Factors/betting method

[jpren37]

Hi everyone,

While I've been a lurker on this board for a few years, this is the first time I've gotten up enough nerve to make a post. I'm so impressed by the skill level of some of the posters here. Anyway, I'm hoping to get some advice regarding the following:

I have been collecting data on 6 different factors (all having to do with how many times a horse has raced). I'd like to apply each factor to each horse in the race; my data shows the win% for each factor can be a low of about 5% to a high of about 18.5%.

I'd then like to take the 6 factors for each horse, add them together, compare them to the horse's current odds and see if I have a bet. E.g., Horse A's 6 factors total 100 percentage points when I add them together. I then divide the 100 points by 6 (to account for each factor) and get 16.62 avg/factor which = 's odds of 5-1. My thinking is that I should only bet this horse if his post time odds are 6-1 or higher. (I don't know the IV for each factor nor how to weight them).
Does this approach make any sense? Are my factor percentages too low for them to be effective?

thanks

Jpren

[misscashalot]

Hi everyone,

While I've been a lurker on this board for a few years, this is the first time I've gotten up enough nerve to make a post. I'm so impressed by the skill level of some of the posters here. Anyway, I'm hoping to get some advice regarding the following:

I have been collecting data on 6 different factors (all having to do with how many times a horse has raced). I'd like to apply each factor to each horse in the race; my data shows the win% for each factor can be a low of about 5% to a high of about 18.5%.

I'd then like to take the 6 factors for each horse, add them together, compare them to the horse's current odds and see if I have a bet. E.g., Horse A's 6 factors total 100 percentage points when I add them together. I then divide the 100 points by 6 (to account for each factor) and get 16.62 avg/factor which = 's odds of 5-1. My thinking is that I should only bet this horse if his post time odds are 6-1 or higher. (I don't know the IV for each factor nor how to weight them).
Does this approach make any sense? Are my factor percentages too low for them to be effective?

thanks

Jpren

It seems to me that your choices are 100% quantifiable therefor studying the charts after making your selections using your method is the way to go. What results have you had up to now?

[jpren37]

My results so far are very good; especially since I don't think anyone is using the same or similar factors. My results get even better when I take into account current form and using posted speed/pace figures. So far,however, none of the factors when taken in the aggregate would give me a bet on any horse less then 4 to 1.

[Overlay]

I'd then like to take the 6 factors for each horse, add them together, compare them to the horse's current odds and see if I have a bet. E.g., Horse A's 6 factors total 100 percentage points when I add them together. I then divide the 100 points by 6 (to account for each factor) and get 16.62 avg/factor which = 's odds of 5-1. My thinking is that I should only bet this horse if his post time odds are 6-1 or higher. (I don't know the IV for each factor nor how to weight them).

I'd say that you would want to assure that the factors that you were using were not interdependent in some way, which would affect the validity of combining them.

Also, your idea of treating the factors equally would depend on whether they all had equal influence in determining race outcomes (as you seem to be aware by your comment on impact values and weighting).

Finally, I would think that you would want to combine the factors through multiplication rather than addition (although adding a combination of factors could be valid also, as Quirin (for example) did with the values for the factors in his multiple-regression formulas, as long as the value that you were using for each of the individual factors was properly weighted to reflect the factor's relative influence upon the horse's overall chance of winning).

[misscashalot]

My results so far are very good; especially since I don't think anyone is using the same or similar factors. My results get even better when I take into account current form and using posted speed/pace figures. So far,however, none of the factors when taken in the aggregate would give me a bet on any horse less then 4 to 1.
If you wind up with just 2 of your 6 factors being silver bullets, you're on to something exciting. Good for you. Your note on the 4/1 problem is not unusual for a beginning study. I expect youre using a spread sheet, database like Excel. You have to manipulate your stand(s) of your 6 factors downward so you do recognize the better bet horse. At 4/1 your throwing out Fav and 2nd bet choice. Maybe this is good, maybe not. Add to your db, new cols and add more info that will suppress the original. then you can sort and delete and add to see what your win rate is and what your roi is for every possibility. I wish you luck
Matt

[misscashalot]

...... as long as the value that you were using for each of the individual factors was properly weighted to reflect the factor's relative influence upon the horse's overall chance of winning).
:ThmbUp:This is the difference between winning and winning big.

[jpren37]

Misscashalot/Overlay,
Thanks for responding to my post. Regarding the weighting of factors, would you suggest I take the winning horse and document the contributing position of each of my 6 factors (e.g # of times it was the highest % factor, 2nd highest etc.)

What then would be my next step. If e.g., I found factor #3 to be the highest % most often, what weight would I give that factor. Not sure I'm asking the question correctly but hope you get what I mean....

Jeff

[misscashalot]

.......What then would be my next step. If e.g., I found factor #3 to be the highest % most often, what weight would I give that factor. Not sure I'm asking the question correctly but hope you get what I mean....Jeff
Yours is not an easy question to answer. It's all according to what your db looks like, and what info you have in addition to your basic 6. R U keeping track of post time odds for each horse (winning and losing), how many runners in each race, the distance (1 or 2 turns)? etc. You'll find if all you're betting is over 4/1 then you'll be doing better in races having 8 runners or more. This will effect how you look at your basic 6. My suggestion is to manipulate all or some of your basic 6 so as to include bettable favorites just to see if there's a market there for you. This may help you with knowing how much importance each of the 6 has. Making a db is as creative as it is mechanical.

[gm10]

I am not sure how you convert your factors into odds?

[jonnielu]

My results so far are very good; especially since I don't think anyone is using the same or similar factors. My results get even better when I take into account current form and using posted speed/pace figures. So far,however, none of the factors when taken in the aggregate would give me a bet on any horse less then 4 to 1.

With this, you are dismissing approx. 50% of winners, your net will probably not get around another 10% of longshot winners in the first place. Which is not a problem, as long as your method enables you to accurately pass at least 20% of that 60%.

The inability to do this accurately/confidently can lead to constant second guessing of your weightings, and the ability to do this will call on you to pass, knowing that a 3-1, or 6-5 is 90% to win.

jdl

[jpren37]

I use 6 factors to rate each horse. I know the % chance of each factor winning. I simply add the factor %'s together, divide by 6 and convert that average % into odds.

[jpren37]

Misscashalot/jonnielu

Perhaps in those races where my data tells me a horse is an underlay but I like his chances and he's one of the first three favorites, I can look at betting him in the exotic pools. Using generally accepted rules like: "68% of all races run are won by won of the first three favorites" therefore, I'd better have one of those three in my exacta ticket somewhere (either on top or bottom).

I guess it all boils down to keeping records; my method will certainy keep my from playing too many "short-priced" horses in the win pool.

jeff

[Fwizard]

My problem always has been when to bet a lower priced horse that may still be an overlay...I know when my handicapping picks a 6-1 to bet it to win because my winning pct(22) will get me a profit if I bet only horses with those odds if my win pct stays the same...but what if I come up with a 3-1 that is far ahead in my handicapping system...while I am typing this I know the answer is to look up what pct I hit with this type of horse...

[misscashalot]

My problem always has been when to bet a lower priced horse that may still be an overlay...I know when my handicapping picks a 6-1 to bet it to win because my winning pct(22) will get me a profit if I bet only horses with those odds if my win pct stays the same...but what if I come up with a 3-1 that is far ahead in my handicapping system...while I am typing this I know the answer is to look up what pct I hit with this type of horse...

:ThmbUp: Very interesting. When you know that, you may have the best of both worlds.

[Harvhorse]

Fred Davis did exactly what you are working on about 40 years ago. His book is called Percentages and Probobilities. It is accompinied with a booklet which converts certain factors into percentages, I have a friend who uses Davis and he said it holds up today. He says it is particularly good with maidens.

[Greyfox]

Several years ago the late Dick Mitchell wrote a book entitled Winning Thoroughbred Strategies. It's available on line at http://www.amazon.com/Winning-Thoroughbred-Strategies-Dick-Mitchell/dp/068807913X

In that book he outlined a fairly accurate way to construct an odds table using Pace and Speed factors. I used it for a number of years and it worked well.

However, if your method for making odds is working my advice is simple:

"Whatever works, use it."

[formula_2002]

Hi everyone,

While I've been a lurker on this board for a few years, this is the first time I've gotten up enough nerve to make a post. I'm so impressed by the skill level of some of the posters here. Anyway, I'm hoping to get some advice regarding the following:

I have been collecting data on 6 different factors (all having to do with how many times a horse has raced). I'd like to apply each factor to each horse in the race; my data shows the win% for each factor can be a low of about 5% to a high of about 18.5%.

I'd then like to take the 6 factors for each horse, add them together, compare them to the horse's current odds and see if I have a bet. E.g., Horse A's 6 factors total 100 percentage points when I add them together. I then divide the 100 points by 6 (to account for each factor) and get 16.62 avg/factor which = 's odds of 5-1. My thinking is that I should only bet this horse if his post time odds are 6-1 or higher. (I don't know the IV for each factor nor how to weight them).
Does this approach make any sense? Are my factor percentages too low for them to be effective?

thanks

Jpren




Here is one way I treat a group of numerical factors.
Take a portion of your data base and determine the Z score for each factor (best if done wrt odds).
Determine the roi for each Z score with respect to factor.

Say 4 of your seven factors each produce a minimum .90 roi when the Z score >=2.

Now, using those 4 factors, determine the roi for the rest of the data base when each of those factors have a Z score >=2.

Test the above for small increments of odds range, say >0<.5
>=5.<1, >=1<1.5. use the smallest increments possible.

Just google Z score and you'll find how to do it.

GL

[markgoldie]

Interesting that every response here involves the technical details of converting handicapping factors into a usable formula for betting at certain odds.

As a theorist, though, I'm much more interested in the posters original statement that he has six separate factors that he has found all based on how many times the horse has raced.

Now. We all understand that a horse can be under raced as well as over raced, but it would seem (off the top of one's head) that the reaction to more frequent or less frequent racing would be highly individualistic to the animal. Furthermore, while racing is the heaviest form of physical activity, hard or frequent training can be a close substitute. Therefore it would seem that training patterns might be nearly as important as actual racing if we are to assume that the gentleman has found some sort of "universal" optimal activity pattern for horses.

And so, I'd be highly interested in the theoretical assumptions or underpinnings that would lead one to such an area of inquiry in the first place. This, I find, is generally necessary before spending a lot of time and energy on research into virtually anything. That is, there has to be a prior rationale for undertaking the research in the first place. For example, (from the top of my head), it may be possible that horses race best during a favored particular phase of the moon. But unless I have some reason to believe the effect is logically connected to specific on-track performance, I'm not pursuing this line of inquiry unless I have unlimited research resources.

So I'd like to know how the gentleman decided to launch this area of inquiry.

[Greyfox]

For example, (from the top of my head), it may be possible that horses race best during a favored particular phase of the moon. .

Shh! Don't tell the world my secret.

[jasperson]

The only difference between % winners and an impact variable is instead of dividing the number of winner by the number of races you must total the number of horse that have this same trait. Like won last race. You can have more than one horse in the race that won his last race, so therefore you must count all of them that won their last race. The iv will allways be = or < than the win%
Jack

[markgoldie]

Shh! Don't tell the world my secret.
Even worse than the unauthorized exposure was my failure to properly attribute the brilliant and inspired insight that led you to this original work. Apologies.

[jpren37]

markgoldie,

I guess I just like to tinker a bit. My prior rationale was from a simple hypotheis that the more times a horse ran, the less it's chances to win. So, to test it, I looked at a series of attempts e.g. horses that have run 6x's lifetime vs those that have run 38x's lifetime (I have stats from 1x to 100x's). I found 6 run horses in a sample of 1647 attempts showed a win % of 15.4%. In 1,048 attempts, a horse with 38 runs showed a win% of 10.8%. There seemed to be a correlation here though i'm not saying there is causality.

I then looked at different factors built around this concept to see if there was a decent spread with these factors e.g in $8,000 purse races, the horse that has the most lifetime runs wins 8% (in a sample of 1165 attempts) while the horse that runs 7th most amount of times wins 19.5% of these races.

These are just two of the factors I'm using but I thought if I looked for factors that showed some correlation to a winning performance and the factors were not frequently considered by cappers, I might be able to build an odds line using the percents and get good prices.

I'm still trying to figure out how to weigh factors and would appreciate a simple algorithm that I could use.

That's it. I'd appreciate any comments neg. or pos.

[formula_2002]

I found 6 run horses in a sample of 1647 attempts showed a win % of 15.4%. In 1,048 attempts, a horse with 38 runs showed a win% of 10.8%. There seemed to be a correlation here though i'm not saying there is causality.
in each group, can you determine the flat bet roi and the actual wins/expected wins ratio?
where expected win = 1/(odds+1) adjusted for track take-out.

[jpren37]

Formula 2002,

I can't and i realize now that is something I should have been looking at while gathering the data. I felt that if I established a win % for each factor and used the post time odds to determine if I had a bet, I assumed I wouldn't need to know the roi. I know, faulty logic.

[jasperson]

Interesting that every response here involves the technical details of converting handicapping factors into a usable formula for betting at certain odds.

As a theorist, though, I'm much more interested in the posters original statement that he has six separate factors that he has found all based on how many times the horse has raced.

Now. We all understand that a horse can be under raced as well as over raced, but it would seem (off the top of one's head) that the reaction to more frequent or less frequent racing would be highly individualistic to the animal. Furthermore, while racing is the heaviest form of physical activity, hard or frequent training can be a close substitute. Therefore it would seem that training patterns might be nearly as important as actual racing if we are to assume that the gentleman has found some sort of "universal" optimal activity pattern for horses.

And so, I'd be highly interested in the theoretical assumptions or underpinnings that would lead one to such an area of inquiry in the first place. This, I find, is generally necessary before spending a lot of time and energy on research into virtually anything. That is, there has to be a prior rationale for undertaking the research in the first place. For example, (from the top of my head), it may be possible that horses race best during a favored particular phase of the moon. But unless I have some reason to believe the effect is logically connected to specific on-track performance, I'm not pursuing this line of inquiry unless I have unlimited research resources.

So I'd like to know how the gentleman decided to launch this area of inquiry.
There is some logic to betting a horse with the fewest starts especially at the cheaper tracks. These tracks card a lot c5000 nw1 and nw2 races. When I handicap those races I look closely at the horses that have met the condition with fewest starts. A horse that has had 20 tries at this condition for the most part is not a good bet. I think that is where he get his best results

[TrifectaMike]

markgoldie,

I guess I just like to tinker a bit. My prior rationale was from a simple hypotheis that the more times a horse ran, the less it's chances to win. So, to test it, I looked at a series of attempts e.g. horses that have run 6x's lifetime vs those that have run 38x's lifetime (I have stats from 1x to 100x's). I found 6 run horses in a sample of 1647 attempts showed a win % of 15.4%. In 1,048 attempts, a horse with 38 runs showed a win% of 10.8%. There seemed to be a correlation here though i'm not saying there is causality.

I then looked at different factors built around this concept to see if there was a decent spread with these factors e.g in $8,000 purse races, the horse that has the most lifetime runs wins 8% (in a sample of 1165 attempts) while the horse that runs 7th most amount of times wins 19.5% of these races.

These are just two of the factors I'm using but I thought if I looked for factors that showed some correlation to a winning performance and the factors were not frequently considered by cappers, I might be able to build an odds line using the percents and get good prices.

I'm still trying to figure out how to weigh factors and would appreciate a simple algorithm that I could use.

That's it. I'd appreciate any comments neg. or pos.

If you continue to filter your basic factor to more factors, you will end up with highly correlated factors. I can't even begin to understand what giving weights to these factors means.

I would suggest that you perform a simple binomial logit regression. Let the number of races run be your predictor variable.

If you are unclear on how to do this shoot me a pm, I'll help you.

Mike