Impact values. the correct spelling and the understood usage

[formula_2002]

I'm certain IV's are of no use in calculating an odds line because;


The calculated IV’s are AVERAGE IV’s and are impacted by the final post time odds that were used in their calculations..
Therefore one is attempting to calculate the fair odds using an item (the IV) which are themselves impacted by the same thing
you are attempting to predict..

The attached file will explain my understanding of the published IV data.

Regards
Joe M

[Tom]

There ae no odds influencing the IV's I have seen- it is just a measure of what percentage of winners have certain factor in their favor compared to what percent of all starters have it. Nothing about odds at all.
If 80% of all winners raced in 14 days, but 90% of all starters
raced in 14 days, the IV is 80/90= .89 IV.

[cj]

The impact values I am familiar with do not use final odds as part of the calculation. They use the factor being tested, number of starters displaying the factor, and number of winners displaying the factor.

For example, 10% of starters in maiden races are first time starters, while 20% of first time starters win. (Not true, example of course). Impact value is 2.00 because 2X the expected number of winner actually win.

Am I missing something?


CJ

[GameTheory]

Impact values are also not averaged, although they are calculated over a sample. But to talk of the IV of a single race makes no sense.

Anyway, while the odds do not affect the calculation of the IV, the odds can very well be highly correlated with the factor you are calculating, making impact value a dubious number to use if you are looking for profits (and who isn't?) I think this is what you are getting at? (i.e. top speed rating has a high IV, but very low odds, so what good is it?)

You might want to look at the articles on the AE ratio on Gordon Pine's site (www.netcapper.com). The AE ratio, or as Dave Schwartz calls it -- the "pool impact value", is calculated by dividing the Actual win % by the Expected win% as determined by the win odds (after adjusting for takeout) for the factor you are looking at. Anything above 1.0 and you have a factor that is out-performing its odds (not neccessarily profitable). Good factors will have a high AE Ratio and an even higher IV.

Unfortunately, AE Ratio can also lead you up a blind alley if you use it blindly. A few longshots can give you a high AE ratio for a factor with no profits AND a low normal impact value. But if you calculate it several times for a several different odds ranges, it will give you a pretty good idea how a factor performs in relation to the odds.

Of course, so will ROI. They are all related...

[formula_2002]

CJ
Here ia an actual case for one of the many Bris factors I have studied.

A # 1 rated factor has an IV of 30/12.5= 2.4. As I have said, that is an average IV for all final odds ranges.

For the odds range of .50 cents to 1.45 the same factor iv is .71

For the odds range of 1.50 to 2.45 the iv is 1.06.

For the odds range of 2.50 to 3.45 the iv is 1.73

In all the above cases the average loss per wagered dollar is -15 cents.

These figures are base on looking at over 100,000 horses.

So if you thing you 1-1 horse is going into the gate with an iv of 2.4, you would be wrong. It's going into the gate with an IV of .71.

As I have stated, The IV is a function of the unknown value you are trying to solve for (the fair odds)..

[formula_2002]

Game theory, I use a/e ratios in my analysis.

If you look at any report on my web page, you should find the last column to be 'A/E'.. I have been using that feature for over 12 years.

Any a/e over 1 is profitable. It's actually the % profit of a ducth play.
A 1.10 a/e yields a profit of 10% dutch, which in the very long run will equal the flat bet profit.

In one of my recents note on the board, I, expressed some concern for my 55 picks in Dec., because although I showed a 6% flat bet profit, the A/e was slighly below 1.

[formula_2002]

Game theory, here is a copy of the post I refered to .

DECEMBER RESULTS AND REPORTS
So far the month of December yielded the following:

55 picks
11% profit
22% winners

actual winner =12
expected winners, sum of (1/(odds=1) =12.78.

Note: I'm much more comfortable when the actual # of wins exceed the expected # of wins.


__________________
http://globalwinningpicks.homestead.com/GLOBALWINNINGPICKSX.html