The ONE RACE WONDER a.k.a the OUTLIER

[Tim Y]

When we evaluate a contender for a race, we review a series of efforts. RARELY are any two the same but rather express a range from objectively "good" to "bad."

If, on the other hand, there appears one line that dwarfts the rest, we have a problem. USE that one or not? It is a definite outlier and not normally representative of what this animal has ever shown before, so it is less likely to ever perform to that degree again in most situations EXCECPT:

1) a horse earned than ranking on a surface that it has not regularly run on. Especailly true with Dirt to Turf like Classic Causeway improved in the Belmont Derby recently, where his first turf outing was much the best of all his runnings up to that point.

2) when it is a quantum leap at the end of several improvements during the Spring-early Summer progression with 3 year olds showing their maturation on the way to the Triple Crown. Even this year's winner, even though nowhere near the relative numbers of lets say Epicenter, did come into the Derby improving. It is often astouding how much a maturing sophomore can improve in one inverval.

BIG changes on the same surface and in mature animals are much harder to accept, so there should be a way to incorporate this wider range in a way that an estimate of project can come of it.

The answer comes from a good book :How to Measure Anything by Douglas Hubbard. In such a series of numbers is best to think in a RANGE of numbers over any specific one. Quote: "In statistics, a range that has a particular chance of containing the correct answer is called a "confidence interval." a CI of 90% is a range that has a 90% chance of containing the correct answer."

The larger the number of entrants in a particular series, the wider the acceptable range included. Small ranges, of course, are more susceptible to sample error (like in a two year old).

I have found that one needs to look at similar lines in a series that can closely match the early pace of the race if came from and looking at that line for an alternative. Using the ORW (One Race Wonder) line is a ticket to overconfidence.

I have found that using the ORW wonder line is a performance that is rarely repeated, NOT even close in most past performances.

[Bill Lyster]

Tim,
I would only add to this by way of being observant of the distances of past performances. Many but not all horses show higher adjusted speed ratings for lower distances. For instance an 80 ASR for 5.5 furlongs, a 77 ASR for 6 furlongs, maybe a 75 ASR for 6.5 furlongs, etc. Works the same for routes. Looking for this trait and keeping within .5 furlong of todays distance meets the comparable standard of best of last three comparable, IMHO

[Tim Y]

ADJUSTED sped rating?

Why would one do that?

[DanBoals]

Originally Posted by Tim Y

ADJUSTED sped rating?

Why would one do that?

The adjusted speed ratings in RDSS are pretty damn good.

If a horse has say a 10 point advantage, even over a lone early speed horse, the 10 point adjusted speed rating advantage tends to hold. The RDSS adjusted speed ratings are the best speed figures I have worked with, and if you have not given them a look yet, you should check them out.

[Tim Y]

Originally Posted by DanBoals

The adjusted speed ratings in RDSS are pretty damn good.

If a horse has say a 10 point advantage, even over a lone early speed horse, the 10 point adjusted speed rating advantage tends to hold. The RDSS adjusted speed ratings are the best speed figures I have worked with, and if you have not given them a look yet, you should check them out.

thanks for the definition

[Ted Craven]

Tim - Bill Lyster, and Dan, are using the term Adjusted Speed Rating to refer to the same internally computed SR from Validator/Speculator, to distinguish it from the other SR accessible in RDSS - the TrackMaster SR. Adjusted by being projected to today's distance and surface, normalized by'Sartin Downs' Par Gaps, equalized by ITV and DTV - the way Sartin (and Wadsworth) did.

Observation of outliers and measurement of representative value from a large enough sample (or whatever sample you've got) can be made iusing any coherent final time Speed Rating (or any other rating), such as Beyer or TimeFormUS or Cramer PSR, or whatever you have accessible.

I studied Hubbard's book How to Measure Anything 10+ years ago - indeed, a good book. In RDSS, the Paceline Selection Strategy named 'Population Sample (2nd Call)' comes straight from that representative measuring strategy, and is also used in the NewPace module and in the EL+ rating. I never recommended it as I never did large enough comparative samples to know if it was better than Best of Last 3/4 as an automated strategy. But it's there.

Part of the strategy to seek ratings not dependent on only 1 paceline, and a 'first cousin' to the '2 lines better than the rest' notion espoused by you and by Dan in a recent PA post: http://www.paceadvantage.com/forum/s...7&postcount=56

Ted

[ranchwest]

I use a complex strategy to select a single pace line. A lot of what I rely on is based loosely on Sartin strategy -- correct distance and surface, etc. As many here do, I prefer a line that is from a competitive race. And many other criteria.

But, I believe any single number can be misleading. So, I look at speed with multiple composite ratings, class measured multiple ways and pace considered in multiple ways and strength of last race. That's how I find my outliers.

I have a problem with multiple pace lines for a few reasons:

1) If the horse is new to this distance/surface, how many pace lines will it take to be relevant?

2) If there is a line outside the norm on the high side, should you pick that line? Or, what other line will combine with that line to make it more representative? Really?

3) If all of the races except one have extreme variants, which two lines are you going to pick?

4) If there is clearly only one line that matches to Sartin guidelines, what other line is going to enhance the perspective of that line?

If two lines works for you, I applaud you. But for me, I look for one representative line.

[DanBoals]

Hey Ranch,

I am not trying to speak for Tim, but I think you might want to reread the thread.

If I understand correctly, he is talking about when is an unusually high ranking is valid and when should you view it with suspicion. He very astutely points out, that if the high rating comes from a first time change, like switching from dirt to turf, it might be due to the horse really being better suited to the new surface and excelling on it. He also pointed out something my dad always used to look for, that a rapidly improving 3 year old is the hardest horse to correctly evaluate. Improving 3 year olds can make amazing leaps in ability seemingly out of nowhere.

In the old "Yellow Book", Howard talks about people combining two lines and how that can be effective in some claiming races, but he recommended against anyone other than experts at the methodology doing this. This is not what Tim is talking about, so I am not sure what you mean by "two lines".

I think anyone who has the time, and is a good handicapper, looks at the horse's history as much as they can, even if that is only 10 or 12 pacelines. The idea is to get a feel for whether the horse is improving, falling off, or consistent. But, only one paceline goes into the software.

You should maybe read Tim's post again. The two situations he points out can lead to cashing some high mutuel tickets.

[Tim Y]

Software is nothing but a tool. Like a screwdriver is supposed to be used to implant screws into surfaces, it also can be used to pry open lids, or scratch one's name into metal.

Anyone who slavishly adheres to 25 year old rules and does not appreciate the BREADTH of what a program CAN OFFER if a bit of innovation is added to the equation, is missing out, BIG time, on what that software is capable of.

Excellent record keeping feedback promotes, or negates these newly attmpted innoavtive pathways...For example, experimentation lead me to discover that with the e/l balance, then with 2nd call madiens, then with
two lines better than the field, then order within chaos, then style over substance, then, (and one of my favorites) the Damon Runyon where one compares all the True Speed ratings in a race, and when they are all with the same whole number (i.e.55.4,55.6,55.2, 55.8 etc) one just looks at the decleration to often find a well paying winner hidden in that data. NO set of rules directed anyone to all of those new angles, and you will never find them if you are limiting your use of a multi-faceted software

[Tim Y]

there is a good example of this phenomenon in the 9th at Saratoga Thursday August 25: 10 A Mo Reay...One big effrot 2nd back and nothing else like it