Any trhought? — Tom

Lots of them - but not quite ready to go public yet.

I'm in the final stages of a very different way to look at pace.

Multiple models - and a way to see which model to use in each race.

Running style is huge - because it predicts how RELIABLE the pace scenario really is.

A strong hint:
You have a model that points to an early horse winning, yet there are no early horses in the race. Now what?

Here's something that Andicap told me that I found useful:

I use energy to detect races that had more of a front-end skew than others by examining the 1st and 2nd calls. I promote horses who pressed the pace in races where the %E is above average in either call, especially those that remained competitive to the end. I demote closers who fail to exploit these set-ups.

I've found a number of high early energy races where the pace figures didn't point to an abnormal front end pressure but where the leaders came back strong when the pace figured to be less energy intensive.

I have decided to try an idea of using pace figures to do something similar.
I am using Dave's partimes, along with Beyer's variants I modify the race final time by Beyer's variant, then I recalculate the race times maintaining the energy relationship. Say the race went in 21 45 111 and the variant was fast 5 , the final time is adjusted to 112. 21 is .2958 of 111, so I use that multiplier on the revised 112, or 21.3.and the 45 becomes 45.6, so the race is now 21.3 45.6 112.
Next, I assign pace and speed ratings to the new times, in Quirin style, so the race might look like this:
115 106 103. The shape is very fast, average. This tells me the race was very fast for one call, but since it was average at the half, I can see the 2nd quarter went very slow (-9), so the race shape by quarters was F15 S9 Average so I can evaluate moves by where they occurr in the quarters.

Now this is all done in an excel sheet, by track, so I can almost automate the process. I just did all of Aqueduct sprints yesterday, from 2017, when the track was winterized to replace the inner dirt track through last weekend. A bonus to this is when I need to look at a race, I can see the whole day at a glance.

Using the race example above, 115 106 103, say the other sprints that day looked like this:

109 106 105
103 105 106
115 106 103 *** This race was the only really fast early, so I am confident it was genuine
101 101 99

But if it looked like this:
111 106 105 F6 A
105 105 106 S1 A
115 106 103 F12 A
108 105 99 F9 F6

The first calls went F27 / S1, or F26 over 4 races, so I would now revise the pace figures by lowering
all four by 6 points (26/4)

105 106 105
99 105 106
109 106 103 *** significant change in how this race was run
102 105 99

By automating the process, I can do more tracks, and only really handicap actual pace variants when I need to use a race.