the 7, 5, and 2 to get a total of 14.

I always assign a percentage to non contenders. Instead of a fixed percentage, I just 1/2 of the non contenders divided by the field size. In this case, there are 3 non contenders and 6 horses in the race, which is 50%. I take half of that, or 25%, and assign it to the non contenders.

I then divide the 14 above by the remaining 75% (.75) and now have a new total of 18.67.

Now, I find the expected win percentage of the contenders. Horse 1 is 7 / 18.67 = 37.5%. Horse 3 is 5 / 18.67 = 26.8%. Horse 6 is 2 / 18.67 = 10.7%.

I convert these to odds as described in an earlier post in this thread to get:
1: 1.67 to 1
3: 2.73 to 1
6: 8.35 to 1

Now, I do some additonal stuff, but again, it is up to each person to determine what works best with their ratings. For example, I would not consider the 6 a contender anymore since his chances are less than his "natural" odds of 1 in 6 of winning the race. I would also add in an overlay percentage. I stagger it such that I demand more of any overlay on a second choice than the first, and more on the third than the second, and so on.

Anyway, hope I didn't lull some to sleep at their keyboard. There isn't much to do in my free time where I'm currently located, so I figured why not!

[punteray]

I do not have a spread sheet.:( Would it be possible for you to write down the formulas for me?:) I can then write an .exe program for them.:cool:


Thanks

Ray Scalise punteray@yahoo.com

[equicom]

Punteray.... you could always download OpenOffice.org (guess where you'll find it?)... Free office suite that can handle MS format plus can do more (and comes with a free PDF converter!).

It uses a random number generator

Why would you use random for something like this? I can't see why you would do that! Maybe you're simulating the fickleness of the public? But odds, as opposed to mutuel dividends, are supposed to be static, not variable. The only thing that should make them change is if there is a change in the conditions, eg. a good horse gets scratched or it starts to rain or something like that.

[shanta]

Punteray.... you could always download OpenOffice.org (guess where you'll find it?)... Free office suite that can handle MS format plus can do more (and comes with a free PDF converter!).


Definitely love the Open Office package Ray. Equicom's suggestion is a keeper!
Richie

[punteray]

LOVE THE PEOPLE ON THIS FORUM:jump:


Thanks

Ray Scalise

[Kelso]

download OpenOffice.org (guess where you'll find it?)... Free office suite that can handle MS format plus can do more (and comes with a free PDF converter!).


Definitely love the Open Office package


Equicom/Shanta,
Have either of you used the "Able2Extract" pdf-xls converter? (It has been the best-recommended converter of which I've read ... here at PA and elsewhere ... but it is very difficult to use with tightly-spaced tables.) Can either of you compare the ease-of-use of the OpenOffice converter with A2E?

Thanks very much.

[dutchboy]

I use both programs. I don't see where open office will convert a pdf file. It will save as a pdf but I do not see where it will convert a pdf file.

The able 2 extract is not 100% accurate. It will miss decimal points. I had an example where it converted 1000.01 to 100001 or it will create 2 columns of 10000 and a column of 01. It can only convert certain types of font. 84 can be converted as B4 if the image is not clear. If you plan on converting a pdf scanned as a pdf document you must buy the PDF pro version. If you have a pdf where part of the line is shown at the bottom of the pdf sheet and the balance is shown on the top of the next page the able 2 extract program will create that line twice. The only way to catch it is to manually check every line after you have converted it. Not a fun project. It will not work with the drf pdf since the data is not in tabular columns.




Equicom/Shanta,
Have either of you used the "Able2Extract" pdf-xls converter? (It has been the best-recommended converter of which I've read ... here at PA and elsewhere ... but it is very difficult to use with tightly-spaced tables.) Can either of you compare the ease-of-use of the OpenOffice converter with A2E?

Thanks very much.

[BCOURTNEY]

I standardize all produced values from whatever "system" I use using the following formula. I want a normalized value from a distribution characterized by
mean and standard deviation of all my values. Z = (X - u)/o. MS Excel has a standardize function for this. With the standardized value I then give each horse an equal chance to win based from the size of the field 1 / n, n being the number of entries. I then add this base number (1/n) + ((1/n)*Z), which will give the decimal odds chance for a horse to win. Z will be between -1 and 1. Any number negative after the calculation is a pass at any price, the remaining numbers are decimal odds and easily converted to "fair price" odds for tote board comparsion. This possibily will result in some negative values, however the odds line will always equal 1. This method has proven extremely accurate for myself. Cheers.

[formula_2002]

RAN THAT CALCULATION THROUGH SEVERAL THOUSAND RACES USING FACTOR "X".

Results: acutal wins/ sum of factor "x:" calculation = appox 1. EXCELLENT!!

Now take me to the next step, how do I make money when the take out is 17%!! :)

[formula_2002]

RAN THAT CALCULATION THROUGH SEVERAL THOUSAND RACES USING FACTOR "X".

Results: acutal wins/ sum of factor "x:" calculation = appox 1. EXCELLENT!!

Now take me to the next step, how do I make money when the take out is 17%!! :)

Here is what I think to be an interesting point;
A proportional odds bet using track odds has the same results (roi) as a proportional odds bet using the "x" factor odds.

[cj]

Here is what I think to be an interesting point;
A proportional odds bet using track odds has the same results (roi) as a proportional odds bet using the "x" factor odds.

Here is a point I find interesting. You had access to my figures, and I know for certain there are many plays that beat the game easily, and have done so consistently, for a few years now. Yet, you didn't find them. It makes we wonder what exactly you are looking for when you search databases.

You know the saying, "Seek and ye shall find", but I think "ye are seeking to lose, and thus that is what ye finds."

[formula_2002]

Here is a point I find interesting. You had access to my figures, and I know for certain there are many plays that beat the game easily, and have done so consistently, for a few years now. Yet, you didn't find them. It makes we wonder what exactly you are looking for when you search databases.

You know the saying, "Seek and ye shall find", but I think "ye are seeking to lose, and thus that is what ye finds."

In the begining, your data format* changed too often for me to get my arms around your figures.
I have used Jeff Platts data, AllWays Data and ofcourse Bris data.
Unlike you, I have not found any of the data, as good as it is, useful enough to outperform (in testing thousands of races) the enormous vig of the race tracks.

*I require consistant comma delimited files with full field descripion like Jcapper, AllWays and Bris.

stay lucky ;)

by the way, I was being complementry when I said

"A proportional odds bet using track odds has the same results (roi) as a proportional odds bet using the "x" factor odds."

[xtb]

Unlike you, I have not found any of the data, as good as it is, useful enough to outperform (in testing thousands of races) the enormous vig of the race tracks.



Joe, if you try to think outside of your box, I think you will find some success.

[formula_2002]

Joe, if you try to think outside of your box, I think you will find some success.

Oh, OK.. I'll give it a try :faint:

[cj]

In the begining, your data format* changed too often for me to get my arms around your figures.
I have used Jeff Platts data, AllWays Data and ofcourse Bris data.
Unlike you, I have not found any of the data, as good as it is, useful enough to outperform (in testing thousands of races) the enormous vig of the race tracks.

*I require consistant comma delimited files with full field descripion like Jcapper, AllWays and Bris.

stay lucky ;)

by the way, I was being complementry when I said

"A proportional odds bet using track odds has the same results (roi) as a proportional odds bet using the "x" factor odds."

Changed how? I think I added a field or two to the end of the file. Other than that it hasn't changed in a few years, which is since inception.

[formula_2002]

Changed how? I think I added a field or two to the end of the file. Other than that it hasn't changed in a few years, which is since inception.

this one is from dec 2005.. still the same today?


AQU1229,1,8.3,Dirt,Clm 16000,12:30,1,SUAVE PRINCE,8.0,122.15
AQUi,20051203,8.3,79,80,74,59,65,73
AQU,20051116,8,80,81,79,80,80,91
BEL,20051028,8,79,77,78,69,72,79
MED,20051021,8,77,80,71,69,70,81
MTH,20050904,8,80,71,79,56,63,71
SAR,20050803,9,84,78,84,74,79,88
BEL,20050612,7,57,66,47,59,56,75
GP,20050321,8,77,63,77,63,71*,81
GP,20050221,7,77,67,73,67,69*,86
GP,20050130,8,84,76,80,70,75,88



I think this one is yours also from a few months earlier.
Not really an issue. I'm not casting any fault with the development of the process.

1,1,VICTORIA FLY,E,HOL,20050630,1650,D,79,62,76,55,49,53.65,61
1,1,VICTORIA FLY,E,HOL,20050527,1870,D,71,47,69,43,32,53.94,54
1,1,VICTORIA FLY,E,HOL,20050504,1870,D,-99,-99,-99,-99,-99,-99.99,-99
1,1,VICTORIA FLY,E,SA,20050415,1430,T,66,77,56,58,60,50.78,60
1,1,VICTORIA FLY,E,SA,20050316,1430,T,89,77,66,57,53,51.52,61
1,2,POWERFUL SISTER,S,HOL,20050630,1650,D,79,62,79,61,56,53.48,67
1,2,POWERFUL SISTER,S,HOL,20050619,1540,D,71,55,56,51,49,52.95,52
1,2,POWERFUL SISTER,S,HOL,20050603,1210,D,76,77,57,60,62,52.18,60
1,2,POWERFUL SISTER,S,SA,20050402,1430,D,72,73,46,65,73,52.07,61
1,2,POWERFUL SISTER,S,SA,20050123,1320,D,70,68,50,68,77,51.83,65
1,2,POWERFUL SISTER,S,HOL,20041119,1320,D,72,68,56,59,60,52.59,59
1,2,POWERFUL SISTER,S,DMR,20040827,1320,D,73,80,59,74,82,51.72,72
1,2,POWERFUL SISTER,S,DMR,20040729,1320,D,88,86,58,71,77,51.91,71

[Kelso]

I standardize all produced values from whatever "system" I use using the following formula. I want a normalized value from a distribution characterized by
mean and standard deviation of all my values. Z = (X - u)/o. MS Excel has a standardize function for this. With the standardized value I then give each horse an equal chance to win based from the size of the field 1 / n, n being the number of entries. I then add this base number (1/n) + ((1/n)*Z), which will give the decimal odds chance for a horse to win. Z will be between -1 and 1. Any number negative after the calculation is a pass at any price, the remaining numbers are decimal odds and easily converted to "fair price" odds for tote board comparsion. This possibily will result in some negative values, however the odds line will always equal 1. This method has proven extremely accurate for myself. Cheers.

BCOURTNEY,
Statistics-illiterate, here, so would you kindly walk me through a (hopefully?) simple exapmle?

********************

Say my "system" is a horse's best BRIS Speed Rating over his past 4 races.
8 horses are entered.
The spread of best Speed Ratings is: 89 - 99 - 93 - 93 - 83 - 85 - 94 - 74

I see that the Excel "standardize" function asks for inputs of "X" (described as "the number you want to normalize"), "Mean" and "Standard_dev."

How do I develop "X?" Do I somehow use the 8 Speed Ratings to get there?
What do I input for Mean and Standard_dev?

How does the above relate to your "Z = (X-u)/o" formula?

Although I think I understand how to get from the "Z" result, determined above, to an odds line ... would you please carry imy example through to that end?

********************

Thank you for any guidance offer.

[formula_2002]

BCOURTNEY,
Statistics-illiterate, here, so would you kindly walk me through a (hopefully?) simple exapmle?

********************

Say my "system" is a horse's best BRIS Speed Rating over his past 4 races.
8 horses are entered.
The spread of best Speed Ratings is: 89 - 99 - 93 - 93 - 83 - 85 - 94 - 74

I see that the Excel "standardize" function asks for inputs of "X" (described as "the number you want to normalize"), "Mean" and "Standard_dev."

How do I develop "X?" Do I somehow use the 8 Speed Ratings to get there?
What do I input for Mean and Standard_dev?

How does the above relate to your "Z = (X-u)/o" formula?

Although I think I understand how to get from the "Z" result, determined above, to an odds line ... would you please carry imy example through to that end?

********************

Thank you for any guidance offer.

if I may..
hopefully, its correct.

see attached xls file..

[Tom]

The Z score expresses a number in standard deviations. The mean is 0 and you would expect to see about 68% of all you numbers between +/- 1 Z
units.
Z units represent the normal curve, so probabilities using it are know and can be looked up in standar tables. I use it to make my pars - I know 50% of the reading will be above and below the mean, and I know that as a number approached +3 or -3 (three sigma) the less likey it is to occurr. I base my pars on +/- 1 for normal, and over andunder as fast or slow.

Say the mean Beyer for a class is 90 and the st dev is 3, so 87- 93 is the normal range for this class to run Beyer-wise. An 85 is a slow race, and a 94 is a fast race, for the class. A 97 is very fast, an 82 is very slow.

[formula_2002]

The Z score expresses a number in standard deviations. The mean is 0 and you would expect to see about 68% of all you numbers between +/- 1 Z
units.
Z units represent the normal curve, so probabilities using it are know and can be looked up in standar tables. I use it to make my pars - I know 50% of the reading will be above and below the mean, and I know that as a number approached +3 or -3 (three sigma) the less likey it is to occurr. I base my pars on +/- 1 for normal, and over andunder as fast or slow.

Say the mean Beyer for a class is 90 and the st dev is 3, so 87- 93 is the normal range for this class to run Beyer-wise. An 85 is a slow race, and a 94 is a fast race, for the class. A 97 is very fast, an 82 is very slow.
the next best thing to a naked breast is a std of 3!!!
how may of those (either one ) do you see in a day?? ;)

[formula_2002]

the next best thing to a naked breast is a std of 3!!!
how may of those (either one ) do you see in a day?? ;)
got so excited about that 3!.. I ment to say "Z" score not stdev.. :)

[proximity]

i haven't tried cj's program yet, but one "problem" with figline is (say you're using beyer figures) that you can put in numbers that are the same distance apart, but you will get different results based on how large or small those numbers are. example of two four horse fields: 90-88-86-84 will yield a tighter figline than a weak field of mcls who run 30-28-26-24. and this shouldn't really be since the "starting point" for all figures (10k clm=80 or 100 or whatever) is arbitrary.

anyhow, i think a player who does this would be better served to track a couple months of their projections and see what the results turn out to be FOR ALL HORSES. example: top horses with an advantage of 1 win what %? top horses with an advantage of 2 win what %? horses projected 1 point below the top horse win what %? horses who project 2 points below the top horse win what %?

based on these broad averages of your historical results you could then just use excel (or open office calc, whatever) to calculate your oddsline. if the pcts based on these historical trends add up to more or less than 100% for the field, then excel can adjust the line up or down so the percentages for THIS particular field add up to 100..... make sense?

[Kelso]

Formula & Tom,

Thank you very much for the education. Trying to relate what both of you told me to BCOURTNEY's formula, and to his Excel "standardize" function reference, I wind up with a few more questions:

1- Formula's spreadsheet has a column for the "average" - 88.75 - of my 8 ratings. The "stdev" column shows 7.869 for the range. The horse with an 89 rating gets a Z Score of only 0.03, , so this all jives nicely with Tom's point about Z Scores measuring standard deviations. (89 is almost -0- stdev's from 88.75.)

So in Formula's spreadsheet, "average" appears to substitute for "mean." Is this simply another way of arriving at the same result as BC's post, wherein he specifies "mean" and refers to an Excel function that includes "mean" as an input ... or is it a distinctly different approach to building an odds line?

2- BC says the Z Scores will range between -1 and 1. However, Formula's Z scores range from -1.87 to 1.30. Same question as #1 ... does this mean that Formula's approach is similar, but not identical, to BCs?

3- The sum of Formula's Z Scores is -0-. How does this tie with Tom's comment that "the mean is 0?" Is this a confirmation that Formula's and BCs approaches are actually the same?

4- I tried inputting -0- to the Excel "standardize" function. I also refered to the column of 8 ratings for the "X" input, and tried 7.869 for the stdev input. Got back a "#VALUE!" error message. Any idea what I'm doing wrong here?


I'm in way over my head with this stuff, but both of your responses gave me sufficient comfort to at least ask these questions. Thank you, again.

[formula_2002]

Formula & Tom,

Thank you very much for the education. Trying to relate what both of you told me to BCOURTNEY's formula, and to his Excel "standardize" function reference, I wind up with a few more questions:

1- Formula's spreadsheet has a column for the "average" - 88.75 - of my 8 ratings. The "stdev" column shows 7.869 for the range. The horse with an 89 rating gets a Z Score of only 0.03, , so this all jives nicely with Tom's point about Z Scores measuring standard deviations. (89 is almost -0- stdev's from 88.75.)

So in Formula's spreadsheet, "average" appears to substitute for "mean." Is this simply another way of arriving at the same result as BC's post, wherein he specifies "mean" and refers to an Excel function that includes "mean" as an input ... or is it a distinctly different approach to building an odds line?

2- BC says the Z Scores will range between -1 and 1. However, Formula's Z scores range from -1.87 to 1.30. Same question as #1 ... does this mean that Formula's approach is similar, but not identical, to BCs?

3- The sum of Formula's Z Scores is -0-. How does this tie with Tom's comment that "the mean is 0?" Is this a confirmation that Formula's and BCs approaches are actually the same?

4- I tried inputting -0- to the Excel "standardize" function. I also refered to the column of 8 ratings for the "X" input, and tried 7.869 for the stdev input. Got back a "#VALUE!" error message. Any idea what I'm doing wrong here?


I'm in way over my head with this stuff, but both of your responses gave me sufficient comfort to at least ask these questions. Thank you, again.

I have been using Z scores for some time and have since incorporated the "little twist" that BC has shown here.
It seems the "(1/n)+(Z*(1/n)) " has the effect of normalizing the "Z" score.
That would be column "E" in my .xls chart, where the values are in the -1 to 1 range.

Ranges of win probabilities as determined by the publics's odds and the win probabilities determined by what we are saying here(and using my figures) , are very close to each other.
That is to say the acutal winners/ expected winners , for both the public probabilities and the normalized "z" scores are within a few points of each other for my figures.
for example in the normalized "Z" probability range >=.1 and <.2., for 64,000 horses. both public and "Z" probability a/e = .79..(very nice) .

the acutal win % was 11%.. adjusted for track take of say 18% , it would be 13%.

[BCOURTNEY]

Example attached using your numbers.

[formula_2002]

thanks for confirming.

let me share with you what seems to increase an already "profitable" roi.

adjusting the bet size as follows:

unit bet x (calculated odds %/ track odds %)
(unit bet is a constant)
I've tested it on thousands of plays.

a conventional proportion odds bet returned 1.06 roi.
a flat bet roi bet returned a 1.08 roi.
a proportional odds bet (using the calculated odds) returned a 1.08 roi.

the bet size mentioned above, returned a 1.18 roi

[completebill]

Gordon Pine---netcapper.com, an advertiser on this site, used to sell (Very inexpensive) a nifty little stand-alone program that did just this. Try him--probably still has it.

[Dave Schwartz]

Gordon Pine---netcapper.com, an advertiser on this site, used to sell (Very inexpensive) a nifty little stand-alone program that did just this. Try him--probably still has it.


Good little program, that.

[BCOURTNEY]

I think it's very important to note the difference between appropriate bet sizing in relation to determining an odds line. They are not the same thing, an odds line can give you the inputs to your bet size, so after you've done all the work to determine an appropriate line, you calculate the difference between your line, and the public's line, determine your edge, the expected return or value, and additionally you analyze the maximum expected return, you now have the job of determining what percent of your bankroll to invest on horses that offer a positive expectation.

[proximity]

Gordon Pine---netcapper.com, an advertiser on this site, used to sell (Very inexpensive) a nifty little stand-alone program that did just this. Try him--probably still has it.


mr pine's program is called FIGLINE and while it can be a useful tool, there are problems with this.... some of which i detailed above in post #51.

[PaceAdvantage]

Pardon the interruption --

Recent off-topic additions to this thread have been deleted so that normal folks can get back to normal, non-confrontational discussions about the actual topic of this thread.

If certain members can't police themselves, and can't resist name calling and personal attacks, then replies will be deleted, and eventually, user names will be deleted.

Now back to your regularly scheduled thread..

[formula_2002]

I've tried the following on a data base of 100,000 + horses.
1. Determine the Z score for each entry in a race.
2. Find the representive area under the curve for each horse in the race.
3. Using the Z score tables, sum all the areas (it sums to about 35)
4. Divide the area found in item #2 by the sum found in item 3.
5. Normalize the value determined in item 4. The normalized value represents the horse’s probability of winning.

One of the more interesting things I found was,
When the calculated probability is in the range of >=.2 and <.4, the actual win percentage was 35%...
When the calculated probability was > public’s probability, the roi was greater (and the win % less) than when the calculated probability was less than the public’s probability.

This was for one of the several factors I use...

[cj]

i haven't tried cj's program yet, but one "problem" with figline is (say you're using beyer figures) that you can put in numbers that are the same distance apart, but you will get different results based on how large or small those numbers are. example of two four horse fields: 90-88-86-84 will yield a tighter figline than a weak field of mcls who run 30-28-26-24. and this shouldn't really be since the "starting point" for all figures (10k clm=80 or 100 or whatever) is arbitrary.



I've been out of the loop the last few weeks, sorry. My program does not have the problem you mention. It uses the difference, not the actual numbers.

[proximity]

I've been out of the loop the last few weeks, sorry. My program does not have the problem you mention. It uses the difference, not the actual numbers.

i haven't had much time to fool around with it yet, but that looks like a VERY good program you made there.

absolutely LOVE the part where you can just enter "fixed odds" for a horse that you're unsure of figurewise!!:ThmbUp:

[formula_2002]

I've tried the following on a data base of 100,000 + horses.
1. Determine the Z score for each entry in a race.
2. Find the representive area under the curve for each horse in the race.
3. Using the Z score tables, sum all the areas (it sums to about 35)
4. Divide the area found in item #2 by the sum found in item 3.
5. Normalize the value determined in item 4. The normalized value represents the horse’s probability of winning.

One of the more interesting things I found was,
When the calculated probability is in the range of >=.2 and <.4, the actual win percentage was 35%...
When the calculated probability was > public’s probability, the roi was greater (and the win % less) than when the calculated probability was less than the public’s probability.

This was for one of the several factors I use...

Using the top Bris Prime Power figure for the above:

in 12490 races
A proportional odds bet, using track odds,returned a .87 roi.
A structured proportional odds bet using the area under the curve, also returned a .87 roi.
flat bet roi was .87, with 31% winning races


When the calculated probability is two or more times greater than the final track probability:

A proportional odds bet, using track odds,returned a .92 roi.
A structured proportional odds bet using the area under the curve, returned a .96 roi.

flat bet roi was 1.02, with 14% winning races
only 5% of all races qualified for this play.

[formula_2002]

I've tried the following on a data base of 100,000 + horses.
1. Determine the Z score for each entry in a race.
2. Find the representive area under the curve for each horse in the race.
3. Using the Z score tables, sum all the areas (it sums to about 35)
4. Divide the area found in item #2 by the sum found in item 3.
5. Normalize the value determined in item 4. The normalized value represents the horse’s probability of winning.

One of the more interesting things I found was,
When the calculated probability is in the range of >=.2 and <.4, the actual win percentage was 35%...
When the calculated probability was > public’s probability, the roi was greater (and the win % less) than when the calculated probability was less than the public’s probability.

This was for one of the several factors I use...

Since this method seems to turn even the popular "Bris Prime Fig", into a winning proposition, anyone who would like to test the method on any figure they use can just send me the figure and I'll run my "Home made" program and let you know what the program picks and the min odds.

dont send me the results, just the figure # and a race identifier. you keep the results so you can determine the value of the picks. Of course you can let me know how I did.

send and xls,dbf or comma delimited file to me.
NO MAIDEN RACES!!! I'll need several hundred races..

[GameTheory]

So if I understand correctly you're taking the cumulative area under the curve according to the z-score (the so-called "lower probability") and then normalizing that over the race, correct? (Isn't step 5 redundant as it would already be normal after dividing by the sum in step 4?)

So the initial figures:

80 82 86 86 88 95

would become probabilities (approx):

0.041 0.073 0.168 0.168 0.220 0.329

Is that correct, or do I have the process wrong?


Also compare this method:

1] Find the avg gap between values in the race: agap = (maxval - minval) / (fieldsize -1)
2] Subtract the value: (minval - agap) from each original value
3] Normalize

For the above given original values, it will give the probabilities:

0.055 0.091 0.164 0.164 0.200 0.327

[formula_2002]

So if I understand correctly you're taking the cumulative area under the curve according to the z-score (the so-called "lower probability") and then normalizing that over the race, correct? (Isn't step 5 redundant as it would already be normal after dividing by the sum in step 4?)

So the initial figures:

80 82 86 86 88 95

would become probabilities (approx):

0.041 0.073 0.168 0.168 0.220 0.329Is that correct, or do I have the process wrong?


Also compare this method:

1] Find the avg gap between values in the race: agap = (maxval - minval) / (fieldsize -1)
2] Subtract the value: (minval - agap) from each original value
3] Normalize

For the above given original values, it will give the probabilities:

0.055 0.091 0.164 0.164 0.200 0.327

close enough..
note: only the top rank is a play and the calculated odds are a min of 2 times the final track odds.

that's what works for the figures I have been using.

depending on the quality of your figure, you maybe able to cut back on these restrictions.
test it on a few hundred qualified plays and let me know the flat bet roi and the proportional odds bet roi
(actual wins/expected wins, where expected wins = 1/(track odds+1)

[formula_2002]

Since this method seems to turn even the popular "Bris Prime Fig", into a winning proposition, anyone who would like to test the method on any figure they use can just send me the figure and I'll run my "Home made" program and let you know what the program picks and the min odds.

dont send me the results, just the figure # and a race identifier. you keep the results so you can determine the value of the picks. Of course you can let me know how I did.

send and xls,dbf or comma delimited file to me.
NO MAIDEN RACES!!! I'll need several hundred races..

ALSO, BEST TO KEEP THE MINIUM VALUE FOR THE AREA UNDER THE CURVE AT >=.90

[JeremyJet]

Do any of these odds calculators take into consideration the distance of the race? If you're projecting numbers based on speed figures, a one point differential shouldn't have the same value across the distance spectrum.

JeremyJet

[formula_2002]

Do any of these odds calculators take into consideration the distance of the race? If you're projecting numbers based on speed figures, a one point differential shouldn't have the same value across the distance spectrum.

JeremyJet

just run the numbers and see what happens :)

[JeremyJet]

just run the numbers and see what happens :)

I did just that. I did a small 5 horse field based on the Beyer methodology. These are the numbers I used and the odds the program came up with.

1. 90 - 1.9/1
2. 89 - 3.3/1
3. 88 - 4.5/1
4. 87 - 5.7/1
5. 86 - 9.5/1

The difference in odds seem to be too great when you consider only 4 beyer points seperate the entire field.

JeremyJet

[formula_2002]

I did just that. I did a small 5 horse field based on the Beyer methodology. These are the numbers I used and the odds the program came up with.

1. 90 - 1.9/1
2. 89 - 3.3/1
3. 88 - 4.5/1
4. 87 - 5.7/1
5. 86 - 9.5/1

The difference in odds seem to be too great when you consider only 4 beyer points seperate the entire field.

JeremyJet

MY PROGRAM GENERATED DIFFERENT NUMBERS;
1.73-1
2.37-1
4.07-1
9.10-1
23.8-1

YOU MAKE A GOOD POINT.
I'LL THINK ABOUT THAT!

[formula_2002]

I did just that. I did a small 5 horse field based on the Beyer methodology. These are the numbers I used and the odds the program came up with.

1. 90 - 1.9/1
2. 89 - 3.3/1
3. 88 - 4.5/1
4. 87 - 5.7/1
5. 86 - 9.5/1

The difference in odds seem to be too great when you consider only 4 beyer points seperate the entire field.

JeremyJetThis is not the way I ment that you should "...run the numbers".In my example, I used about 100,000 horses.
You should try a similar test.
Perhaps things are not what they now seem to be.
It's possiblt that a larger sample will allow for a better picture of the way things are.

[JeremyJet]

This is not the way I ment that you should "...run the numbers".In my example, I used about 100,000 horses.
You should try a similar test.
Perhaps things are not what they now seem to be.
It's possiblt that a larger sample will allow for a better picture of the way things are.

I wouldn't know how to begin testing 100,000 horses other than taking it race by race. And that would obviously take too long.

If you're able to do it, tell me what the [overall] win percentage is for each odds range. And then find out the win percentage for each odds range when an [overlay] occures.

Also, indicate what speed figures you're useing in the testing process.

I'd love to see the results if you could do it.

Regards.

JeremyJet

[formula_2002]

in 12490 races
A proportional odds bet, using track odds,returned a .87 roi.
A structured proportional odds bet using the area under the curve, also returned a .87 roi.
flat bet roi was .87, with 31% winning races

After much more testing, I find that the roi's for each of the above bets remain just about equal to each other.

here are the results for different system plays

A/E FLAT BET A/MY E
1.20 1.21 1.17
1.07 1.13 1.08

0.95 0.97 0.96
0.97 0.98 0.97

0.76 0.76 0.76
0.75 0.74 0.76

the first two lines are the results of long shot plays (ave odds about 20-1. 888 races)

the second two lines are the results of favorites(ave odds 1.10. 137 races)

the last two lines are BAD favorites (ave odds 1.49. 1088 races)