Post Position Impact Value

[Jay Trotter]

Using the BRIS Track Bias Stats for Tampa Bay 9 on Wednesday, March 16th (Free PP attached) I'm trying to figure out the true Impact Values for Posts 2 and 3.

Total Races = 124

Rail Impact Value = 1.41 Win = 16%
1-3 Impact Value = 1.15 Win = 13%
4-7 Impact Value = 0.93 Win = 10%
8+ Impact Value = 0.90 Win = 9%

The way I read these stats is that the Rail has an IV of 1.41 and Win % of 16; and that Posts 1 to 3 inclusive have an IV of 1.15 and a Win % of 13.

How would you calculate the IV and Win % for Posts 2 and 3 only not 1,2 and 3 as the stats seem to show.

Any input would be appreciated,

Trotter:ThmbUp:

[RXB]

For posts 2 & 3 combined, 3 x 1.15 - 1.41 = 2.04, then divide by 2 to get 1.02 IV. This is based on the perfectly acceptable assumption that there have been at least three horses in every race.

[GameTheory]

Using the BRIS Track Bias Stats for Tampa Bay 9 on Wednesday, March 16th (Free PP attached) I'm trying to figure out the true Impact Values for Posts 2 and 3.

Total Races = 124

Rail Impact Value = 1.41 Win = 16%
1-3 Impact Value = 1.15 Win = 13%
4-7 Impact Value = 0.93 Win = 10%
8+ Impact Value = 0.90 Win = 9%

The way I read these stats is that the Rail has an IV of 1.41 and Win % of 16; and that Posts 1 to 3 inclusive have an IV of 1.15 and a Win % of 13.

How would you calculate the IV and Win % for Posts 2 and 3 only not 1,2 and 3 as the stats seem to show.

Any input would be appreciated,

Trotter:ThmbUp:

Assuming there were at least 3 horses in all those races:

Rail (post#1) = .16 * 124 = 20 wins (rounded)
post#1-#3 = .13 * 124 * 3 = 48 wins

posts #2-#3 = 48-20 = 28 wins = 11% winpct (28/(124*2))

To get the IV, we'd need the total runners, but again if we assume that there were at least 3 horses in all those races, we know that that the rail IV was 1.41, which means the expected number of winners was approx 14 for the rail. So that means that 14*2 = 28 expected wins for 2-3, which is just what we have. So posts 2-3 get an IV of approx 1.0.

Right?

[Jay Trotter]

For posts 2 & 3 combined, 3 x 1.15 - 1.41 = 2.04, then divide by 2 to get 1.02 IV. This is based on the perfectly acceptable assumption that there have been at least three horses in every race.

Thanks RXB. That makes sense. You would think they would just post the IV for Posts 2-3? :ThmbUp:

[hrspwr]

Glad this question came up and thank goodness some provided a formula cause I am clueless on the math, but I am curious how many players utilize post position IV's? I never do, but maybe I should?

[Overlay]

Glad this question came up and thank goodness some provided a formula cause I am clueless on the math, but I am curious how many players utilize post position IV's? I never do, but maybe I should?

I only take into account those positions that produce a percentage of winners that is significantly higher (or lower) from a statistical standpoint than would be expected just as a result of normal variance. (For example, at the time that Quirin wrote Winning at the Races, he found post position 1 to be a positive factor in one-turn dirt sprints of six furlongs or less, as well as in two-turn dirt routes of 1 to 1-1/8 miles. Post positions 1-3 were positive factors in two-turn turf routes from 1 to 1-1/8 miles at tracks where the main dirt oval was one mile or more in circumference, since the turns on the turf course were generally tighter than on the main track. (Outside post positions were at a disadvantage from an impact value standpoint, but I don't believe that any qualified as being significantly bad from a statistical perspective on either dirt or turf.) Individual tracks and surfaces can vary from day to day, but I still find Quirin's research a useful "plus factor" to remember.

[Turfday]

Comparing a 7-8 horse field to one of, say, 10-11 is like apples to oranges in regards to impact value.

[BIG49010]

I know I have big problems when the entrees drop below 90, other wise I do quite well with 10 to 11 in each race. Any insight into this problem I would appreciate

[Dave Schwartz]

Turfday,

In the original approach, which was %starters / %winners this was certainly true.

However, most people producing IVs today are, I believe, smart enough to actually count the "expected winners" based upon actual field size of the races. Thus, a horse in PP1 is expected to win 0.20 of a race in a 5-horse field and 0.10 in a 10-horse field.

The new formula of IV becomes:

Wins / ExpectedWins

Granted, this is not perfect, but it is probably pretty close to reality.

I think a bigger concern with PP should be that some tracks actually favor the outside posts over inside posts. In our software we have factors called "insidedness" and "outsidedness," which is probably a better way to go.


Regards,
Dave Schwartz

[Twenty Seven]

The trouble with a lot of studies representing IV according to (in this case) post positions is that the sample size during a meet, even at the midway point or beyond, is too small. And since track maintenance and bias can vary dramatically from year to year even at the same track, the stats have to be looked at with temporal limitations.

In addition, and in tune with the first point above, the winners coming from those particular posts -- #2 and #3 in this thread's study -- can be heavily skewed towards overlays or underlays. In short, there might be a lot of value coming from post 2 and little in post 3. All this is the result of random post drawings, and not at all from track configurations, pace scenarios, and the like.

IMO, relative value, trip info, and raw pace data have to be used with all post position studies to lend any context.

[Dave Schwartz]

IMO, relative value, trip info, and raw pace data have to be used with all post position studies to lend any context.

Twenty Seven,

Are you aware of any way to accomplish that statistically?

In other words, can you put a value on particular moves or is all just "seat of the pants analysis?"

[Turfday]

A switch from dirt to synthetic to dirt (e.g. Santa Anita).

[Twenty Seven]

Twenty Seven,

Are you aware of any way to accomplish that statistically?

In other words, can you put a value on particular moves or is all just "seat of the pants analysis?"

This is one of those dilemmas that I don't think can be resolved by either an objective or subjective solution. I agree with Davidowitz who emphasized a comprehensive approach which needed a high degree of discrimination in order to concentrate on which particular factors needed to be given priority for the specific race under study.

I will say, though, that the crowd's love affair with quantification (I'm guilty of overemphasizing it, too) means creative options can have more than enough value.

This can all sound so general, so an example:

Fair Grounds used to favor posts 1 and 2 at 6 furlongs in the mid 90s to a stunning degree. Posts 3 through 5 were acceptable. And if you bet on #6 through #12, you were either poorly informed, a masochist, or someone with a very unusual case to be made, along with high corresponding odds. I didn't know how long this had been going on, but obviously I bet accordingly.

The next year the inside bias disappeared (track maintenance and/or unusual weather). The funny thing is that post position strengths were all over the map. I don't have the exact data on hand, but post #1 (for example) would be great, #2 poor, #3 decent, #4 great, #5 poor, etc. I had and have no idea why this was so -- certainly the wide swings in adjacent posts meant that trips couldn't account for it -- but it destroyed all confidence in putting much stock in those results. The only thing I could think of was highly atypical value due to the vagaries of post draw.

IV for all trainer categories are very important. But I can't see the value in post position draw having any meaning in isolation other than the usual and obvious trip issues to the first turn.

[Cratos]

Using the BRIS Track Bias Stats for Tampa Bay 9 on Wednesday, March 16th (Free PP attached) I'm trying to figure out the true Impact Values for Posts 2 and 3.

Total Races = 124

Rail Impact Value = 1.41 Win = 16%
1-3 Impact Value = 1.15 Win = 13%
4-7 Impact Value = 0.93 Win = 10%
8+ Impact Value = 0.90 Win = 9%

The way I read these stats is that the Rail has an IV of 1.41 and Win % of 16; and that Posts 1 to 3 inclusive have an IV of 1.15 and a Win % of 13.

How would you calculate the IV and Win % for Posts 2 and 3 only not 1,2 and 3 as the stats seem to show.

Any input would be appreciated,

Trotter:ThmbUp:

It is not just post position; it is post position with respect to turn length (radius). For instance at Saratoga the turn length is 1476 feet, at Belmont it is over 2,000 feet and at Churchill Downs it is 1,296 feet.

Why do these numbers matter? It is because the race starting positions relative to the turn is different given the same race distance and this affects the post position impact value.

[Jay Trotter]

It is not just post position; it is post position with respect to turn length (radius). For instance at Saratoga the turn length is 1476 feet, at Belmont it is over 2,000 feet and at Churchill Downs it is 1,296 feet.

Why do these numbers matter? It is because the race starting positions relative to the turn is different given the same race distance and this affects the post position impact value.

Well, its a factor to consider just like anything else. Perhaps its significant and perhaps it isn't. It depends on the race and the numbers.

Based on the formula RXB presented the Rail has an IV = 1.41 but Posts 2 & 3 actually only have an IV = 1.02. With a decent sample size of 124 races I would think that could be a fairly significant difference. The original weight of 1.15 was obviously slanted because of the impact of the Rail.

[Dave Schwartz]

If you break that sample out by running style or early speed points, I think you will find that not all horses are effected the same way.


Dave