Selection Process
Hi, I'm Jimmy Steward. I’m the guy that owns Horse Track Stats and JRR. I've been selling horse racing “tip-sheets” since 2015. I closed my website in July of 2018 to work on a new algorithm that will bring more accuracy to my horse racing selections. I launched my new website Monday October 1st, 2018. I have made many friends on twitter and because of twitter, have built a close following in the horse racing community.
How did I get into to predicting outcomes of Horse Races? For years I was a normal track handicapper, I studied the program, looked at Jockeys, Trainers, Owners, Horse Speeds, Sprinters, Closers. Overall Speed, etc… I also played a ton of vertical bets, Exacta, Trifecta and Supers!. If I hit 3 out of ten races I thought that was a good day. I also paid $20-$50 for those online tip sheets and found that a monkey picking numbers from a hat could do just as good!
Something changed in 2015, a light came on, and I thought what if I did not pay any attention to the daily race programs, etc.. and just look for a consensus of which horse will win the race? So I searched the internet and found 50-60 free websites with amateur and professional handicappers that listed their picks as to which horse will come in 1st, 2nd, and 3rd. Of course if you ask 20 to 30 people who will win the race you get many different opinions. Hooray! that’s what I was looking for. If 20 people look at one race you will find that 70% of the people select the same horse to win and so on and so on.
Next, I assigned value to those assumptions and in a hierarchical/ordinal level of confidence. I looked at my Bible “The handbook of parametric and non-parametric statistical procedures” and there it was the “Wilcoxon Rank Sum Test” ….“Let X1, ... , XN be a set of N observations. The method presented here does not use the observations but rather the ranks of the observations. The rank of Xi among the N observations, denoted R(Xi), equals the number of X'j s ≤ xi. There are many cases where ranks are the observed values. For example, consider a contest where the judge has to rank the contests. It is also helpful to consider the ranks of the data where there are extreme outliers present”
First, in order to run the test you need data. There are numerous people posting picks online. Next, place their picks in order as to finish order. What you will find is a consensus of which horse will win and a decrease in value for the next horse OR sometimes it might be a tie. For example: if six horses enter a race and you use 20 different handicappers you might see ten people pick the #1 to WIN, five people pick the #2 to WIN and the remaining five people may pick the 3, 4 and 5 to WIN. Next look at which handicappers pick for PLACE…Again out of 20 handicappers ten people pick the # 1 and seven people pick the #2 and three people pick #3,4 and 5...to PLACE. Again look at 3rd place and repeat the process. What was difficult about the process was assigning a weighted value for track bias, that took some time to master and after watching a years worth of racing (no betting) I nailed down the correct weighted formula per track.
So now you know how I get 3/4 of my numbers - the other 1/4 th of my data involves times and distance averages per horse. In fact, I shut down my website of two-years "Horse Track Stats" just to work on making this aspect of my research more accurate. What I did was add a simple test of independence,using a 2x2 contingency table that looks something like this:
| Low Odds | High Odds |
------------------------------------------------
(In the $)
By Track |
By Grade |
By Distance |
------------------------------------------------
(NOT in the $)
By Track |
By Grade |
By Distance |
------------------------------------------------
I use SPSS software to analyze my data. I look for results that are significant. Once you put ALL the data together - BOOM! you get my selections!
|