Traditional Pars vs. new EPR

[km]

Had some questions come up regarding the new EPR rating (estimated performance rating) vs. the old Par and Level numbers.

The EPR is on a scale that corresponds to the previous Quirin pars and levels. But it is a much more realistic number that targets an individual field of horses, rather than a static "par" that just averages a group of similar past races. The EPR looks over the field and determines how weak or how strong it is, how much improvement is likely and predicts what the winner's performance rating will be. That prediction: 'strength of field' is far more accurate with this method than the par solution.

The EPR is also a dynamic rating. If one or more of the top contenders scratch, the EPR will go down a notch or two as the core of the race is now weakened. The EPR is even slightly dependent on the paceline mode or user line selections. About 25% of the rating is calculated from the performance ratings chosen in the PL selection (the EPR will rarely change more than 1pt when switching PL modes). In contrast, the previous Levels would never change no matter how diluted the composition of the field or its pacelines had reckoned.

Now the past-performances also reflect the correct strength of the field from the past races as well. Among some fields of OCL, STR and various claiming conditions the quality of the fields can vary as much as 4-10 points from the par for that level. The previous par/level numbers could only average them but never give us information on how weak or strong individual races turned out.

In the 21st century the traditional Beyer or Quirin pars fail due to two basic problems: 1) they are based on written race conditions, purses and class charts that are supposed to be arranged in perfect order. For example, $8k claimers should be slower than $10k claimers in a large sample, but at a short meet like KEE or DMR, this doesn't always hold true. Yet anyone making pars would contend that the 10k par must be written higher to hold the symmetry of the par chart -- but that means will have underrated $8k races and overrated $10k races...... 2) the massive variety of race conditions and purse structures nowadays is either too complicated to comprehend, or the sample size too small to realize an accurate target par. The variance is just too wide for establishing conclusive numbers.

I realize the EPR is a progressive new strategy. But the improvement in class understanding and the accuracy of race assessment keeps us a step ahead of the crowd.

[MikeDee]

Ken As I understand it, in the HX5, export the ppXpar values for items 61 through 70 are now EPR values for those past races. Is this correct?

[km]

'Dee; correct; nPAR = EPR in all cases.

All the export files have nPAR, but since there were no other changes to those, I felt it was un necessary to revise them with a single field name replacement. HX4 had other changes anyway so nPAR was renamed nEPR permanently.

[zimal2]

Is the EPR what I've thought of as the class rating located just before the age on PPQ or FIG 2? Or is it elsewhere?
Zim

[km]

Zim,

EPR is found in the top header of every report only. It is the estimated performance number for the current race.

The past-peformances have the actual SOR (strength of race rating) in the location you specified. But it is not an "estimate" or "par" it is the actual quality of race rating computed when the race is over. Read this month's newsletter page-11 for examples. This number can be labeled a "class rating" if that helps. Below is a chart that relates it with the FIGs.

The first listed is the Quirin style figure which is used for the new EPR and SOR, as well as the PER and PAC ratings found in the PPQ and other screens. The 2nd number is the equivalent FIG from FIG1 or 2 ==>

080 --- 34
085 --- 30
090 --- 25
095 --- 21
100 --- 16
105 --- 12
110 --- 8
115 --- 3

If you see an EPR: 95 for instance, the expected winning FIG for the race will be 21. Older horses that have never run close to the 21 are in trouble. If it is a maiden race and the experienced runners have all been running around 26 or worse, then a FTS gets a big push to beat them.

If the EPR seems high compared to the recent performances of all the entrants, then HTR is warning you that one of the horses is likely to "break out" so check all changes, workouts and early speed, surface switchers, etc. carefully to tip off a possible big move forward - maybe at a nice price ($ and $$).

[tbrown]

Like the new EPR - makes the race shapes much more effective and meaningfull, after a weekend of playing around.
Just caught a nice FG horse who was exiting a SS race using nPar but was coming out a AF using Epr. It was the favorite, but that used to be a throw out for me-fav coming from a slow race shape.
This number looks really promising, especially at the smaller tracks.
Again, Ken, you just keep making it all better.
Sa-lute!

[km]

Appreciate that T'

While testing the EPR in realtime I found situations where the rating seemed way too high. Then I watched those races. In almost all cases, one or more of the horses exceeded expectations and exploded with a big race.

So my hope is that the EPR will lead us to understand when improvement is imminent. If the EPR seems too high compared to the recent performances of the entrants, trust the number and look for tell-tale signs of forward progress. Here is possible list, maybe some of you can add to it ==>

Past Class
Moving from Turf to Dirt
Positive Jockey Switch
Recent Claim or acquisition by "Super Trainer" (rating over 400, such as Dutrow at AQU)
Workout Rating is strong (over 83)
Blinkers on or off
Pace rating exceeds the EPR
Switch from route to sprint or vice-versa
Shipper from a stronger circuit
Freshening (30-90 day layoff) combined with a good wk rating

2yr and 3yr automatically improve with experience and maturity. The higher the PED rating, the greater the improvement will likely be.

[tbrown]

Case in point - FG R9 - LA Derby today.
EPR = 107 , seemed way too high. 100-102 more likely.
OK, #3 Wimboldon, WO 83+, triple advantage horse in Pac ratings last three, huge F1 advantage. Vi=22. Ordinarily, I would not look too hard at a maiden winner goin ginto a stake, but in this one, I had to use it.
The guy wins and pays $16.40.
This is one of the oldest plays in my book, predating HTR by 20 years at least - when nothing can run to par, look for something to steal the race on the front end.
I have always been on the lookout for good early speed plays and this new EPR looks like it might really set up some of these on more positive handicapping grounds.
All in all, a very good weekend using EPR as the basis of all my plays and the Pac Per race shapes to pick my horses.

[DOUG]

Tom,

When you speak about race shapes are refering to early, midpack, and late runners?

As Ken mentioned this horse looked like a big time breakout horse.

Am going to start checking out sor's for hidden class drops.

Thanks,

Doug

[Glen]

Cool, sounds like an interesting concept. I bet a good play would be dirt maiden horses with a PAC rank1 (2nd call quirin type fig) who have a PAC fig higher than todays EPR.

[DOUG]

Ken or somebody

Just want to make sure I know what I am looking at when I view the sor's.

The sor rating is a strength of field rating and has nothing to do with the race class itself (i.e. mc $25000) . Or is it something different?

THANKS,

DOUG

[km]

No more "Levels" or "Pars" Doug, the SOR is based entirely on the realtime output of the race and the quality of the field that competed. Written race conditions are no longer applicable.

Example: with the old par method a 3up male 10k claimer at SA; Par=100 (State bred = 99). This would be true for every past-peformance line no matter the quality of the field. With the SOR, we evaluate each field after the race is over. Two races of identical conditions run on the same afternoon could receive entirely different ratings depending on how strong or weak the race turned out. This gives you more insight when they return to race again. You are already aware that "10k claimers" are supposed to be better than "5k claimers" with a Par; but with the SOR you'll know for sure who is really moving up or down in class and there will be times when a 5k race was much stronger than anticipated and the returning horses will be facing easier company even though apparently stepping up in claiming price.

There are so many tricky conditions to figure out when creating a par chart that it is a minefield and inaccurate - so I gave up on it and brainstormed the solid EPR and SOR.

Any label applies to a race-rating like the SOR - "class rating", "race performance index", "strength of field", etc.

[DOUG]

Thanks Ken,

Don't want to be too dense here but I would like to make sure I am on the right track here.

So if a 5000 claimer comes out of a race with a 98 sor you would take a good look at that horse going to a 10000 claimer with an sor of say 94. Or would you prefer seeing a bigger difference between the two sor numbers.

Also say a horse was running for a nw2 alw and was running in an alw nw1 and the nw2 race had an sor of 100 and the egr for todays race is 98 would you consider that dropper to be taking like a double drop in class (one drop for the dtrength of field and one drop for the nw2 to nw1)?

DOUG

[km]

These are critical questions Doug, nothing dense about it, thanks for posting.

In both cases, compare the horse's own pace and performance ratings with today's EPR.
Can it compete at this level? Has it run any PAC or PER numbers close to the EPR today? Check the pp lines and SOR ratings -- has the horse competed at this range of EPR before, if so how did he do? Was he competitive?

You can estimate class moves in your own mind such as dropping from C20 to C10, but it is no longer quantified in HTR2 by written conditions. Conditions are nuts these days; where will this field below fit on a par chart when the horses return to open company ....?

"Starter Alw for for Non winners twice at 1mile or less in the last 180 days at a claiming price of 12k or less; Illinois bred fillies; 3yr and 4yr olds; races of 5k or less not considered; starters for a claiming price of 10k or higher preferred....................."

The SOR and EPR help you deal with this by stating a clear number that genuinely evaluates the quality of the field both before and after this race; instead of trying to guess at a Par.

[MVMcKee]

I was trying to clarify something for myself here, I probably got confused by the correlation to the FIGs.
If a race is won by a horse by 10 lengths, and the winner gets a strong figure, but all the other horses in the race are mediocre at best, what kind of effect would that have on the EPR (and SOR)?
I believe you said it measured the strength of the field as a whole and not just the winner. In the case of the SOR (post race), with the same result (10 length winner, etc) would it make a difference if a number of the other horses in the race dueled with the winner, or set a very quick pace, as opposed to just being outrun throughout?
Also, are the SOR ratings subject to "hindsight adjustments", like the Cramer figs? I am just curious if a number of horses exiting a race come back to run very well, or surprisingly poorly, will the SOR change?

[km]

The race fractions themselves are the computation McKee, the winner is not singled out. But there is a clear hold between the quality of the race and its strongest link (the winner).

If a field of maidens had a powerful 10 length winner and the rest of the runners were dogs, should we average all the dogs together to determine the quality of the race? That won't help us and we dont' need it anyway because when those losers return from the race their beaten lengths against the strong SOR will tell us the story instantly.

And yes, you smartly point out that the internal fractions for the losers are a key "watch" item despite their poor finish or speed figures. In fact, that is basically the secret to Fr1 and EP longshots. Years of research and experience leave us no doubt that early good effort expended during a highly rated race, regardless of a bad finish, is often a tell-tale sign of impending improvement and live bombs. But that shows up in the velocity ratings automatically - the race rating (SOR) is not a preequisite for expected live performance when the horse is FR1 = 1, odds 20/1; at 6.0f on dirt.

The numbers can change if Cramer re-sets the variant and race figures; that's called "back propagation"; and while that is controversial concept, it is a correction which can help us later.
But that will rarely be an issue with the SOR unless it is later learned that race was far slower or faster than originally listed.

The EPR and SOR are guidelines and benchmarks to help us understand where an individual horse fits and how much improvement is necessary to win. Nothing more.

[Paladin]

Will we still be able to get/use the "par" numbers if using a previous version of HTR2? (before the change to SOR and EPR)

[km]

Paladin - yes, any HTR version prior to March 2004 will have the Par/ Level ratings in them.

[hurrikane]

Ken,

have you found a range where the EPR/SOR becomes definitive. Say if a horse has no SOR within 3 pts of todays EPR is he a thowout if others have run to the EPR

Did you find the EPR/SOR more effective/stable as levels rise? More reliable. Was the reliability consistant across age/sex/surface/dist etc.

[km]

HK -

Water will always seek its proper level and so do maidens and young horses. Despite the fact that the EPR appears too high in many cases with maidens or any 2yr and 3yr race, one or more horses will tend to rise to the level. This puts a burden on the handicapper to make an educated guess as to which horses will move forward. I listed the most likely reasons for improvement earlier in this thread.

With older horses, the 3pt margin is probably a good guide. For example; if the EPR = 105 and an older horse has never run a race > 101, then it can only win if the other contenders fail badly or it gains some sort of pace advantage. By age 5, it is a rare thoroughbred that can suddenly improve 3 pts higher that its previous best.

The previous PAR ratings were definitely more accurate at class levels between 99 - 107. Those are the natural levels for quality thoroughbreds at major tracks. But stakes races with Par > 107 were often listed with Levels that were unattainable for most fields. The Ky Derby was listed with Par = 113 for example, yet not a single Derby winner ever ran that fast. But that is par for $1million 3yr Gr1 race in May according to the correlated par tables.

The PAR levels for cheaper races and complicated conditions were not accurate in a lot of cases. This was not because the par charts were off; it happened because there was a huge variance in the performance ratings of these races despite apparent purse, claiming price or race-conditions equality.

The EPR takes care of the problem by scanning the records of entrants and the races they have run in and putting up a number that is realistic for a individual field. Cheap races and bizzarre written conditions, Graded Stakes, etc are not an issue with the EPR since the number is based on actual performance and expected improvement among a particular group of horses only.

[km]

To follow up my example; I looked up the Derby last year and see the following =

PAR = 114
A regular Gr1 3yr race in May gets a Par = 113, but the Derby is listed a bit higher since it was $1million dollar purse and a Classic. Neither of these ratings was realistic for that field, or for most Derby winners sans Secretariat or Seattle Slew.

EPR = 110
The actual SOR for the Derby last year was 108, but Funny Cide did improve to a 110 in the Preakness and the SOR for that race was a strong 111. But nowhere near 114. The EPR was much more realistic than the PAR in relating the outcome and the quality of the Triple Crown contenders last year.

Funny Cide entered the Br Cup Classic in October at SA. The $4 million race-->
PAR = 115
EPR = 114

It was obvious in his PP's that Funny Cide simply couldn't muster a 114 and was over his head in the Classic.

[hurrikane]

my chi is good oh Massa one.....now..where is that in hand thing you spoke of.

[tomcat]

This is beginning to sound like the old speed number minimum thing... Look at the ML favorite, not his highest, but his lowest speed in the last 3...this then becomes the cutoff for anyother horse in the race.
This group becomes the contenders. If more than half the field, pass the race.