New Morning Lines

[Ray2000]

A recent article (May 7) by Bob Pandolfo at DRF on the need for a better Morning Line has drawn several "agree with" comments.

http://www.drf.com/news/pandolfo-search-better-morning-line

(sorry about the link, the search string is the only way I know to post it)


I also agree and would hope all tracks adopt the new TrackMaster Lines for the ML.
(Even tho it'll create a lot of work for me to adapt my robot :) )


I'm not sure how the TM Line beat the track ML in 2011.
It can be measured several ways like
Correlation to actual final odds?, ...Strike rate? ...ROI?
maybe Bob can answer?

In any case I would like to know if and when the individual tracks change over,
maybe they'll tell us. :rolleyes:

Here's what I have for 2013 to date for Morning Line Pick 1 performance.


track races strike roi
DD 2354 40% 0.848
HAR 1202 39% 0.846
STGA 2430 38% 0.886
MR 2341 38% 0.811
RIDC 1106 37% 0.839
FRD 806 37% 0.887
FHLD 1661 36% 0.796
PHL 2123 36% 0.809
PCD 2176 35% 0.801
BTVA 788 35% 0.893
NP 493 35% 0.911
BR 1802 35% 0.847
NFLD 3612 35% 0.793
MAY 1275 35% 0.806
GRVR 658 34% 0.890
HP 554 34% 0.911
CALX 851 34% 0.822
M 1216 31% 0.870
BMLP 1154 33% 0.804
ACES 252 33% 0.932
HOP 1460 32% 0.854
SCD 870 32% 0.798
YR 3647 32% 0.770
MEA 3281 32% 0.770
PPK 1401 32% 0.752
FLMD 1692 31% 0.718
PRC 569 31% 0.785
GEOD 184 30% 0.799
TGDN 599 30% 0.796
MOH 687 29% 0.747
VD 874 29% 0.731
WDB 1662 27% 0.728

[RaceTrackDaddy]

Ray
By strike rate I am assuming that it is the win percentage of the lowest morning line odds and the ROI is that one's payouts with respect to the total number of races run.

Most track odds makers (for a loss of a better term) would tell us that their job is not to predict winners; rather to predict the final odds of the horse.

To me that is something that cannot be accomplished as there are too many variables. A mistake of a large wager on the track or race number can change those final odds.

I think all odds makers should use this study as a guide. They should try to pick out the strongest horses and assigned them accordingly. To try to predict human performances is tougher that to predict animal performances when driven or ridden by men.

As always, Ray, you have the database that can be relied upon. It is great being part of your circle.

rtd

[LottaKash]

Thx for posting that list Ray....

And, an interesting article Pandy...

Funny thing tho, I had no idea that TM had anything like this going for it...

Me personally, I don't put too much stock in the ML's....But, on the other hand I know who's ML's are better than who's, so I guess I see it or use it more than I would care to admit...I am not sure which it is tho..haha..

One thing that I do know is, except for Pocono and Chester the tracks where I play and win the most "kash" at, are in the bottom 15 of Ray's-List...And, I believe, even tho Pocono and Chester are in the top tier, I think I do well there is because they are both tracks that I spend a lot of time doing the homework at....

So, at least on the surface of things, the tracks where I do best at are the tracks with the most unpredicatable ML's....Now, why would I want them to go and change anything ?...:D

[Ray2000]

Yes Bob, these numbers are for playing $2 Win bets on the Morning Line choice in each race, Jan 1, 2013 to May 26, 2014. I did a quick pull on the data so some races are excluded... races with 6 or less horses, Maiden races, races with coupled entries and races ending in dead heats.


The discussion on whether the morning line should reflect some expert's opinion of the probability of each horse winning or should it be an estimate of how the crowd will set the odds, has been debated ad nauseam among many sources. One can probably find some here at PACEAD by searching on Morning Line.


LK

As much as I complain about dumb Morning Lines, it still has the largest impact value on my bot's selections.

If I pass on all races where the robot's top pick is ranked 1,2 or 3 in the ML ...
then the ROI swings from negative to positive. (Or at least used to.. :D )

[mrroyboy]

Are the morning line odds in the Trackmaster PP's the trackmaster morning line or the track's morning line?

[David Siegel]

As always, I am happy to post to PA. Here is the definitive word on Morning Lines (MLs). Since the goal of the ML is to predict the off-odds, the way to measure accuracy is to measure how far a ML is from the off-odds (the difference). Here is how we did this. Then read on.

The average difference could be one meaningful measure, but if the ML
lines were balanced to start with, the average difference would be close to zero, assuming no scratches, so measuring the average difference would not be completely telling.

What would be telling, would be measuring the spread of the differences - but the differences of what? Is a line that has a horse at 5/1 with off odds of 10/1 more or less accurate than one that has a line of 1/2 that goes off at 7/5? TrackMaster established a scale of “levels”. Each level is one “step” up or down the table of industry-accepted standard odds levels as they appear on tote boards around the county. In the above example, 5/1 is level 18 and 10/1 is level 23. So a 5/1 ML horse off at 10/1 would have a difference of 5 levels. 1/2 is level 3 and 7/5 is level 8. So a 1/2 ML horse off at 7/5 would have a difference of 5 levels as well. In these two cases, we would conclude that the ML maker did an equally proficient job in these two cases, each being “wrong” by 5 levels.

This is a very reasonable approach and it has been accepted as such by a number of race secretaries and GMs as a reasonable measurement system.

Once this is in place, you can use a very standard statistical measurement system called STANDARD DEVIATION to measure the "spread" of the TrackMaster to Off Odds differences against the "spread" of the Track-generated to Off Odds differences. We did this (this is what Bob had), and TM had much less "spread" meaning, we were much more accurate.

ROI or Win% has nothing to do with this. Those are great measures for handicappers, or drivers, and the like. But the ML is supposed to, in theory, be one person's best estimate of where the PUBLIC WILL END UP in terms of off-odds.

While the data Bob showed was from 2011, we monitor the tracks that we do the lines for, or those we are talking to now, and nothing has changed. The morning lines we produce day in and day out do a much better job (measured like described above) predicting the off-odds than the humans. The sole exception (for the tracks we are reviewing) is the Meadowlands. They are on-par to ever so slightly better than what the computer has done of late.

[Ray2000]

David

Thanks for the description, looks like a well thought-out approach.

Can you tell us if any tracks are using or will be using Trackmaster morning lines?

[David Siegel]

As of this post, five tracks are currently using our Automated Morning Line (AML) - they are Tioga, Vernon, Monticello, Harrahs Philadelphia and Running Aces. There are some others that I expect to use them upon the opening of their next meet, but until they do, would prefer not to mention them until I know they have been used.

Let me be frank here, I am amazed that more are not using them. Many are still not aware of this great, free service provided by us and the USTA. But for those that are, when I can demonstrate absolutely that our lines are more accurate (closer to predicting the off odds than their current line maker), it is beyond my understanding why they would not use the superior line that also may save them money, save them time, is always balanced, is never influenced by outside factors, etc. etc.

Anyway, those are the (progressive) tracks that are using the line today.

[mrroyboy]

Ok but again the odds with your TM pp's? Those are YOUR morning lines not the tracks, correct???

[Ray2000]

Thanks for the information David,

Guess I'll need to revamp my "bettable" races at some tracks, they were based on my Pick 1 not being in the Morning line top 3. With a more predictable ML, the number of my wagers should decrease but I won't know until a new base is constructed. Fewer bets---improved ROI? Gotta hope :)

Can't see why some tracks wouldn't jump on it, ... should reduce complaints, but old ways die hard.



mrroyboy
I would assume whatever Line appears in the program, (bought at the track or purchased digitally from Track Master) is the racetrack's Morning line as made by their guy or by TM's formula. It's up to the track to inform the customer on how they make the line.

[David Siegel]

Bob has it right. The morning line that appears in the TM program, the printed track program, DRF products or any simulcast program are all the same. It is whatever morning line the track chooses to use. I am not sure why it matters where it came from, as even before the automated ML, no reader could know where it comes from as "Joe" could make it one day at the track and "Frank" another day. That said, even when a track uses the TM-generated line, they can chose to "override" any particular line at their pleasure. But for the most part, the five tracks currently using our line don't typcially override. In fact, I think there have been 7-8 such overrides and all but one resulted in a less accurate line.

As I mentioned in an earlier post, I am still puzzled why tracks continue with their own when a better one exists at no charge. But, that is what makes the world go 'round I guess.

[mrroyboy]

Thank you guys.

[Capper Al]

Raybo,

You should be rich by just playing your 35%+ tracks.

[dannyhill]

Raybo,

You should be rich by just playing your 35%+ tracks.
Raybo has not even posted in this thread. :lol::lol:

[Capper Al]

Raybo or ray2000, sorry for the mix up. Ray2000 then should be the rich man.

[dannyhill]

Raybo or ray2000, sorry for the mix up. Ray2000 then should be the rich man.
He's speaking about the ML and their strike rate. Nothing to do with his personal strike rate.
Although i do believe that Ray2000 on the weekends posts from his yacht.:)

[mrroyboy]

Definitely from his yacht. King Ray.

[Sea Biscuit]

Bob has it right. The morning line that appears in the TM program, the printed track program, DRF products or any simulcast program are all the same. It is whatever morning line the track chooses to use. I am not sure why it matters where it came from, as even before the automated ML, no reader could know where it comes from as "Joe" could make it one day at the track and "Frank" another day. That said, even when a track uses the TM-generated line, they can chose to "override" any particular line at their pleasure. But for the most part, the five tracks currently using our line don't typcially override. In fact, I think there have been 7-8 such overrides and all but one resulted in a less accurate line.

As I mentioned in an earlier post, I am still puzzled why tracks continue with their own when a better one exists at no charge. But, that is what makes the world go 'round I guess.

Mr Siegel: At the risk of sounding stupid, may I know what good is the ML to the average bettor/gambler that predicts how the public will bet a particular horse come post time?

Why not just make the track handicapper's most likely to win picks the morning lines and be done away with it.

Let us decide how good or bad the track handicapper is at doing his job.

[David Siegel]

You happen to raise an excellent point - which really is not about how good our ML is vs the human-created line, but rather is about why a morning line exists in the first place.

First, I fully support any track utilizing a handicappers picks or suggestions. Many tracks do this, as do independent selectors. But this should not be confused with a morning line.

That raises the issue then of the use of a morning line as a predictor of off-odds. Since the public is, in general, a better handicapper en-mass than any one individual (quite often), the ML can be used as a decent proxy for win likelihood when a track handicapper is non-existent.

The other very practical use is a forecast of off-odds. Lets take one simple situation where you "like" horse A and horse B, and you think A has a 25% chance to win, and B has a 20% chance to win. If the best morning line has A at 2/1 and B and 8/1, a wiser bet, all other things equal would be on B. Now, you may argue that one need not use a ML, but rather wager 30 seconds before post. This WOULD be a better route rather than rely on a morning line, but if one could not see odds (e.g. had to bet in advance), then the best ML is better than none. A similar situation exists in a multi-leg wager (say a pick 4), and you are weighing what horse to include in a late leg, where you have no sense of what the public will do. Using the best morning line may help you find a better later-leg overlay to include in your ticket than having no clue about what the public might do.

Finally, while I think the above arguments have merit, the primary reason for a morning line is, like Tevye said, "Tradition". So if a track is going to have one, may as well have the best one possible. And of course, no one has to pay attention to it if they don't want to.

Hope that helps.

Dave

[Stillriledup]

As always, I am happy to post to PA. Here is the definitive word on Morning Lines (MLs). Since the goal of the ML is to predict the off-odds, the way to measure accuracy is to measure how far a ML is from the off-odds (the difference). Here is how we did this. Then read on.

The average difference could be one meaningful measure, but if the ML
lines were balanced to start with, the average difference would be close to zero, assuming no scratches, so measuring the average difference would not be completely telling.

What would be telling, would be measuring the spread of the differences - but the differences of what? Is a line that has a horse at 5/1 with off odds of 10/1 more or less accurate than one that has a line of 1/2 that goes off at 7/5? TrackMaster established a scale of “levels”. Each level is one “step” up or down the table of industry-accepted standard odds levels as they appear on tote boards around the county. In the above example, 5/1 is level 18 and 10/1 is level 23. So a 5/1 ML horse off at 10/1 would have a difference of 5 levels. 1/2 is level 3 and 7/5 is level 8. So a 1/2 ML horse off at 7/5 would have a difference of 5 levels as well. In these two cases, we would conclude that the ML maker did an equally proficient job in these two cases, each being “wrong” by 5 levels.

This is a very reasonable approach and it has been accepted as such by a number of race secretaries and GMs as a reasonable measurement system.

Once this is in place, you can use a very standard statistical measurement system called STANDARD DEVIATION to measure the "spread" of the TrackMaster to Off Odds differences against the "spread" of the Track-generated to Off Odds differences. We did this (this is what Bob had), and TM had much less "spread" meaning, we were much more accurate.

ROI or Win% has nothing to do with this. Those are great measures for handicappers, or drivers, and the like. But the ML is supposed to, in theory, be one person's best estimate of where the PUBLIC WILL END UP in terms of off-odds.

While the data Bob showed was from 2011, we monitor the tracks that we do the lines for, or those we are talking to now, and nothing has changed. The morning lines we produce day in and day out do a much better job (measured like described above) predicting the off-odds than the humans. The sole exception (for the tracks we are reviewing) is the Meadowlands. They are on-par to ever so slightly better than what the computer has done of late.

Good post.

I feel the "5 levels" between a 5-1 and 10-1 is much less severe than a 1-2 ML who goes off at 7-5 and here's why.

If there is a carryover at one of these harness tracks and some bettors want to "get in" but don't really know the horses all that well, a great ML is really vital to whether or not a player can scrap together a ticket. An example would be your 1-2 ML and off at 7-5....if a player sees 1-2 ML and its a really good and accurate ML, that player can essentially have a "Free square" in the pick 4 or pick 5 carryover and that might get a player to wager when he or she would otherwise not.

Sometimes M1 has a carryover, but the ML guy is a bozo, and i can't trust that a 3-1 ML horse won't go off at 3-5 and beat me.....imagine going 6 deep in a M1 pick 5 carryover and someone ends up 3-5 and jogs, you feel stupid and you blame the ML maker for not knowing that the horse was a "Stick" and making a dumb ML.

So, personally, when i see a C/O in harness racing, i just skip it because i can' t trust these morning lines to be accurate, i can't take a chance that what appears to be a wide open race on the ML will turn into a laugher with some horse paying 2.80. Bad ML costs them handle.

[Ray2000]

I do believe that Ray2000 on the weekends posts from his yacht.:)
:D Yes, but the yacht basin is a beaver pond in this Clarion corn field . :lol:




Sea

For me the Morning Line Rank (I really don't care about the actual value) is the single most important impact value of all the factors I use.

Since Jan 2012 the bot has come up with a selection in 30955 races,

of that group 23,350 were either the Moline's 1st or 2nd choice
betting those gives a ROI% of -7.5%

but when bot's pick is 3rd ranked (or higher lower?) in the Moline, 7,605 races
betting those gives a ROI% of +3.6%

The graph shows betting those picks sorted on Morning Line Rank.