06/29/2010 1:30PM

A question for the pace handicappers among you

Email   For the last month, I've been working on an upgrade to the Moss Pace Figures.  It is more of a statistical fine tuning than an overhaul.  But in trying to reconcile database numbers with common-sense handicapping, I again came up against a dilemma that might be of interest to pacefig junkies, whether you prefer my pace numbers or anyone else's.

   My usual warning prior to these kinds of blog entries:  this is not light reading for the casual horseplayer.  This is definitely "inside baseball" stuff, as they say.

    First, some background.  One of my goals in the creation of Moss Pace Figures was to give horseplayers "true" numbers that maintain integrity over all distances.  Most pace figure methods use time charts based on pars for each class and distance, so that if Monmouth $20,000 claiming races for older males at 6 furlongs average fractions of :22.2 and :45.1 and a final time of 1:10.3 (which is exactly what they do average, by the way), the pace line for those clockings could be, say, 88-88-88, signifying that the pace and final time are exactly in line with par.  A 93-92-85 pace line would indicate at a glance that the quarter-mile and half-mile pace was faster than par, and the final time slightly slower than par. There is much to be said for this simplicity.  But I don't do this.  My problem with this style of pace figures is that the accuracy begins to diminish when horses switch distances: a $20,000 claimer who runs at a par pace at 5 furlongs almost always has more speed than a $20,000 claimer who runs at a par pace at 1 mile, yet in the above methodology their pacefigs might be identical.  I'll admit Moss Pace Figures aren't as easy to compare against par -- but - and I hope this makes sense - when I'm trying to ascertain if a horse can make an uncontested early lead in a race, I prefer pacefigs that tell me with clarity that a 90 is a faster performance than an 85, regardless of pars and distances.

To get there, I have overloaded Daily Racing Form's system by collecting data from tens of thousands of races at dozens of racetracks at all distances, and based my figures on what the numbers told me.

     As I said, I'm preparing to update Moss Pace Figures soon to reflect new numbers. No one will ever accuse me of skimping on data. For example, I recently lumped into Excel spreadsheets 101,844 races from one-mile ovals at Beulah, Calder, Churchill, Emerald, Fair Grounds, Fairmount, Finger Lakes, Hawthorne, Hoosier, Lone Star, Louisiana Downs, Meadowlands, Monmouth, Mountaineer, Oaklawn, Philadelphia Park, Pimlico, Prairie Meadows, Remington, Retama, River Downs, Sam Houston, Suffolk, Sunland, Tampa Bay Downs, Thistledown, Turf Paradise - and even pre-synthetic races from Santa Anita, Del Mar, Turfway, Woodbine and Golden Gate, and pre-renovation races from Gulfstream.  That's quite a contrast from the old pre-computer days of Picking Winners, when Andy Beyer came up with his figures by holing up with a poster board, Flair pens and a bottle of Jack Daniels.  

    From that cross-section of major- and not-so-major racing, here's what the number crunching showed:  At 5 1-2 furlongs, races that averaged a final number of 88 on my number scale (slightly more truncated than the Beyer Speed Figure scale we're familiar with) had a paceline like this: 87-88-88.  At 6 furlongs, the national average paceline was 85-87-88.   At 6 1-2 furlongs, it was 82-86-88, and at 7 furlongs it was 79-83-88-88 with the six-furlong call included.

    Of course, it makes sense that as distances get longer, horses are ridden less energetically from the gate, if only slightly.  The data shows that, and thus my pace figures reflect it.

    But truly accurate pace figures also require one additional and important "tweak."  At the average one-mile track with a 990-foot stretch run, the first half-mile of a 6-furlong race is run around 76% of the stretch turn. At 7 furlongs, the first half-mile is run around only 29% of the turn.  Running around a turn slows a horse, but how much?  I have a formula I've used for 30 years, ever since I first read about this phenomenon in an old Gordon Jones handicapping book. I used that formula to adjust the above pacefigs, but I'd like to hear other opinions.

    This tweaking based on turn times also comes into play -- in a big way - when attempting to calculate pace figures for distance races.  The spreadsheet shows that at a two-turn 1 mile distance, the national average pacefig line for a race with a final figure of 87 is as follows:  67-76-83-88.  The average opening quarter-mile of 67 is a substantial decrease from the first-quarter average of 79 at 7 furlongs.  Again, one logical explanation is that jockeys ride less aggressively in two-turn races than one-turn races, but an additional and perhaps more substantial factor is this one: the opening quarter of a 1 mile race is around most of the clubhouse turn, and the first quarter of most sprints is run in a straight line with no turn involved.  To a slightly lesser extent, the half-mile fraction is similarly affected.  The first half-mile of a 1 mile race includes the entire clubhouse turn, while the opening half of a 6 furlong races covers only 76% of a turn.

Clearly, pace figure points need to be added to the first two calls in route races to bring them in line with sprints and give us an apples-to-apples comparison, but exactly how much does a turn at a one-mile track slow progress?

I have great respect for the intellect of horseplayers, and I'm certain there are mathematicians and computer programmers among you.  If you care to, please pass along your thoughts on this - they might even put more money in your pocket someday. 
Proper Reality More than 1 year ago
Re. blog post of June 29. Here's what I would try. I would start by focusing on only two- (or one-turn) races for which the difference in distance in minimal. For example, 1m70 v 1m16th around two turns. This might minimize the effect of riding tactics (but that is only an hypothesis that would have to be verified—I'll say more about this later. Also, we'll eventually be looking for a data plot that may or may not be linear; again more later). I would then group tracks by percentage of the various pace-call distances that are run on the turn. Let's use four tracks for the sake of example: Track A: 30% of the first half at 1m70 is turn; 26% at 1m16th (ignore whether the difference in percentage is realistic). Track B: 28% 1m70; 24% 1m16th. Track C: 32% 1m70; 29% 1m16th. Track D: 27% 1m70; 23% 1m16th. The next step, of course, is to collect data on the pace figs at the different distances. Once there is a sufficiently large sample for each distance and track (and, ideally, track condition), I would take averages (rather than medians) as follows. Take all the races run at 1m70. Average the percentage of half mile that is turn. This naturally is necessary, because of the different number of races run at the distance at each track. Then average the pace figs, for all races at the distance, for the various pace calls. Do the same for the races at 1m16th. Comparing the averages will give the average pace-figure difference for the average percentage. This will be just one data point for our eventual curve. Now repeat the process, comparing races at, say, 1m16th v. 1m8th (again, using an appropriate grouping of tracks). This will give you another pace-fig-difference/percentage data point. Note that at this stage we also want to do this for all groups of tracks grouped in the way mentioned above. Ideally, this could be (and I suppose eventually should be) done for each track. I'm suggesting grouping as a short cut for purposes of getting enough data in the short run. Doing this will help verify if this method is feasible to begin with. Plot all the data points. We'd probably get a nonlinear curve. It should be easy to find an equation that fits the resulting curve. It should be easy to find someone or a program that can do that. Of course, the same procedure for one-turn races. Now, the promised worry about riding tactics. A few years back I was making my own pace figures, based on the Beyer scale. I found that using a parallel time chart to project for races at near distances was asking for big trouble. A year's worth of data for Pha gave me these fractions for races run in 1:44 4/5: 1m70: 23.16, 47.53, 1:13.54 1m16th: 23.15, 46.79, 111.58 ! To find such a difference, needless to say, paid off. The above exercise, I think, would probably work for many tracks. And it would be extremely interesting if I'm right about this and just a few tracks don't fit the pattern. Best, Mike
Anita More than 1 year ago
I use your figures instead of figuring FPS for a quick quide, plus the actual times. I believe the more adjustments that are made the more actual handicapping your doing and the raw data is lost. Handicapping is an art and we each need to come up with our own information. i.e. 2 horses run the 68-77-83 in 6 fl. one loves to be covered up and the other like to go in the 3-4 path. You offer a great tool, please keep it that way. If we could go strickly by numbers Beyer would be rich! Or the game would be over. Thanks!
ctfrombt More than 1 year ago
Randy, Shouldn't you be able to use the Aqueduct 1 mile pace figures to differentiate some of the factors. A good part of the first half is straightaway. I would think comparing the 1 mile and 1 1/8 mile (2 turn) early paces might tell you something.
Dennis Tiernan More than 1 year ago
This comment is for Chuck who posted earlier. Final beaten lengths in races is not calculated based on how far the horse was behind the winner at the time the winner crossed the wire but how long it took that horse to cross the wire behind the winner. So if a horse is 10th 20 lengths behind the winner as the winner crosses the wire, and he passed 5 horses in the last 70 yards and finishes 3.6 seconds after the winner, his chart would be 5th beaten 18 and not 10th beaten 20. Conversely, when chart callers call out the order at the points of call poles, they go by distance behind the leader at said point. The points of call are not based on time. The actual finishes are based on actual time. Also, 1 length at the 1/2 call is not equal to 1 length at the 1/4 pole or finish. A horse making up 3 lengths from the 1st call to the 2nd call is going much faster than a horse making up 3 lengths from the 1/4 pole to the wire. Good luck with all of the math.
Mike L More than 1 year ago
I read a great book called the Art and Science of Handicapping. I think both are equally important. I've always had bad days at the track when I rely too much on figs of any kind. If horse 'A' figs are 90 and horse 'B' figs are 88, then bet on 'A'. Very dangerous to the bankroll. As handicapping skills evolve, we should be able to look at how a horse got that fig and wager accordingly. Now, of course, the more accurate the fig, the better. But, it can also create to much emphasis 'science' and not enough on the 'art' when handicapping. See You at Oaklawn 2011!
Mike Gara More than 1 year ago
As a statistician, my suggestion, not far off from what is done on chef-de-race.com for graded stakes horses, is regression analysis ( with random intercepts.) For example, take all your mile races, enter a dummy variable for one turn (zero) and two turns (1). Then add other factors (also dummy variables)such as class , sex, age, and whatever else seems important (track.)Dependent variable is the observed split (1/4, 1/2. etc), also dummy coded. Random intercepts are necessary because splits are not independent. The regression weight obtained for one turn v. two turn could be a serviceable correction factor. Depending on the sample size, there are other even more technical adjustments to be made (e.g internal penalization of regression coefficients (weights) but I'd have to see the data
Bruce Friday More than 1 year ago
I'll take a shot at estimating the loss of speed caused by a turn early in the race. In the straightaway, the horse expends energy to carry the added weight of the jockey and equipment. In negotiating the turn, it must overcome the additional effects of centrifugal forces from the added weight. Assuming the horse travels a quarter mile in 24 seconds (V = 16.5 meters/second), the jockey weighs 120# (54.5 Kg) and the turn radius is 140 Km (Saratoga), the added centfifugal force is : Fc = (M x V x V)/R or 54.5 x 16.5 x 16.5 / 140 which is 106 Newtons. This should have about the same effect as carrying an additional 24 pounds of added weight (4.45 Newtons per pound). Using the Jockey Club Scale of Weight I estimate the effect of added weight at 4 furlongs as 1 length per 6 pounds added weight. 24 additional pounds would slow the horse down by 4 lengths or (at 1.28 Moss points per length) the equivalent of 5.1 Moss Pace Figs. (Alternatively, if you assume that 5# added weight yields a 1 length loss, the Moss Fig equivalent would be (24/5 x 1.28 or) 6.1 Moss Figs. I think this is pretty close to the time differential you measured. While I'm at it, here's my guess at loss turn length per added path. So how big is a path? Well, according to Google Pedometer the Belmont starting gate is 52.2 feet wide and accomodates 14 horses - 3.73 feet between horses. Let's assume that's a path width. Going around a half circle one added path wide yields (Pi x 3.73) 11.7 feet. Assuming an average length of 9 feet suggests one added path costs a horse 1.3 lengths. This will all be so much easier when we have Trakes data.... or will it?
Wayne Crimi More than 1 year ago
Randy, I am really pleased you decided to take this huge complex task on. The issue is that turns physically slow the horses down to some extent and jockeys also tend to slow down on turns as a matter of safety to lower the risk of bearing out etc... It's hard to isolate the impact of one factor from the other. I think if you looked at one turn 1 mile races and two turn 1 miles races you could probably get a good ballpark idea of how much the overall turn impact is (on average). Then you might be able to gain further insights by looking at the super fast paces in both types of races. The assumption being that in those situations the jockey was not holding back, so any difference would be attributed to the horse not being able to run as fast on the turn. I hope that makes some sense.
Jim Hughes More than 1 year ago
We all know that there is currently no way to accurately measure the ground loss for each individual horse when running on a turn. Nor the importance of other variables, like pace etc... But you should be able to get some type of insight about the "turn factor" by using bull ring racetracks. Because some tracks, like Charlestown in West Virgina, or Fairplex in California, have some fairly decent races to go by, even if the data base might be small. So some of the races, like the 6 1/2 & 7F races at Charlestown, which go around two turns, should allow you to come up with a better turn percentage variable to incorporate within your own numbers after you compare all the final times to similar class races at the mile tracks.
RCW More than 1 year ago
How about trying a simple data sort that compares the pace figs for all wire to wire winners in sprints to the same horse that went wire to wire in a follow up route race? Churchill, Gulfstream, Belmont(3 distances around one turn), Aqeduct, and Arlington, would be good data sources since one turn miles and 8.5, 9 furlong races could be included. That should be a fairly small sample but could be statistically significant. The theory is that such horses stretching out and winning back, are fit, and will have run a truly good race. The difference in the pace figs might be the adjustment you are looking for.
Bruce Friday More than 1 year ago
I think I have an understanding of the complexity you face in developing meaningful pace figures. I am a fan of incremental pace figs (as opposed to accumlated pace figs)and so spent the better part of a month working on a methodology to derive quarter mile (incremental) figs from your pace numbers. By doing so I would be able to calculate Sartin metrics for "average pace" and "sustained pace" as handicapping tools. With incremental figures I could also evaluate "turn times" without having the details of track configuration hidden in 1/2 mile or 3/4 mile numbers (29% here, 76% there). Unfortunately, by "normalizing" your turn times (as you are doing for two turn routes), you make estimation of incremental splits from your numbers virtually impossible. Oh well. As long as we're discussing "inside baseball stuff", let's look at some of the metrics. The PRIMARY unit of measurement in racing is TIME. It's understood that this is time over a fixed (more or less) distance. Speed on the other hand is a DERIVED metric. If one is willing to take into account the distance variabilities (post position, turn path, run up, gate position, etc. you can calculate a SPEED metric. To do this accurately you need the type of data provide by TRAKUS. Since the Beyer methodology doesn't do this, I sometimes think the product should be called Beyer TIME Figures. Now, by making the turn adjustments, you are taking the rating unit another degree away from the primary measurement of time and deriving, not a speed metric, but something more like an "effort" metric. One gets more credit for running a 1/4 mile turn in :24 than in running a 1/4 mile straight in :24. I don't mean to imply that such an effort is "wrong", but it does require that you invent any number of "tricks" (reference global climatic change brouhaha), in order to derive a number that normalizes sprints and routes and 1 turn 2 turn races. I might be inclined to let the (variant adjusted) time figures speak for themselves and leave the handicapper to understand he must take precautions when using route ratings to handicap sprint races and vice-versa. I think most of these tricks and uncertainties will be significantly diminished when TRAKUS data becomes generally available for most tracks. In the meantime, thanks for your Herculean effort to sort through the vagaries of pace.
felix jakubowski More than 1 year ago
20 Beyers for a mile is using 8ft as length of horse (20/2.5). Trainers know their horse best distance by 3 ft., if 1/5 sec = length of horse, then 3/5 = 24 ft. the energy depletion for 8 furlongs, 2 turns. I use 10 ft as the base length, adjusted all parameters to this standard, any errors can be adjusted by % of error. The second turn is a wall for many types thus is used as a conditioner by trainer. I eliminate the first turn and use 30 ft. as the mean, or 3 adjustment using 1 point for each fifth, 30 ft for a middle claimer, the number drops for classier levels. Track to track adjustments can be made for dirt races. In this manner, weight can be adjusted, beaten lengths can be adjusted, post position can be adjusted, 1,2, or 3 path can be adjusted, variant can be adjusted, any measureable parameter can be applied to achieve a rating on the horse-when a sizeable advantage is found, a sound betting formula is needed, but this method does explain why many horses get bet down yet cannot run to the low odds presented.
Alvy Singer More than 1 year ago
Should MEDIAN fractions in a specific class be used in formulas, instead of AVERAGES? Averages can be skewed by absolute bests. In a large enough sample size, the middle number is probably more accurate? Just puttin' that out there.
Paulie More than 1 year ago
As a recently retired data analyst now turning my attention to the game we all love, I want to note my enviousness of your ability to get such a stupendous database to work with. One way to calculate your current turn time task might be to normalize as best you can tracks with two turn miles versus tracks with one turn miles. The differences in the pace times would give you your answer. By the way, I want to applaud your quest for as much accuracy as possible in the face of the rampant noise--randomness--that pervades everything, including racing data. While I have found some "statistically significant" correlations, I value even those at the 10 percent level because they clarify ever so slightly comparisons of the horses racing in the perpetual data fog. Keep up the good work!
Mike L More than 1 year ago
Randy I see you did not publish my opinion, that is fine. I didn't expect it. I didn't think you would publish anything that goes against the Beyers. I'm not against Beyer figures, if people want to use them that's fine, I believe better to use them than nothing. However I don't. Years ago (back in the 60's) my dad developed speed figures and used to use the fractions from The Morning Telegraph to come with internal and closing moves as well as a rudimentary pace figures. In the 80's I worked on pace based off the class variant. I also tried to use the Beyer figures, actually worked off of them to try and establish pace figures. However that is when I noticed that the figures were off. In many cases where my figures (Delta )showed horses improving the Beyers did not reflect it until a race or two later. To me that just made no sense. Also on the low end of the figures I could not see how horses could run a 15 or display a negative number, I still do not understand the rationality. because of this I eventually stopped using them or trying to use them. As a handicapper in many ways you have to examine the theory of the handicapping you are using and you have to look at all aspects to come up with a conclusion. Once you are there you have to except the facts. This is much like a scientist. Keplar comes to mind. Keplar was a devout Catholic who believed in God. He theorized that if there is a god and if he made the universe than it, the universe, had to be perfect. He was an astronomer who studied the movement of the stars and planets. Through his observations after decades of studies he finally came to the conclusion that earth was not the center of the universe,that the planets did not revolve in perfect circles that the perfect design was not in place as he had theorized through most his life. In the end he discovered many things that made him cast aside his theory to accept the truth. That is the essence of science. The same holds true to an extent in the science of handicapping. You theorize what you believe should work, but to be successful you may have to cast aside that belief based on facts. Mike La Rosa SimRacing Form horseracingpark.com
Ralph D'Agostino More than 1 year ago
Your quest to establish "true" pace figures is truly admirable, and you have opened the eyes of many a pace freak (such as myself) to the many variables that have to be considered. I have been doing my own pace figures since the '70s with inconsistant success at sprints and total futility at route races. I have finally come to the personal realization that "pace" may be the wrong term and the misleading parts or positions of a race to best interpret or analize its significance. I do believe, however, that each race has a telling point - and only one point (simplification sorely necessary) - that may lead to enlightening interpretation. This point I call the "pace peak," and it occurs when par is supposed to be reached, such as the 1/2 mile point in 5 1/2 FURLONG race, and slightly after the 1/2 mile point at 6 FURLONGS, and so on and on as the distances increase. I realize that the "peak point" of very long races will have no significance to any point of much shorter races - but common sense tells me it is fruitless (apples - oranges?) to compare. I wish that TRAKUS or something comparable could be installed at every race track and timers installed at those positions which consensus (MOSS studies) determine are the "peak points."
felix Jakubowski More than 1 year ago
Trainers know their charges best effort within 3 feet of the wire; our job is to know how to measure that effort. Breeding knowledge is important, that alone can give one a winner. Super horses can give a strong run perhaps of 5 furlongs in anerobic state, weak horses may give a run of 3 furlongs in anerobic state, these efforts are measureable. Use the track record as base, assume a 4 furlong run to achieve anerobic state, now attempt to measure a 3 furlong run with pace, variant, distance, weight, trip, chaos, beaten lengths, and any other measureable parameter obeying the rules of mathmatics. Turf must be handled in a different environment, two turns require tweaking, track adjustments are needed. Try to explain the running of 20,000 claimers in 1:08+. (Yikes) More on this if interested. Felix 316
Laura Skamser More than 1 year ago
Randy I have a question for you. How are you going to handle the Arlington Park 1 1/16 race. I would be interested since I don't think of it has a good route race since the stretch run starts at the 2nd turn. Doesn't leave the horses much time to get ready for the stretch run. Would love to hear your opinion. Laura Skamser
greg s More than 1 year ago
I've been handicapping since I was 7, I am now 50, and do very well at making money, I have changed my handicapping process very little in 40+ years as everything that changes makes the same changes for all horses, so basically any change become a moot point. However in regard to your issue, I think the only time this would or should be an issue at all is the size of the track i.e. 5/8 mile track 7/8 mile track 1 mile track, etc due to the sharpness of said turn or TURNS otherwise it's a statistic that is moot. Oh and also track condition i.e. fast, good, sloppy, muddy, etc. Hope this helps you as much as it has me, now back to the pool to handicap the late pick-4 at Hollywood
Bill Legge More than 1 year ago
Great question and I'm interested in reading the comments. I have been handicapping and playing the races for forty years and how the game has changed. When I started, you played one track and had an early double, maybe an exacta or two and win, place and show. Today the sky is the limit with regard to types of bets. I use pace figures in an attempt to determine how the race will unfold,and I find them invaluable in this sense. In doing this, I'm not looking for exact numbers, but ones that put me in the ballpark. I want to know how much eartly speed is there and does one stand out over the others. If so, he's a contender and I'll move on to the pressers and closers. When I play, I'll generally play two tracks and due to time constraints I prefer to move on and concentrate on other handicapping factors and putting together a tentative game plan before going to the track or playing on line. From this point on, the most important factors are determining value and structuring your plays. Bill
ron More than 1 year ago
Randy, I have been making my own pace figures for decades now and have been over and over all this stuff many times. Please don't take this as a put-down, but your figure system is too complicated and that may explain their lack of popularity. I, for one, would love to give up the toil involved in making my own and use yours, but they are not really conducive to good pace analysis. There is something to be said for simplicity. I applaud you for the effort you've made, and for looking to make a much needed improvement. I wish I could convey my approach but space restricts any debate we could have, but I believe the discussion would be fun and rewarding. Drop me an email, and we'll talk pace-figs. doc
Ken Wiener More than 1 year ago
Randy: thanks for the great work you do. I would strongly suggest you not modify pace figures for routes to bring them in line with sprints. If a horse on the lead in a route drops to a 7F sprint, we want to be able to compare its 67 to the 79 for the horse that led in a 7F race last time. The router in fact may not have the speed to be on the lead in a 7F race and assigning a 79 to the router will cause real confusion. I think it would be very useful however to bring all one turns sprints in line with the 6F pars for a track and two turn routes from one mile to say 1 and 1/8 mile in line with the 1 and 1/16 par. The most common and important things we need to do are to be able to accurately compare all sprinters with each other and to be able to accurately compare all routers with each other. Thanks and regards.
Dave More than 1 year ago
All of this "Pace" stuff is interesting thought material. The fly in the ointment of pace handicapping is the -hole shot- analogy of drag racing. The thoroughbred misses the break and the numbers go up in smoke.
Chuck Sakach More than 1 year ago
One final comment before I leave this one alone: I have been observing a phenomenon I call "pack density". I can only describe this as the resistance to speed when a group of animals are bunched together as they travel through the race. For instance, a horse running second that is 2 lengths behind the leader faces much less resistance than a horse that is fifth and possibly 3 lengths from the lead. There is no way to quantify it so that it can be applied to pace figures and yet I swear that the phenomenon exists.
Chitownsteve75 More than 1 year ago
This is a very interesting topic. I think that the Suburban HC this Saturday at Belmont is a perfect race to determine who will be fastest to the lead in a one turn 1 1/8th mile race. Maybe Randy can give us a break down of this race with his figs. It's all Speed. I'm really liking Haynesfield on the outside. Perfect tactical advantage on the outside? A couple of these are coming off layoffs as well. Randy, stop by Rocky's Corner across from Oaklawn next meet in the spring. I'll buy you a beer. Keep up the great work. -Steve Simonovich