• mikesal57
    61
    With recent Tiers , there might be some advancement into using Turf Races.
    From what I see, HCP & IRA are OK to use as selectors but ESx isn't in my
    opinion. I put up , in Facebook, a comparison of ESx & my own OBJ for Turf
    races 7/10 to 7/15.
    My question is how can we understand the stats that this study has given?
    These are the "Probabilities" of each ....I dont know if any other stat is usable

    Esx

    f0pgqvtnocv8kxxx.jpg



    My OBJ


    3o29wcmy4bnbdfht.jpg



    thxs
  • Dave Schwartz
    46
    My question is how can we understand the stats that this study has given?
    These are the "Probabilities" of each ....I don't know if any other stat is usable
    mikesal57

    [Edited: I actually said "ESx" but it should have been "Turf" because that was what Mike was addressing. Works the same way.)

    I absolutely love your post because it demonstrates the big problem with looking at an object or factor from a purely statistical POV.

    Your results are taken out of context FOR WHAT WE ARE TRYING TO ACCOMPLISH.

    Remember that this object has one purpose: To designate "points" awarded to a low-odds horse (i.e. below 1.21/1) based upon his rank (without ties).

    A good template for the questions we want to answer about factors would look like:

    "Which _________ produce the highest __________ among _____________ using factors that __________?"

    Example
    "Which Factors produce the highest Impact Value among Odds<1.01 using factors that Represent E Horses ?"

    Note: Yes, I know I said "Odds <1.01" and not "Odds <1.21."
    What I really want is "aML<1.21" but that would have over-complicated the template.


    Does this make sense?

    REAL CONTEXT
    The real context is we want to know: which of these factors actually work well in this particular actual purpose, which is to award points to horses that we predict will go off at low odds, suggesting that they DO NOT DESERVE TO BE LOW ODDS.

    So, building "better" IV Tables is not the end result.

    Rest assured that I spent a lot of time in designing that object. Can it be improved? Yes. And maybe your object will do that. Please let me know at your earliest convenience.
  • mikesal57
    61
    So what I need to do is add LOW odds horses to my equation?

    Because my example above is using ALL horses...
  • Dave Schwartz
    46
    No, what you want to do is:

    1. Figure out what pct of pool a 1.21 horse gets. (I did not actually bother to do this, but you should since you're trying to over-turn an object that seems to be working.)
    2. Create an analyst that uses the aML to handicap.
    3. Run that study against a filter.
    4. In that study, create a "pick" file of all the "low odds horses." (They would be the ones with probabilities above what you found in step #1.)
    5. Then run your IV table against the picks.

    Note: I seriously doubt that you can do this all in Excel but am very willing to be taught.
  • Dave Schwartz
    46
    6. Handicap 50-100 races using the handicapping approach you've chosen. (For this to have value to a Tiers Player, logically it would have to be using Tiers.

    Look at the performance of the low-odds horses that are play against vs. play as you integrate it with the other 2 objects. You only need to do look at races where the aML has a horse under 1.21.

    You could build a new filter for ward-testing and capture the picks as you did earlier. Then, handicap just those races.
  • mikesal57
    61
    Thxs Dave , this will take a while till I play around with it...
  • mikesal57
    61
    Dave...Out of the above study , I took out stats for PUB Choices 1 & 2...this is actually what we what to compare.....?
    Again these are vs TURF Races 7/10 thru 7/15

    ESx vs TURFOBJ

    Comments?

    0e9tuwzm83egl0rb.jpg
    w6rk9s0jnw87h95u.jpg
    2x48y7jj1fztpp1s.jpg
  • Dave Schwartz
    46
    How did you compute the probabilities?
    (i.e. what did you handicap with?)
  • mikesal57
    61
    Process:

    1) Filter out TURF RACE 7/10 thru 7/15 ALL TURFS
    2) Create Study.....RAN MY TURFOBJ vs Filtered races...
    3) In that study I picked 1 & 2 Pubch Range
    That 2nd chart above is the results

    That 1st chart is 1 & 2 choices using ESx object

    TURFOBJ
    gp5f7x7xgjlgw4z5.jpg
  • Dave Schwartz
    46
    Not the best plan.

    Start with this:
    1. Figure out what pct of pool a 1.21 horse gets. (I did not actually bother to do this, but you should since you're trying to over-turn an object that seems to be working.)
    2. Create an analyst that uses the aML to handicap.
    3. Run that study against a filter.
    4. In that study, create a "pick" file of all the "low odds horses." (They would be the ones with probabilities above what you found in step #1.)

    Capture those horses. Goes to "list 0."

    Using the PickList Manager, copy List0 to List1.

    mqyen4sxj2lbd20f.jpg

    Then...
    5. Make your object the only handicapping object and run the analyst against the same filter again.
    6. Capture the top ?? horses for probability to picks.
    7. In the PickList Manager, intersect the new List0 with List1 and put it into List2.
    8. Empty List0.
    9. Run your IV Tables against the Picks. (List0)

    NOW you have some meaningful answers.
  • mikesal57
    61
    I like to verify my results based on above approach....

    My process....

    1) Ran an aML object vs all races Jan-Jun 2019
    2) I filtered out horses that I'm studying
    3) Captured to Picks File
    4) I ran captured picks vs my IV Factors ...I did one factor -six times
    5) Created Object
    6)Test Object vs July 2019 races...

    Got this from Study
    r6i93ds261wa3qt7.jpg

    So this is telling me that I am profitable if horses are going off 7/2 or less?

    Thxs
  • Dave Schwartz
    46
    It would appear so.
  • Gino
    32
    your tenacity meets the gimlet eye of juan q publico...whatever these plugs are is fairly obvious....
    29/84 is a lot of losses, too many to expect greatness going forward....the dropoff above 7/2 is like
    stepping in a hole when you are at the ocean and trying to get into deep enough water to take a leak...
    gurgle riptide byebye....
    as your sample size goes up you will need to discern more about the money...7/5 to 7/2 is an odds
    range the size of texas...you are on to something for sure, but to sustain profits you are going to
    have to figure out way more about both your winners and your losers...
    think of it this way...you are looking at this horse and making your educated guess as to how much
    loot this horse is gonna take...ok, that is a doable deal right now, but as more and more losing bets start to nibble at your butt, you are going to be forced to figure out what you know that juan q doesn't...so right now, eyeball scan the bejesus out of the 84 picks, do some fine tuning, rerun your
    ever growing sample and focus on the meaning of profitable odds ranges, especially in light of
    today's bet...
    it ain't easy, and it has to be done.
    dos and a half centavos,
    gino
bold
italic
underline
strike
code
quote
ulist
image
url
mention
reveal
youtube
tweet
Add a Comment

Please register to see more

Forum Members always see the latest updates and news first. Sign up today.