Racetrack computer generated morning lines

[Quesmark]

At Laurel Park on the video feed crawl it says :

"Todays morning line were computer generated."

How many other tracks are doing this now; & when this becomes common will it make the ML a reliable reference point since the inputs will be consistent all the time.Human generated #'s are subject to many more variables,and are not uniform even @ a single track.How are changes in the field/conditions accounted for though?
Can a computer morning line be used as a solid jumping off market opener,and the betting action gauged from there on.

[Dave Schwartz]

It is all that I use.

I've not looked at the tote board to make a wagering decision in just about 3 years.

Regards,
Dave Schwartz

[Franco Santiago]

It is all that I use.

I've not looked at the tote board to make a wagering decision in just about 3 years.

Regards,
Dave Schwartz

Purely out of curiosity, have you compared your computer generated prediction of the odds to the morning line to see how much more accurate the computer model is than the track morning line? In other words, is it better at predicting the favorite?

[green80]

It is all that I use.

I've not looked at the tote board to make a wagering decision in just about 3 years.

Regards,
Dave Schwartz

How do you tell if your bet has any value? Do you have a minimum odds you will take on your bets?

[Tape Reader]

It may turn out to be “garbage in, garbage out.” I would prefer to see syndicated ML makers, with names that one can trust. I think there are a few here that would qualify.

Separately, with the IRS tax change on exotics, I am really looking forward to Saratoga this year. I was really disappointed when mabred informed me that Eric Donavan no longer does the ML at Saratoga. Eric was really good!

[AltonKelsey]

It is all that I use.

I've not looked at the tote board to make a wagering decision in just about 3 years.

Regards,
Dave Schwartz

So you're saying the opinion of one person contains more information than the collective $$$ vote of everyone?

[AltonKelsey]

Apparently nothing new

http://www.drf.com/news/bergman-automated-morning-line-has-limitations

[Quesmark]

Apparently nothing new

http://www.drf.com/news/bergman-automated-morning-line-has-limitations

From the article:
Any morning line can lead some people in a direction. If we are truly in search of something that works perfectly and has no bias we should totally eliminate the morning line. Only theoretical, because even if there's no racetrack issued morning line there would still be one available,or made elsewhere.

[AltonKelsey]

Years ago , there was the National Armstrong Scratch Sheet (great name, back when men were men and you could tell the diff tween them and women)

That's where folks got the morning line, adjusted for scratches and track condition changes.

I'm skeptical it can be done WELL by computer, but considering some of the ML's we've seen in NY over the season, probably not a lot worse.

It's a very difficult job to get right, even IF its done by a pro.

[thaskalos]

It is all that I use.

I've not looked at the tote board to make a wagering decision in just about 3 years.

Regards,
Dave Schwartz

Does this include the exotics?

[Quesmark]

Getting back to the Laurel Park example,is this program created in house,or is there a vendor which sells software to the entire racing industry.
Any algorithm needs to account for a wide range of conditions,like 2yo MSW with many FTS,an off the turf race,or a race with lots of shippers,what kind of inputs would be useful to generate a ML for a full card,is a large data base used as a starting point,and would the ability to make modifications/adjustments be built in to change the process if necessary.

[Dave Schwartz]

Does this include the exotics?

Yes.

But I only play exactas and not in every race.

In tri's and supers there is no tote anyway, but not my cup of tea. (Working on it, though.)

[Dave Schwartz]

So you're saying the opinion of one person contains more information than the collective $$$ vote of everyone?

Absolutely not.

I am saying that using a handful of key factors (one being the ML itself) I can do a better job of predicting where the tote would be than using just the ML.

Further, I don't use the tote because it is like hitting a moving target. Make a betting decision at even 1 minute to post based upon "current odds" is pretty much hopeless.

The odds on use are rigid. I can make my value-based selections at 8 minutes to post. Barring a late scratch (which would likely cause me to pass the race), I can upload my bets and just wait to pull the trigger.


More on the "key factors."

If one follows BRIS Prime Power, you can easily see how predictive it is of the final odds.

For HDW users there is PSR and Pwr (an older version).

For Beyer number users, the average of last 3 Beyers also correlates highly.

Put these together, and suddenly you have an odds prediction system that is pretty powerful.


Regards,
Dave Schwartz

[dnlgfnk]

Absolutely not.

I am saying that using a handful of key factors (one being the ML itself) I can do a better job of predicting where the tote would be than using just the ML.

Further, I don't use the tote because it is like hitting a moving target. Make a betting decision at even 1 minute to post based upon "current odds" is pretty much hopeless.

The odds on use are rigid. I can make my value-based selections at 8 minutes to post. Barring a late scratch (which would likely cause me to pass the race), I can upload my bets and just wait to pull the trigger.


More on the "key factors."

If one follows BRIS Prime Power, you can easily see how predictive it is of the final odds.

For HDW users there is PSR and Pwr (an older version).

For Beyer number users, the average of last 3 Beyers also correlates highly.

Put these together, and suddenly you have an odds prediction system that is pretty powerful.


Regards,
Dave Schwartz

Totally agree, Dave.

I have found the public to be the personification of a typical conventional speed handicapper-- impressed by consistent figs while excusing a lower fig for good reason. Checking the resulting rankings of the field after speed handicapping with BRIS Prime Power for any minor adjustments in those rankings, will produce a very reliable line.

I have no problem with NY, but the Parx ML can be occasionally bewildering.

[Quesmark]

Totally agree, Dave.

I have found the public to be the personification of a typical conventional speed handicapper-- impressed by consistent figs while excusing a lower fig for good reason. Checking the resulting rankings of the field after speed handicapping with BRIS Prime Power for any minor adjustments in those rankings, will produce a very reliable line.

I have no problem with NY, but the Parx ML can be occasionally bewildering.
If the ML correlates with certain #'s than could capping differently generate better returns...

[Dave Schwartz]

If the ML correlates with certain #'s than could capping differently generate better returns...

In my opinion, these are two distinct issues:

1. Predicting the odds.

2. Predicting winners (i.e. probabilities).


I use #2 to bet into #1.

Before I do that, I pick contenders and severely penalize the non-contenders. My record with those non-contenders is excellent, which is why I can penalize them far more than they truly deserve.

In that way I am merely deciding which of my contenders are the logical best bets. (I never bet a non-contender.)

BTW, that "use #2 to bet into #1" thing does not mean I generate accurate probabilities. They are meant to be "relative." That is, of my contenders, which are likely to be the better bets?



Regards,
Dave Schwartz

[FakeNameChanged]

In my opinion, these are two distinct issues:

1. Predicting the odds.

2. Predicting winners (i.e. probabilities).


I use #2 to bet into #1.

Before I do that, I pick contenders and severely penalize the non-contenders. My record with those non-contenders is excellent, which is why I can penalize them far more than they truly deserve.

In that way I am merely deciding which of my contenders are the logical best bets. (I never bet a non-contender.)

BTW, that "use #2 to bet into #1" thing does not mean I generate accurate probabilities. They are meant to be "relative." That is, of my contenders, which are likely to be the better bets?



Regards,
Dave Schwartz

Dave, Without giving away the key to the mint, I'm interested in what you consider a non contender. For example, if your program says a horse has only a 15-1 or 20-1 probability, do you throw the non-contender tag onto them? Or is it a combination of odds probability and apparent poor form? Thx.

[Dave Schwartz]

Dave, Without giving away the key to the mint, I'm interested in what you consider a non contender. For example, if your program says a horse has only a 15-1 or 20-1 probability, do you throw the non-contender tag onto them? Or is it a combination of odds probability and apparent poor form? Thx.

First, it isn't just "my program." It is an approach I have developed.

Does it work better in my software than doing it by hand? Sure. It's just easier because we have so many factors to work with.

When I say a horse is a "non-contender," specifically, I am looking for an under-performing horse. That is, one that figures to be way overbet relative to his chance of winning.

Even more specifically, I am concentrating my efforts on low-odds horses. My measuring stick is "how well do non-contenders who ultimately go off at 3/1 and below perform?"

For the past 3 years those horses have returned approximately $0.89 per $2 wager for me. In other words, they lose 55% for there backers.

For most players a realistic target is a $1.20 $net. That is, a 40% loss.

So, to address your original question, I might have a N/C that goes off at 3/1 and loses 40%, leaving him with a projected hit rate of 15% --- ($120 / $8.00 = 15%) This is the perfect example of a bet-against horse.

Another example: A 3/5 horse at -40% has a projected hit rate of 37.5% --- ($120 / $3.20 = 37.5%) Another bet against horse.

These "bet against horses" are what fuels positive results.


Dave

[Franco Santiago]

Even more specifically, I am concentrating my efforts on low-odds horses. My measuring stick is "how well do non-contenders who ultimately go off at 3/1 and below perform?"

For the past 3 years those horses have returned approximately $0.89 per $2 wager for me. In other words, they lose 55% for there backers.


Purely out of curiosity, do you analyze your performance of "low odds" horses as a group, or do you break 'em down even further? The reason I ask is that I can see you getting a return of .89 per $2 bet on horses in the 2-1 to 3-1 range, but getting that type of return on 4/5s and less seems quite difficult. If you are doing it, congrats! But I'd have to see it to believe it (and no, I don't care to :-) (saying this in case you were gonna invite me...I am comfortable with my own methods)).

[mountainman]

With strong factions making move and countermove now in the win pool at Mnr, it's extremely difficult to do an accurate morning line.

In this era of the volatile tote, I more pride myself on watching where the early money lands and then accurately predicting tote movement and closing odds. It's a more doable than crafting a good morning line, but does require guts and a feel for live action.

[Dave Schwartz]

Purely out of curiosity, do you analyze your performance of "low odds" horses as a group, or do you break 'em down even further? The reason I ask is that I can see you getting a return of .89 per $2 bet on horses in the 2-1 to 3-1 range, but getting that type of return on 4/5s and less seems quite difficult. If you are doing it, congrats! But I'd have to see it to believe it (and no, I don't care to :-) (saying this in case you were gonna invite me...I am comfortable with my own methods)).

While it is more difficult to find a 3/5 or 4/5 horse to toss, the tosses are actually better (at losing money).