I spent a year in the 1980s handicapping with pretty much nothing but trainer stats. I came out of that year a little poorer and a little wiser, as is often the case with horseplayers. I still use trainer stats in a supporting role. And there are a few things I learned during that year-long losing streak.

Handicapping with trainer stats is rife with contradictory subcategories. Here�s a case in point from the first race at Hollywood Park on Wednesday, July 11: John Sadler, the trainer of Asking Bid, has won 16 of 55 or 29% of his races with horses in the second race after a layoff, for a 1.77 ROI. Good positive stat, huh? All you�ve got to see is that 77 cents profit � sign me up! But if you look further, you see that John Sadler and Chance Rollins, the rider on Asking Bid, have an abysmal 1/17, 6%, 0.21 ROI record.

What to do? You�ve got contradictory subcategories. It�s amazing how many treatises you can read about trainer stats without hearing a peep about them. That brings me to my second point:

As Ron Ambrose used to say, positives outweigh negatives. All else being equal, if you�ve got a negative subcategory and a positive subcategory on the same horse, discount the negative and pay attention to the positive. Why is this true? The random chance of a horse winning is about 12%, given an average field size of about 8 horses. A low win% trainer stat like Sadler and Rollins� 1/17, 6% record fits much more easily under the bell curve of normal performance than the 16/55, 29% positive trainer stat. The positive trainer stat is much more statistically significant. By pure happenstance, this brings me to my third point.

Small negative trainer stats are almost meaningless. I remember going to seminars where guys would quote stats like, "this horse is 0 for 7 the second time off a layoff," as if that was a justification for throwing him out. Frankly, it don�t mean a thing. Going 0 for 7 or 1 for 17 is well within the normal range of outcomes. Maybe one of Sadler/Rollins� losing horses came in second by a nose � if their head bobs had gone the other way, they�d have a perfectly normal 2/17, 11% stat. Never throw out a horse because of a bad stat when the sample size is less than 20. I don�t give negative stats much credence until there are at least 30 races in the sample, and even then I take them with a large grain of salt.