This is the second part of a major article by American expert David E. Schwartz. Last issue he outlined various key strategies and in this article he sums up his approach. You can visit his website at for more details.

So, what am I doing? Well, the truth is, it does not matter. What does matter is my method for getting here.


  1. Select a handicapping path. Choose one that seems right to you. If you feel that early speed is where the answer lies, then start with an analysis of early speed.
  2. Develop a research log system. Formalise your research. Become a scientist at this. Determine an approach to try and lay it out. Take the time to type out a description of the system you are interested in.

    I did this by creating a folder named “Research”, and sub folders under it with titles like “test 001-ES System.” The first file in that folder is the main file. It contains a complete description of the theory behind the system, as well as the other fields, buttons, etc. that I think might improve the system. (These become my “mistake list” items.) Every “sub test” I run will be documented within this file.

    Whenever I do a “test run” it gets a letter code. Thus, the actual test file becomes “test001a.txt”
  3. Test one race at a time, looking at the results before you begin the handicapping process. Although I use the word “test,” this is really the development phase, not the testing phase. (Actually, in A.I. circles, we would call this the “exploration phase.”)
  4. Make notes in your race-by-race log by categorising the “mistakes” the system makes. Use similar wording for the mistakes whenever possible. It is much easier to look back at the races and say, “I played 50 races which had a PLF and did much worse in those races.”
  5. Stick with the original system for the number of races you committed to before the test began. Resist the urge to dump a system because it lost the first 10 races. Maybe in this sample there will be a hot streak in the last 20 races. How will you know if you never get to the last 20?
  6. When your sample is over, you should have a feel for how well it is working and, by looking at the mistake log, get some ideas for improvement.

    Implement whatever changes you need to make and do some more “exploration” by returning to step two. That is, create a new test log entry in the main file, document what you are changing and continue.
  7. Once you are confident that you have something that is looking pretty good, you finalise your “exploration” by documenting what you have found. Maybe I should say “what you THINK” you have found because you now have a hypotheses to test. You need to write out the final test  description in the main file before beginning the test.
  8. Set some reasonable number of races to test. Not 50, but not 2,000 either. Something like a couple of hundred. Remember to completely document how your test will be run. This includes days of the week and tracks to play. Ever noticed how the first 10 races in a test are always excellent? That is because you are modifying the system to make sure it works from the start. Heaven forbid you should lose at the beginning of a test.

    A great way to overcome this is to test the first day and then remove the first 20 or 30 races from the sample. Of course, don’t do this until the day is over.

    Another tip: When testing days, I like to jump forward eight days at a time. In other words, I start on a Wednesday because that is the day I start my handicapping week. Then I move to the following Thursday, then the Friday eight days after that. This way, a seven-day test covers seven weeks.
  9. After your test is finished, ask yourself if your hypothesis still appears to be true. Notice that I did not use the word “proven.” We are simply not going to do enough races to “prove” it works.
  10. Once we have determined that the system is performing as anticipated, we move on to the “exploitation phase.” This is where we begin looking at the races we wagered for signs of failure. Do we win in sprints as well as routes? How do our picks do when they are under 5/1? Under 2/1?

Do not attempt this until you have several hundred races under your belt. Just because you are 0-4 in Allowance races does not mean you can’t win in Allowance.

The exploration phase is an ongoing process. It will even continue after you have begun using the system in real play. But if you can hit on the obvious ones upfront, you can save a lot of disappointment.

Click here to read Part 1.

By David Schwarz