Trakus idea

[teddy]

I read an interesting article about it yesterday. Where they did a study on turf races in the daily racing form. Probably most of you have read this it's about grout races on the turf and obviously it shows great rio on horses with a 30 or greater delta. I noticed they were races where all of the horses have a higher Delta then the winner. That points to a better than looked win. love to hear if anyone plays those next out.

[ReplayRandall]

Originally Posted by teddy

I read an interesting article about it yesterday. Where they did a study on turf races in the daily racing form. Probably most of you have read this it's about grout races on the turf and obviously it shows great rio on horses with a 30 or greater delta. I noticed they were races where all of the horses have a higher Delta then the winner. That points to a better than looked win. love to hear if anyone plays those next out.

In 2014, John Shufon came up with a methodology for horses with a delta of 32 or more. I will post this in 2 parts, the second part will answer teddy's question about follow-up plays:

PART 1- Written by John Shufon

"The Trakus T-Charts, which are maintained on each participating track’s website, show time and distance measurements by segment and cumulatively for each point of call. The finish screen of the charts provides the actual distance run by each horse and computes a delta value that signifies the difference in ground covered by each runner relative to the winner. For example, if the second-place horse received a delta of 32, it means the horse ran 32 more feet than the winner, presumably due to a wide trip.

To evaluate the usefulness of Trakus ground-loss data for finding horses to bet back who have a good chance to be successful, turf races at two meets were selected for analysis: the 2013-14 Gulfstream winter meet and the 2014 Saratoga meet. Turf races were selected because of their relatively tight turns, where ground loss is particularly disadvantageous.

The methodology used to find qualifying horses:
1. The Trakus T-Charts on the track’s websites were used.
2. Only two-turn races on turf were analyzed.

To find qualifying horses, the following criteria were used, based on the principle that a length is about eight feet:
a. If the delta value for any horse was between 32 and 40 feet and the horse finished within four lengths of the winner, the horse was selected for follow-up; and
b. If the delta value was 40 feet or greater and the horse finished within five lengths of the winner, the horse was selected for follow-up.

The methodology was designed to find horses who were in contention and might have even won had they not raced wide. It should be noted that only races where horses received a positive delta (wide relative to the winner) were selected for the study. Some leeway (two feet max) also was granted for horses who may have been just under the delta criteria if other traffic trouble warranted follow-up."

[teddy]

yes I read this study too.. I added 30 horses last night and will add more from nyc tracks. I think they are too sharp in NYC for this to do much. I also did GP. Not as sharp down there. Well test it out. I used Some dirt too and horses that were the widest and still won. Fun to track....

[ReplayRandall]

Originally Posted by teddy

yes I read this study too.. I added 30 horses last night and will add more from nyc tracks. I think they are too sharp in NYC for this to do much. I also did GP. Not as sharp down there. Well test it out. I used Some dirt too and horses that were the widest and still won. Fun to track....

Here's part 2, with results from GP and Sar/Bel:

For the Gulfstream analysis, the runback race had to occur within the time frame of that meet (Nov.30, 2013, through April 6, 2014), but the start could occur at any track, distance, or surface. For Saratoga, the runback
race could occur either during that meet or during the subsequent Belmont
fall meet (covering a cumulative time period of July 18 through Oct. 26). As with the Gulfstream analysis, the first race back could occur at any track, distance, or surface.
There were 117 total qualifying runbacks between Gulfstream and Saratoga.
Gulfstream Park results
The results of the study for the Gulfstream turf course can be summarized in one word: impressive.
Of the 66 horses who qualified, 21 horses (32 percent)came back to win at an average parimutuel win payoff of $10.08. The $2 return on investment was a very healthy $3.21.
Twentyone other horses hit the board at average odds of 3.701, and six nonwinning runbacks won their next start at odds of just over 6.101.
The average odds for the 24 horses finishing off the board was 13.601.
This relatively high value suggests that several of these horses may have been eliminated from consideration if traditional handicapping principles were applied, which would further increase the ROI.

Saratoga results-
The results of the Saratoga runbacks, are not as impressive as at Gulfstream but certainly are decent and confirm the benefits of including Trakus groundloss data when handicapping. Of the 51 horses who qualified based on the criteria, 13 horses (25 percent) came back to win at an average win mutuel of $9.83 and a $2 ROI of $2.51. Twenty other horses hit the board at average odds of 4.801, and 10 nonwinning runbacks won their next start at odds of just over 6-1.
The average odds for the 18 horses finishing off the board was 9.301.

Summary
Summarizing the results of qualifying horses for the two meets, and the runbacks collectively won 29 percent of their next starts, with an average win payoff of $9.98. Approximately twothirds of the runbacks hit the board, a percentage that interestingly is consistent across both venues. The cumulative $2 ROI for the 117 races is a respectable $2.90. Achieving a flatbet profit given such diverse circumstances without the benefit of conventional handicapping is certainly noteworthy. In addition, 16 nonwinning runbacks won their next start at average odds of 5.501.
This means 50 horses, or 43 percent of all qualifying horses, won one of
their next two starts.

To the question of whether Trakus data can be used to find winners, the answer is a qualified yes.Complementing that data with traditional handicapping methods should improve on the results achieved.

Shufon is a retired analyst and bureau director from the New York State Department of Transportation. He is an avid recreational horseplayer and an original member of Epona Racing Stable.

[AltonKelsey]

let me know when trackus grows a pair and puts the data up in downloadable form.

its 2016, im not going to sit there and type the data in.

[rsetup]

Originally Posted by AltonKelsey

let me know when trackus grows a pair and puts the data up in downloadable form.

its 2016, im not going to sit there and type the data in.

Do you have any programming skills?

[AltonKelsey]

Considerable. But the last time I looked it was non trivial to grab the data programmatically.

I cooked up something, but I wasn't happy with the amount of manual work involved.

I've asked in the past for ideas but never got a response.

[cj]

Originally Posted by AltonKelsey

Considerable. But the last time I looked it was non trivial to grab the data programmatically.

I cooked up something, but I wasn't happy with the amount of manual work involved.

I've asked in the past for ideas but never got a response.

Also, the problem with programming for the Trakus data as it is currently presented is they can change it up at the drop of that. I've been down that road with Equibase many times. It is much better the last few years since they worked with HANA, but Trakus has no such agreements.

[classhandicapper]

Originally Posted by cj

Also, the problem with programming for the Trakus data as it is currently presented is they can change it up at the drop of that. I've been down that road with Equibase many times. It is much better the last few years since they worked with HANA, but Trakus has no such agreements.

The other problem is that when it malfunctions you sometimes get no data at all.

[classhandicapper]

The results of that study surprised me when I originally read it. To the extent that ground loss matters, I thought most of it was built into the odds because some of the TG and Sheets players are huge bettors and they all have that information built right into their figures.

I thought the real value would be in understanding when ground loss matters a lot and when it doesn't matter much.

I have heard that some TG/Sheets players do especially well on turf, but I can't imagine they do anywhere near as well as this study indicates.

[AltonKelsey]

I could deal with an occasional change of format.

Basically these guys have been around for years now, and they are giving the middle finger to horseplayers who would like to use the data.

I called them about it long ago, and got a non answer.

I don't even know of a PAY method of getting the data in CSV format.

[AltonKelsey]

Originally Posted by classhandicapper

The results of that study surprised me when I originally read it. To the extent that ground loss matters, I thought most of it was built into the odds because some of the TG and Sheets players are huge bettors and they all have that information built right into their figures.

I thought the real value would be in understanding when ground loss matters a lot and when it doesn't matter much.

I have heard that some TG/Sheets players do especially well on turf, but I can't imagine they do anywhere near as well as this study indicates.

Good point. TG and Ragozin have included GL in the number from day 1, which is why they might come up with a 'buried' performance more often then pure speed figs might reveal.

So yes, the sample was probably too small to be reflective of the real roi.

[rsetup]

I suggest 'web scraping with python' Ryan Mitchell. Might be a free pdf of it on the web.

[teddy]

You can hire data entry people dirt cheap in India. I have use them on websites for programming and you're talking about 1/10 of the cost. If you can get the data and format to them they can do it for pennies.

[AltonKelsey]

Originally Posted by teddy

You can hire data entry people dirt cheap in India. I have use them on websites for programming and you're talking about 1/10 of the cost. If you can get the data and format to them they can do it for pennies.

Bottom line is, I shouldn't have to hack python or hire an Indian to get a little data into a spreadsheet.

What are they afraid of?