Database Handicapping Software- JCapper

JCapper Message Board

          General Discussion
                      -- Faves

Home Register
Log In
By Faves
TPW
5/30/2018
5:43:05 AM
Very disheartening when this is what the Top 5 looks like in my Key Factors report. I realize field size is down, but winning faves and the ROI is ridiculous.


code:
     UPRMLPROB           12917   3874    0.2999  2.3145    0.8595    6471      0.501   0.8721  
BETTORSTOTEPROB 12918 4636 0.3589 2.7699 0.8568 7349 0.5689 0.8819
JPRTOTEPROB 12918 4824 0.3734 2.8818 0.8529 7538 0.5835 0.8733
UPR 12916 3717 0.2878 2.2211 0.8528 6301 0.4878 0.8721
POST TIME FAVS 13387 5025 0.3754 2.8972 0.8518 7858 0.587 0.8771



Terry

Reply
jeff
5/31/2018
10:31:14 AM
Here's what the top of my 01-01-2018 through 05-30-2018 Key Factors Report looks like using HDW Data:

****************************************************************************************
Key Factors Rank = 1 sorted by Win ROI Run Date: 5/31/2018 7:35:06 AM
****************************************************************************************
WIN WIN WIN PLACE PLACE
FACTOR PLAYS WINS PCT IMPACT ROI PLACES PCT ROI
****************************************************************************************
CPACE F20 13430 3097 0.2306 1.7564 0.886 5442 0.4052 0.869
JPR 13423 3750 0.2794 2.1281 0.883 6353 0.4733 0.8773
USERFACTOR3 F33 13423 3857 0.2873 2.1883 0.8806 6434 0.4793 0.8677
EARLYCONSENSUS F19 14218 3167 0.2227 1.6963 0.8776 5655 0.3977 0.8757
CLASSCOMPOS F17 13860 3138 0.2264 1.7245 0.8772 5493 0.3963 0.8509
JPRMLPROB 13423 4002 0.2981 2.2706 0.8655 6652 0.4956 0.8669
JRATING F29 13430 3084 0.2296 1.7488 0.8646 5539 0.4124 0.8736
CFA F08 13614 3199 0.235 1.79 0.8632 5658 0.4156 0.8646
POST TIME FAVS 13902 5308 0.3818 2.9081 0.8627 8297 0.5968 0.8904
CLASSCONSENSUS F27 14579 4242 0.291 2.2165 0.8625 7111 0.4878 0.8665
BETTORSTOTEPROB 13448 5123 0.3809 2.9013 0.8597 8019 0.5963 0.8897
FIGCONSENSUS F13 14225 4104 0.2885 2.1975 0.8591 6930 0.4872 0.8618
_CLASS100_b23 F28 14984 4138 0.2762 2.1038 0.8532 7046 0.4702 0.8588
_FIGS100_b23 F14 14565 3970 0.2726 2.0763 0.8519 6824 0.4685 0.8577
USERFACTOR1 F35 13609 3376 0.2481 1.8897 0.8517 5914 0.4346 0.8549



Similar to yours in that no one factor has broken the 0.90 roi mark (and remained there) for the year.

Also similar to yours in that the big picture stats for post time favorites (win pct at 0.3818 and $1.00 roi at 0.8627) are the highest I ever recall seeing.

Imo, the root cause is the result of an alarming trend.

According to the Jockey Club we are now getting roughly half the number of starts per year from each runner vs. what we were getting in 1960.

Link - here: http://www.jockeyclub.com/default.asp?section=FB&area=10

Imo, it goes without saying that today's game is much harder to beat than the game of even a few years ago -- and that players have to work harder than ever to find (and maintain) an edge.

But it doesn't mean finding an edge is impossible.

More on that last part in a bit...



-jp

.

Reply
jeff
5/31/2018
12:01:04 PM
Speaking strictly for myself, I've been forced to gravitate away from what has historically been my bread and butter: UDMs designed to capitalize on speed-pace-form horses (that used to win) in the 6-1 to 10-1 range.

Not because those horses don't win at similar percentages.

And not because those horses don't show up in the pps with the occasional ridiculously high morning line.

But because the money in today's pools is SHARP (compared to even two years ago) and not enough of those horses win at high enough odds.

Here is a link to a text file that contains a cut and paste from my Data Window of a UDM that I created late last year:
http://www.JCapper.com/MessageBoard/Reports/__ItFigures-a.txt


The UDM is designed to flag horses that really should be in the 7-5 to 9-5 odds range that end up going off in the 2-1 to 3-1 odds range.

Here, I am using three primary factor constraints:

• RANKF14=1

• AND RANKF35 <= 2 AND VALF35 >= 0.1125

• AND RANKF33 <= 3


Fyi, the primary factor RANKF14=1 is a Prob Expression I am using to score the most recent n number of starters for FigConsensus at the track-intsurface-dist.

And while rank=1 for this factor only has an roi of 0.8519 on my ytd Key Factors Report... which is hardly magical:

I have found that Prob Expressions help my betting avoid too much "against the grain" from a track bias standpoint.

The second factor RANKF35 is a UserFactor based on an MLR model where the output is a prob estimate between 0 and 1 after running exported JCapper data for speed-pace-form through the mlogit module in r.

Here I am basically scoring each horse for many of the same things I used to include in my speed-pace-form UDMs. Imo, speed-pace-form is still valid. But with the state of the game being what it is today you aren't likely to find a lasting edge by basing your world on a handful of rules for speed-pace-form. Imo, you are far better off using speed-pace-form as an integral part of an overall model.


And while rank=1 for this factor only has an roi of 0.8517 on my ytd Key Factors Report... which again is hardly magical:

The factor scores out a few points better than that at the tracks I am playing because the coefficients for the factors in the model are track-intsurface-dist specific because I am using SubGroups per The MLR Help Doc.

The third factor RANKF33 is a UserFactor based on an MLR model where the output is a prob estimate between 0 and 1 after running exported JCapper data for rider-trainer-railposition at the track-intsurface-dist through the mlogit module in r.

I was actually surprised to see this factor ($1.00 roi of 0.8806 on my Key Factors Report) score as well as it did.

My own cognitive bias, which gets in the way sometimes, says no way should this factor score out as well as it does.

The top three areas?

• Track bias from a FigConsensus standpoint.

• A pricing model from a speed-pace-form standpoint.

• And a pricing model from rider-trainer-gate draw standpoint.


The rest of the UDM is just filtering designed to eliminate sets of horses that have obvious negative attributes.



-jp

.

Reply
jeff
5/31/2018
12:25:26 PM
Couple more comments...

Even after all of that -- which was only made possible after thousands of hours of work over the past 4-5 years:

Still only getting a handful of +EV plays per day.

My significant other likes to joke I'd be better off working a job at McDonalds or Walmart.

Not 100% convinced she's really joking. She is joking...Right?


-jp

.



Reply
TPW
6/1/2018
7:23:32 PM
I hear that Jeff, I’m thankful I have a day job. I’m not nearly advanced as you are using JCapper, and as I’ve mentioned many times, I am so deep in playlist mode, I could never extract myself.

On the sharp money comment, I do see a lot of strong moves at post time. I don’t know how many strong movers are on losers, but it hurts when it is on a winner.

As always, appreciate your insight.

Terry

Reply
Reply

Copyright © 2018 JCapper Software              back to the JCapper Message Board              www.JCapper.com