Making Track Models - Page 2 - Pace and Cap

For The Lead · 05-24-2010, 05:23 PM

Quote:

Originally Posted by RichieP

Mountaineer, Charlestown, Philly, Penn National, Woodbine and Calder run the longest meets.

Thanks for picking me up, Richie.

I should have mentioned the tracks with the longest meets as well as some of those that have the shortest meets.

For The Lead · 05-24-2010, 05:51 PM

Quote:

Originally Posted by tizlinda

I also believe in work. I am not trying to shortcut anything, but am trying to figure out what you guys do. There is always, however, a point of diminishing returns. Has anyone found that to be true when creating track models? How many races do you need in your model to feel comfortable using it? Do you really find a big difference between 10k claimers going 5.5F vs 6F or 6.5 F? Is it really possible to get enough races even in a long meet to fill so many different models?

I am interested in which tracks have the longest meets, can you let me know what your favorites are?

I'll answer your last question first. YES!

Now, how many races do you need in a model to feel comfortable?
Let's look at this way. If we just start with one and ask ourselves, is that enough(?), we can begin to answer the question ourselves.
How about "two"? Well, suppose there is an early and a late? I guess two isn't enough.
How about "three"? Well, maybe now you have two lates and an early. Is that conclusive enough yet? Not hardly.
At this point wouldn't you have to look at at least one more? Suppose one more adds another "late" to the model. At this point you have one early and three lates. All of a sudden we are seeing a possible trend. Are you comfortable with just those four races in your model?
You see, all of a sudden it becomes and individual question.
What YOU are comfortable with, someone else may not be comfortable with.

A point worthy of note is this. Not every model will produce something that you can point to and say,"wow, look at this. All I have to do in this model is bet ...." whatever it is the model shows.

Some models are all over the place. So what do you do with those models? Keep adding to them. Perhaps something will show itself that you can take advantage of later.

There will be other models that start out good and stay that way.
I have posted this before. I once had a model at a particular track, distance, sex, surface, class and condition, that showed the #1 or #2 first fraction horse won more than 30 consecutive races.
Had I NOT keep models in this way, those races would have been mixed in with other races and I would have never seen this trend.

If I have not answered you question well enough for you, post any remaining questions and I'll try my best to answer them.

tizlinda · 05-25-2010, 01:24 PM

I guess it is up to the individual to decide. Ideally, I would like to be able to break my models down to the extent you do, but I don't see it as a reality if I'm going to play So Cal tracks. By the time I get sufficient data I'll have 2 weeks left to wager. I certainly have the capability to play other tracks and will have to seriously consider that.

Ted Craven · 05-25-2010, 03:16 PM

Porkchop, here are my thoughts, and I'll try to identify where I believe my practice differs from Sartin's recommendations (at least his modern recommendations, circa latter 1990's early 2000's).

Re terminology, and per what For the Lead has mentioned above - can we establish that a collection of Past Performance lines chosen to analyse horses in a race, and specifically, the line chosen for the eventual Winner of the race - be called generically a Decision Model (or Brohamer model if you like); and that Winners' running data or line culled from Result Charts for the race in question (i.e. how the winner actually ran the race) be called a Track Profile. If we do this, we'll have a common nomenclature.

So, talking about Decision Models...

I believe we need to distinguish between the Analysis process of identifying contenders, potential Winners (and in-the-money candidates), i.e. handicapping, on the one hand, and making a Wager Decision, on the other hand. Obviously we don't skip over and ignore the Morning Line favourite, or if we have real-time odds, the too-low-odds betting favourite: we want to know where it fits on our odds line (e.g. BL/BL) - does it tower over the field, is it vulnerable, does another higher odds horse rank closely to it, etc. The fact of public odds at bet time is unrelated to the matchup of the horses and not the Top-ranked horse's fault that it's held at 9/5 at 2 MTP and we are disinclined to Win-bet it.

In my opinion, Sartin advised hiding a top ranked horse too low in odds to bet as a mental device to get his clients to give themselves permission to look further down the contender list for legitimate bets. I recall innumerable missives where he stated 'horseplayers' were too oriented on top-rank-itis, and the only way to cure that and induce them to consider other horses at potential average longer mutuels was to 'Hide' the top-rank horse from sight so you would then see the remaining Top 3 (most likely, the former top 2,3,4). IMO, that was a Wager Decision-related device to be employed after the primary Analysis process.

If you're seeking to make a Decision Model of factors, visible before the race, (and always from a collection of consistent paceline selection strategies), you should be always including that (possibly) top-ranked Analysis horse in your rank orders (i.e. if it won: its own rank in the various factors; if it lost: its rank effect on the possibly lower ranked eventual winner). In your Decision Model, however you maintain it, you could flag the factor rank 2,3,4 winners where a higher ranked horse in your Analysis set closed at odds below your Win betting threshhold and from a Wager Decision standpoint, your method called to (possibly) pass over that one for Win and consider from among the next 3 (depending on the accumulating impact of that factor's significance for the track, distance, surface). That way, when consulting your Analysis Decision Models, the percent hit rate of your Top 3 for a given factor is boosted from a Wager Decision point of view (i.e. because you know that you'll skip over a LPR rank 1 (if important) or BL/BL rank 1 if the bet-time odds are too low and thus your #2 BL/BL becomes your top ranked bettable horse, etc).

Sartin advised at one point considering keeping parallel Decision Models of before-hides and after-hides to accomplish the same thing. Maybe some people did (or still do) that - but I could not - way too much work! The above is the way I kept factor models - until I stopped keeping them.

My understanding from reading the later Follow Ups (and talking to Doc over the past few years of his life) is that he felt that voluminous model keeping eventually ended up becoming of diminishing value compared to the effort. Rather, I believe he advocated simply keeping a Wager Decision Model comprising BL/BL and VDC rank factors (and perhaps both before and after hiding unbettable favourites) and recording from which BL/BL rank (or tier) winners were coming, with what frequency and at what average mutuel. If a 'hidden' rank #1 winner won, record that event (as a 'Miss'). For the record, I am trying to unify such before and after-hiding record keeping in an automated modeling system, and some ideas of that are shown in some of the RDSS 2.0 screenshots (more to come).

From a Wager Decision point of view, leaving the top ranked unbettable horse in your Analysis set also makes it easier to see the principal player (the favourite) in a race and apply alternates to Win betting: for example Exacta as Win (key favourite over the next few or over a counter-energy horse), Hedge a bet to breakeven if the low odds favourite wins and make money if the next close-on-the-oddsline horse wins, Pass the race, single the Top horse in a horizontal bet (DD, P3, P4...).

I think a recent 'dozen or so' races makes a useful model. If you accumulate more (or much more) and it indicates a strong and continuing impact value for one factor's rank or range (e.g. range of E/L) - à la For the Lead's 30 consecutive races - so much the better! If you have no model for a new season, or a newly reconditioned track (e.g. Monmouth 2010), go with what you have and bet lightly and keep notes from the get-go.

I never saw Sartin demonstrate or heard of him recommending models broken down as finely by class as advocated above. Maybe it works. I believe that all the aggregate factors of the later Methodology can be modeled simply by track/surface and distance grouping (i.e. factors which are compounds of other basic factors: BL/BL, VDC, TPP, CPR, Total Energy).

The ideal Decision Model bookkeeping would clearly by to store them in a database (or spreadsheet) where you could filter them, sort and group them by choosing predefined templates by whatever criterea pleased you or looked to be producing good spot-play results - inlcuding sorting and grouping by class distinctions if it suited you or you had enough data to make it meaningful. Model results co-related to hit rate, presence of 'hides' and mutuel payoffs.

My thoughts...

Ted

For The Lead · 05-25-2010, 04:29 PM

I guess it is up to the individual to decide. I agree. Ideally, I would like to be able to break my models down to the extent you do, but I don't see it as a reality if I'm going to play So Cal tracks. By the time I get sufficient data I'll have 2 weeks left to wager. I agree. This is the reason I play tracks with long meets. I certainly have the capability to play other tracks and will have to seriously consider that.

For The Lead · 05-25-2010, 04:42 PM

Quote:

Originally Posted by Ted Craven

I never saw Sartin demonstrate or heard of him recommending models broken down as finely by class as advocated above. Maybe it works.

Ted

For the record, I keep my models and profiles in this fashion based on a recommendation given to me, first hand, by Tom Brohamer at a Sartin Seminar at Saratoga several years ago. I thought advice coming directly from the guy who invented it was good enough advice for me to follow.
One of the first things that became quite obviously was that it would not be useful with tracks that conduct short meets. Therefore, my advice to anyone who wishes to employ such a modeling/profiling technique, is to search out your favorite tracks(s) that have longer meets.
LIKE WOODBINE!!!

Ted Craven · 05-25-2010, 04:54 PM

Quote:

Originally Posted by For The Lead

LIKE WOODBINE!!!

Point taken! And I hope to get back to you some time in the future about how that works out for me there!

Ted

For The Lead · 05-25-2010, 05:00 PM

Quote:

Originally Posted by Ted Craven

Point taken! And I hope to get back to you some time in the future about how that works out for me there!

Ted

And I hope it works out well.

For The Lead · 05-25-2010, 10:20 PM

Quote:

Originally Posted by Ted Craven

Sartin advised at one point considering keeping parallel Decision Models of before-hides and after-hides to accomplish the same thing.

Ted

I do not see the value in keeping two models, one with "hides" and one without "hides". In fact, other than to "trick" someone into looking at the other contenders in the race, such as "Doc" was attempting to do, ones time would be better spent "evaluating" those "hides" to see if they are vulnerable and can be beat. After all, horses that go off at odds of 5/2 and less win almost 60% of all races, so I feel it is important to try and find those low odds horses that ARE beatable.

Keeping a model with ALL contenders included should be sufficient. After all, what a model is telling you, is the various corollaries of the winner as it pertains to the other contenders in the race, NOT its' "odds" compared to the other contenders. If one wants to keep track of the "odds" (prices), then add the prices of the winners to the model.

Ted Craven · 05-26-2010, 02:04 PM

I would have to agree with this (and said more succinctly than I did above!)

Ted

05-25-2010, 03:16 PM	#14
Ted Craven Grade 1 Join Date: Jun 2005 Location: Nanaimo, British Columbia, Canada Posts: 9,255	Porkchop, here are my thoughts, and I'll try to identify where I believe my practice differs from Sartin's recommendations (at least his modern recommendations, circa latter 1990's early 2000's). Re terminology, and per what For the Lead has mentioned above - can we establish that a collection of Past Performance lines chosen to analyse horses in a race, and specifically, the line chosen for the eventual Winner of the race - be called generically a Decision Model (or Brohamer model if you like); and that Winners' running data or line culled from Result Charts for the race in question (i.e. how the winner actually ran the race) be called a Track Profile. If we do this, we'll have a common nomenclature. So, talking about Decision Models... I believe we need to distinguish between the Analysis process of identifying contenders, potential Winners (and in-the-money candidates), i.e. handicapping, on the one hand, and making a Wager Decision, on the other hand. Obviously we don't skip over and ignore the Morning Line favourite, or if we have real-time odds, the too-low-odds betting favourite: we want to know where it fits on our odds line (e.g. BL/BL) - does it tower over the field, is it vulnerable, does another higher odds horse rank closely to it, etc. The fact of public odds at bet time is unrelated to the matchup of the horses and not the Top-ranked horse's fault that it's held at 9/5 at 2 MTP and we are disinclined to Win-bet it. In my opinion, Sartin advised hiding a top ranked horse too low in odds to bet as a mental device to get his clients to give themselves permission to look further down the contender list for legitimate bets. I recall innumerable missives where he stated 'horseplayers' were too oriented on top-rank-itis, and the only way to cure that and induce them to consider other horses at potential average longer mutuels was to 'Hide' the top-rank horse from sight so you would then see the remaining Top 3 (most likely, the former top 2,3,4). IMO, that was a Wager Decision-related device to be employed after the primary Analysis process. If you're seeking to make a Decision Model of factors, visible before the race, (*and always from a collection of consistent paceline selection strategies), you should be always* including that (possibly) top-ranked Analysis horse in your rank orders (i.e. if it won: its own rank in the various factors; if it lost: its rank effect on the possibly lower ranked eventual winner). In your Decision Model, however you maintain it, you could flag the factor rank 2,3,4 winners where a higher ranked horse in your Analysis set closed at odds below your Win betting threshhold and from a Wager Decision standpoint, your method called to (possibly) pass over that one for Win and consider from among the next 3 (depending on the accumulating impact of that factor's significance for the track, distance, surface). That way, when consulting your Analysis Decision Models, the percent hit rate of your Top 3 for a given factor is boosted from a Wager Decision point of view (i.e. because you know that you'll skip over a LPR rank 1 (if important) or BL/BL rank 1 if the bet-time odds are too low and thus your #2 BL/BL becomes your top ranked bettable horse, etc). Sartin advised at one point considering keeping parallel Decision Models of before-hides and after-hides to accomplish the same thing. Maybe some people did (or still do) that - but I could not - way too much work! The above is the way I kept factor models - until I stopped keeping them. My understanding from reading the later Follow Ups (and talking to Doc over the past few years of his life) is that he felt that voluminous model keeping eventually ended up becoming of diminishing value compared to the effort. Rather, I believe he advocated simply keeping a Wager Decision Model comprising BL/BL and VDC rank factors (and perhaps both before and after hiding unbettable favourites) and recording from which BL/BL rank (or tier) winners were coming, with what frequency and at what average mutuel. If a 'hidden' rank #1 winner won, record that event (as a 'Miss'). For the record, I am trying to unify such before and after-hiding record keeping in an automated modeling system, and some ideas of that are shown in some of the RDSS 2.0 screenshots (more to come). From a Wager Decision point of view, leaving the top ranked unbettable horse in your Analysis set also makes it easier to see the principal player (the favourite) in a race and apply alternates to Win betting: for example Exacta as Win (key favourite over the next few or over a counter-energy horse), Hedge a bet to breakeven if the low odds favourite wins and make money if the next close-on-the-oddsline horse wins, Pass the race, single the Top horse in a horizontal bet (DD, P3, P4...). I think a recent 'dozen or so' races makes a useful model. If you accumulate more (or much more) and it indicates a strong and continuing impact value for one factor's rank or range (e.g. range of E/L) - à la For the Lead's 30 consecutive races - so much the better! If you have no model for a new season, or a newly reconditioned track (e.g. Monmouth 2010), go with what you have and bet lightly and keep notes from the get-go. I never saw Sartin demonstrate or heard of him recommending models broken down as finely by class as advocated above. Maybe it works. I believe that all the aggregate factors of the later Methodology can be modeled simply by track/surface and distance grouping (i.e. factors which are compounds of other basic factors: BL/BL, VDC, TPP, CPR, Total Energy). The ideal Decision Model bookkeeping would clearly by to store them in a database (or spreadsheet) where you could filter them, sort and group them by choosing predefined templates by whatever criterea pleased you or looked to be producing good spot-play results - inlcuding sorting and grouping by class distinctions if it suited you or you had enough data to make it meaningful. Model results co-related to hit rate, presence of 'hides' and mutuel payoffs. My thoughts... Ted __________________ RDSS - Racing Decision Support System™

05-25-2010, 04:29 PM	#15
For The Lead Grade 1 Join Date: Nov 2008 Posts: 1,292	I guess it is up to the individual to decide. I agree. Ideally, I would like to be able to break my models down to the extent you do, but I don't see it as a reality if I'm going to play So Cal tracks. By the time I get sufficient data I'll have 2 weeks left to wager. I agree. This is the reason I play tracks with long meets. I certainly have the capability to play other tracks and will have to seriously consider that. __________________ "It's suppose to be hard. If it was easy, everybody would do it." Jimmy Dugan, A League of Their Own

05-26-2010, 02:04 PM	#20
Ted Craven Grade 1 Join Date: Jun 2005 Location: Nanaimo, British Columbia, Canada Posts: 9,255	I would have to agree with this (and said more succinctly than I did above!) Ted __________________ RDSS - Racing Decision Support System™

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Track Models	alydar_ David	General Discussion	0	02-01-2010 12:21 AM
Does Anyone Stick to a track or circuit	trotman	Previous 'Handicapping Discussion' Forum	0	11-05-2009 10:40 PM
To Model your track or wagers	pktruckdriver	Previous 'Handicapping Discussion' Forum	26	09-03-2009 12:32 PM
Release Notes - Version 0.98.8	Ted Craven	RDSS Info, Reference	3	07-28-2009 08:35 PM
Track Comparisons	Houndog	Classic Sartin Programs - Support, Discussion	31	04-17-2009 11:27 AM

05-25-2010, 01:24 PM	#13
tizlinda AlwNW2X Join Date: Apr 2010 Posts: 20	I guess it is up to the individual to decide. Ideally, I would like to be able to break my models down to the extent you do, but I don't see it as a reality if I'm going to play So Cal tracks. By the time I get sufficient data I'll have 2 weeks left to wager. I certainly have the capability to play other tracks and will have to seriously consider that.