Friedman fires smoking gun, shoots foot.

[TGJB]

Okay, I\'m back. There\'s a lot to get to, and I\'ll try to get to all of it, but I want to use this thread for a very specific subject-- Friedman\'s post about the Chilukki race-- because it is of overriding concern in evaluating the figures made by both companies, and impacts  several of the other subjects that came up.

Here\'s what Friedman said:

On the day in question (4/28/99) Chilukki won the first race at CD. The track was worked on several times, but specifically between the first and second races. The track was AM muddy, track listed as good for the first, fast for races 2-5, sloppy 6-9 when more rain hit. Friedman says \"accordingly\", they used 3 variants for the card. He then goes on to talk about subsequent events indicating the number they assigned turned out to be correct, which I\'ll get to later.

Now the key stuff, and you have to read this pretty carefully-- I had to read it twice and show it to someone else before I could really believe he said it. Friedman says they came up with the figure by going over the historical data about the relationship between 4 1/2 furlong races and other sprint distances at CD (he says Kee by mistake, but that doesn\'t matter). Now, I could have a field day with that alone, but right now I have other fish to fry-- actually herrings, red ones. In this case, THE RELATIONSHIP BETWEEN THE DISTANCES IS NOT THE ISSUE-- THE RELATIONSHIP BETWEEN THE TRACK SPEEDS IS. The two races could both be at 6 furlongs and it wouldn\'t make any difference.

So, how did they come up with a variant for a race that was not preceded by another race, run over a track that had a different moisture content than the following race, and that was worked on between races?

Possibilities:

1- They did it at the same variant as the second race. This is what I thought they did, and it would have been wrong to do it that way, but Friedman says they did not.

2- They cut the race loose, and did it based on the horses that ran in the race itself. This would be a real news story, a feat worthy of Karnak. There were exactly 2 fillies in the race who had run before, one start each over the downhill 4 1/2 course at Keenland (which Ragozin gave figures to, another future field day). Even if you think those numbers were accurate, they both had run in the mid thirties, and both were assigned jumps of at least 15 points in this race, so they obviously weren\'t used to construct the variant here. So if they did the figures based on the horses that ran in the race they did so using workouts and/or pedigree, which seems unlikely.

3- Which seems to leave only the following-- they used the variant for the second race, but adjusted it with a mechanical correction. And the obvious question is, how did they come up with the correction? Possibility 1 is that it is arbitrary. Possibility 2 is that they used \"careful attention to the OBJECTIVE condition, (and) the track history in SIMILAR situations\" (emphasis added).
   Okay, what is the \"objective\" condition? When Friedman said that the listed track designation changed, and they \"accordingly\" changed the variant, is he serious? Is he actually claiming that a good track at CD always has the same variant, and always  differs from the fast track variants (which of course are always the same) by an exact amount? And what are \"similar\" circumstances? All times where the track goes  from good to fast have the same variant relationship? Really? How about the ones where they worked on the track between races, do those have the same relationship? Do you know all the times they worked on the track between races, at all tracks?
    My guess-- at best, this is another example of the Ragozin use of broad averages. They may have taken a bunch of times that tracks went from good to fast (independent of work being done) and averaged them to come up with a rough correction. First of all, the only reason averages are used at all, ever, is because there is variability in the results you are testing-- if the average is a 4 point difference, some are 8, some are no change, etc.. That means the number you are assigning may be assumed to be AT BEST relatively close (in the above example, within 8 points), meaning only if the situation you are looking at actually fits exactly into the situations you measured with your average, and the results you measured were not too variable. And since they did work on the track, all that goes out the window, unless you did a seperate average for all the times they did work, which for starters means you have to know every time it happened.


Friedman\'s other point is that the subsequent figures the fillies ran vindicated the 4/28 figures because all the fillies (except Chilukki, of course) ran back to those numbers in the next 3 starts. First of all, as I\'ve pointed out before, you use earlier figures to make the later ones, so it is to some degree self-fulfilling. But more importantly, PLEASE-- these are 2 year olds, in April. OF COURSE they\'re going forward, rapidly-- if you were to believe Ragozin, the two making their second starts in this race both moved forward 15 points.

As for Awesome Humor winning-- it\'s meaningless. We don\'t leave boxes because the races come up fast-- we do it because there is not enough info to make figures with. It wouldn\'t have mattered if the Chilukki or Awsome Humor races had come up 2 seconds slower.

I urge everyone to read the above and Friedman\'s posts again, carefully. Please keep all comments on this string to the narrow confines of the figure making questions discussed here-- there are other strings for \"all figures are imperfect\", or \"Friedman wins so the figures must be good\".

[Silver Charm]

Who started this mess anyway?

Seriously, I\'ll followup with more on this later but his explanation has more holes in it than SWISS CHEESE.

I would like to leave with this thought and I challenge anyone to research what I am about to say and point out if I\'m wrong.

The runner-up in that race was Shattered Heart, (Bobby Barnett-trained, Robby Albarado-ridden). She ran back two weeks later and won. Her Ragozin number was a 21, SIX POINTS SLOWER than what he gave her in the Chilukki race.

As Freidman himself wrote:


\"It is not admirable to encourage others to use factors in making handicapping decisions that are not reliable indicators of anything at all.\"

[TGJB]

Just wanted to make sure this string gets moved up so it doesn\'t get lost.

Okay, I\'m back. There\'s a lot to get to, and I\'ll try to get to all of it, but I want to use this thread for a very specific subject-- Friedman\'s post about the Chilukki race-- because it is of overriding concern in evaluating the figures made by both companies, and impacts several of the other subjects that came up.

Here\'s what Friedman said:

On the day in question (4/28/99) Chilukki won the first race at CD. The track was worked on several times, but specifically between the first and second races. The track was AM muddy, track listed as good for the first, fast for races 2-5, sloppy 6-9 when more rain hit. Friedman says \"accordingly\", they used 3 variants for the card. He then goes on to talk about subsequent events indicating the number they assigned turned out to be correct, which I\'ll get to later.

Now the key stuff, and you have to read this pretty carefully-- I had to read it twice and show it to someone else before I could really believe he said it. Friedman says they came up with the figure by going over the historical data about the relationship between 4 1/2 furlong races and other sprint distances at CD (he says Kee by mistake, but that doesn\'t matter). Now, I could have a field day with that alone, but right now I have other fish to fry-- actually herrings, red ones. In this case, THE RELATIONSHIP BETWEEN THE DISTANCES IS NOT THE ISSUE-- THE RELATIONSHIP BETWEEN THE TRACK SPEEDS IS. The two races could both be at 6 furlongs and it wouldn\'t make any difference.

So, how did they come up with a variant for a race that was not preceded by another race, run over a track that had a different moisture content than the following race, and that was worked on between races?

Possibilities:

1- They did it at the same variant as the second race. This is what I thought they did, and it would have been wrong to do it that way, but Friedman says they did not.

2- They cut the race loose, and did it based on the horses that ran in the race itself. This would be a real news story, a feat worthy of Karnak. There were exactly 2 fillies in the race who had run before, one start each over the downhill 4 1/2 course at Keenland (which Ragozin gave figures to, another future field day). Even if you think those numbers were accurate, they both had run in the mid thirties, and both were assigned jumps of at least 15 points in this race, so they obviously weren\'t used to construct the variant here. So if they did the figures based on the horses that ran in the race they did so using workouts and/or pedigree, which seems unlikely.

3- Which seems to leave only the following-- they used the variant for the second race, but adjusted it with a mechanical correction. And the obvious question is, how did they come up with the correction? Possibility 1 is that it is arbitrary. Possibility 2 is that they used \"careful attention to the OBJECTIVE condition, (and) the track history in SIMILAR situations\" (emphasis added).
Okay, what is the \"objective\" condition? When Friedman said that the listed track designation changed, and they \"accordingly\" changed the variant, is he serious? Is he actually claiming that a good track at CD always has the same variant, and always differs from the fast track variants (which of course are always the same) by an exact amount? And what are \"similar\" circumstances? All times where the track goes from good to fast have the same variant relationship? Really? How about the ones where they worked on the track between races, do those have the same relationship? Do you know all the times they worked on the track between races, at all tracks?
My guess-- at best, this is another example of the Ragozin use of broad averages. They may have taken a bunch of times that tracks went from good to fast (independent of work being done) and averaged them to come up with a rough correction. First of all, the only reason averages are used at all, ever, is because there is variability in the results you are testing-- if the average is a 4 point difference, some are 8, some are no change, etc.. That means the number you are assigning may be assumed to be AT BEST relatively close (in the above example, within 8 points), meaning only if the situation you are looking at actually fits exactly into the situations you measured with your average, and the results you measured were not too variable. And since they did work on the track, all that goes out the window, unless you did a seperate average for all the times they did work, which for starters means you have to know every time it happened.


Friedman\'s other point is that the subsequent figures the fillies ran vindicated the 4/28 figures because all the fillies (except Chilukki, of course) ran back to those numbers in the next 3 starts. First of all, as I\'ve pointed out before, you use earlier figures to make the later ones, so it is to some degree self-fulfilling. But more importantly, PLEASE-- these are 2 year olds, in April. OF COURSE they\'re going forward, rapidly-- if you were to believe Ragozin, the two making their second starts in this race both moved forward 15 points.

As for Awesome Humor winning-- it\'s meaningless. We don\'t leave boxes because the races come up fast-- we do it because there is not enough info to make figures with. It wouldn\'t have mattered if the Chilukki or Awsome Humor races had come up 2 seconds slower.

I urge everyone to read the above and Friedman\'s posts again, carefully. Please keep all comments on this string to the narrow confines of the figure making questions discussed here-- there are other strings for \"all figures are imperfect\", or \"Friedman wins so the figures must be good\".