Effects on removal of the "No Bad" negatives

[SteveS]

Referring to the Orange block " race winners", I filtered out the route distances for GG, and found a total of 658 Sprint winner races. Below are a few isolated factors for illustration:

K=1 34.6%
RS "P" 38.4%
FR3-1 23.5%

Now, I filtered out all the NO BAD" negatives and reran Learn All.

The above three factors now read:

K=1 36.8%
RS "P" 32.3%
FR3=1 17.5%

Some significant changes, both up and down. My question is it appears it is a mistake to eliminate the Negatives at this point of review - is that a reasonable conclusion?
Thanks for any comments

Steve

[Rick]

I always use the reports to tell which "NO BAD" items to remove.

Look at the Learn All report and see which ones really look bad.

I normally look at the Win ROI.

Start with the worse looking ones and rerun the report.

[completebill]

Steve--------I, also, run Learn All as my first step and, like Rick, look at Win R.O.I. On my first run through, I start by eliminating all factors with a Win R.O.I. about 12-15% below the "All" win r.o.i.

RARELY, will ALL the negative items be big losers, in a given situation. Indeed, as you've discovered, eliminating ALL of them, carte blanche, will often hurt your r.o.i.

After I make my initial eliminations from the "All" report, I then run the TOUR report, looking at the same thing---factors whose win r.o.i.s are significantly less than the remaining overall win % total.

I'll then run the Learn more Report for further "fine-tuning" If I still have a large enough total of races remaining in my total, I will often then repeat the whole process. The factors will change as you make eliminations.

Of course, if you're "building" spot plays, or just trying to learn what is WINNING with your chosen parameters, you will, in addition to looking for fruitful negative eliminations, be looking for various POSITIVE HTR handicapping factors which you may want to incorporate into your model