%early Vs. %median

[socantra]

I've heard several people bemoaning the change from %early to %median in the later Sartin programs. I don't know whether it was for marketing or whether Doc was just sick of everyone ripping off the readout.

Doc did claim in the Follow Up that it was a better readout,but its really the same readout on a different scale. I ran the attached spreadsheet the other day on a few cards just to make sure there was no difference. It's not track specific but I did seperate out distance and surface for comparative purposes.

It is a small sample, but its all winners and should have a fair mix of tracks in it.

My math is not good enough to come up with an actual conversion formula. Maybe someone else can. If not, we should be able to make comparisons off the spredsheet.

Dick...

[forecast]

Hi Dick,

Really good question, one that I've asked myself for a long time now. With all the database handicappers out there, someone must have done a comparison, if not on this site, then maybe on the PA site. Maybe you could post your findings there as well, and look for responses. In any case, thanks for sharing your findings with us.

Forecast

[tom]

I have a conversion formula at home. I'll post in a couple hours when I get home from work.

One thing to remember is that Doc used 10 feet per length early on and then changed to various numbe of feet depending on the distance of the segment.
My work used 10 feet for all BL adjustments.

[Fast4522]

Here is a sample of some turf winners in Excel, I just copy and paste the line I use in my program to the model. The only thing I have to data enter is the win payoff. Its clear to see the % E of these 1.16 blade runners have a range that they peg inside of, and of course dirt winners will peg a tad higher on a norm.

[tom]

Here is some data from TP showing the two ratings side by side, and a chart that shows the formula to convert.

[socantra]

thanks, Tom, but either I am confused (which is entirely possible) or that is not the conversion formula from %early to %median or vice versa. Maybe its the slope of the line?

At a %early of 50%, the correlation for %median is 66.7%, basically 1/2 to 2/3. The multiplier is %early times 1.333. It offsets a bit on either side of that, but the equivalent of your formula of y=1.157x-0.2426 is not close whether I use 50 one way or .67 the other way.

My apologies if I'm wrong on this. I can make a calculator dance and am pretty good at basic arithmetic, but I never learned much math beyond that level.

Dick...

[forecast]

Hi Tom,

First off, great modeling! With records like that, you must be doing very well. Thanks for letting us have a look at your records. The first thing that strikes me is the small comparative difference in the total range of %M to %E. Here is the high, low, and range for both figures from your spreadsheet:

%M %E
70.17% 54.04%
65.74% 49.11%
______ ______
4.43% 4.93%

Doc Sartin said, in one of the Follow-ups, that %M was more predictive and had a tighter range to make modeling and contender elimination easier. If your sample is representative of most racetracks, then a 0.50% doesn't seem to offer that much difference. Ideally, we would like to have parameters that excluded as many horses as possible, while not eliminating winners from consideration before the race is run. It has been my experience that %E can be used as a loose guideline to eliminate non-contenders, but not a firm throw-out for horses above or below the established parameters.

As we are the remaining users of the Methodology, we should forge forward as a group and try to develop an energy formula that could deliver on the promise of %E and %M. I would be very interested in working to that end.

I'm looking forward to hearing more views and opinions on this subject.

Forecast

[tom]

Forecast - excellant idea.
I have tried to model the infamous E/L readouts from this data, but can never get the same numbers as Val3 spits out. I have also been able to model the old Energy@ readouts of percent for each fraction - ie, 35.44% - 33.67% - 30.12% with some luck, at least in coming up with recognizable energy distributon patterns.
An idea I have just started working is using the horse's three fraction energy percentages against the dream race three fractions to set a "par", or I am calling a target, for today's race. Then, comparing the differences in each fraction, come up with a number that represents how close each horse matches up to the whole race. Sounded good at the time, but it is not the quick fix I had hoped for!

[tom]

Socantra, try this revised spreadsheet.
I made the graph so that %M is a fucntion of %E.
When you plug values into the y=ax+b formula, you have to express the number as .5312, not 53.12. That got me last night, too.
I also made a EZ reference chart so you can go back and forth from one to the other.
Let me know if this makes more sense.

[Houndog]

It would be interesting to compare Median Energy from the old Hollywood pre-"cushion" track to the Median Energy post "cushion track". The same would be true of Keeneland, Turfway, and Woodbine.

I would assume Median Energy would be lower for post "cushion track". I could see a shift in read-outs where other read-outs may be more significant. Factor X in sprints and perhaps Hidden Energy would have added significance.