Turf vs Dirt question
 

I did some research today that was pretty enlightening. I've always known turf finishes are tighter than dirt finishes in general, but I didn't know to what extent. The value of a beaten length in a turf race is much greater than a dirt race. It actually breaks out really clearly by distance, less than a mile and a mile and over:

< 1m Value of a beaten length in turf races is 1.6 times that of a similar dirt race

> 1m Value of a beaten length in turf races is 2.0 times that of a similar dirt race

I use different values for the two, but not to that extent. Beyer made and adjustment to use his 6.5 furlong chart for turf routes, but that is about the 1.3 to 1.4 range, not 2.0. I don't believe Rags or Thorograph or BRIS or Equibase differentiate between the surfaces.

This is definitely something I'll be digging into a lot deeper the next few weeks. It sure looks like also rans in turf races are being given more credit than they deserve by speed figure guys, myself included.

While I am sure that your 1.6 and 2.0 values will give you a fairly accurate number, I have found that turf courses can be unique. Things such as length of grass, moisture, etc. can change the characteristics enough to make a difference.

This is 100% accurate. Turf races "bunch up" at the wire for 2 reasons mostly, less or no kickback and the slower nature of the internal pace. If i see a turf horse lose by 3 or 4 lengths i really try and remind myself that the horse was far from competitive and really got "drilled". Good post and good reminder to not give too much credit to those turfers who lose by 3, 4 and 5 lengths, a dirt runner who does the same thing is much more impressive from what i've seen.

I did some similar research, but not for figure making purposes. It was for a class oriented application. At the time I also broke it out by synthetic (which was still relevant) and off tracks (because margins on off tracks tend to be larger). I'm sure I still have the spreadsheet somewhere.

I also found evidence that winning margins don't expand as much as theory suggests they should as distances stretch out.

It's an interesting issue. I mentally filed it away as another reason to be skeptical about some figures, but found no real way to use it for my own purposes because margins also vary with pace and I couldn't control for that unless I used pace figures (which defeated the purpose of my class research). I shifted my attention to finishing positions, but ran into complications there also.

I'm going to revisit this soon for myself related to class. It's on my "to do" list. Now that I have a database to work with maybe I can make some progress.

Good luck with the figures.

Isn't there a connection to the energy distribution in a turf race vs dirt race? In 8.5/9f races, isn't there a much different spread and average time in the final fraction times between a dirt vs turf race

That is Very cool.

Yes, there is a difference in the two surfaces which is defined by the surface resistance to speed between the two surfaces. There is a very good "white paper" on racing surfaces and can be found on the Internet. It is somewhat quantitative with math and statistics, but it is an excellent read.

Of course, the races are run differently in general, though obviously there are numerous exceptions on all surfaces.

Here is why it matters in regards to speed figures which is what I was discussing. Imagine a speed figure scale where a beaten length counts as 2 points on dirt and turf. Horse A and Horse B both exit 9f races on turf.

Horse A exits a race won with a 100, was beaten 2 lengths. He gets a speed figure of 96

Horse B exits a race won with a 94, and he won the race. His speed figure is a 94.

In this scenario, horse A appears a length faster. But is he? What I am finding is that Horse A actually should have gotten a 92 if you double the value of a beaten length. He is now rated a length slower.

I'm not advocating that this is a better way to do things in every situation, but I think in the majority of the scenarios, it is. More to come in the next few days...

TURF vs DIRT
 

Any possibility of evaluating Kittens Joy offspring on Turf vs the rest of
the compositing Sires on grass.

It appears the KJ offspring ,if not winning, generally hit the board across
the major tracks in the U.S.

It's an interesting theoretical question.

Do we want to know who ran the faster final time or do we want to know who actually ran better?

Sometimes they are different answers.

This was basically the rationale Beyer used when using the 6f dirt beaten length scale for turf races. Since turf sprints were (evidently by comments I've read) much rarer and probably not well considered, this lines up well with the turf route observation you make.

Agreed, but I'm also saying his adjustment was based on observation and not actual numbers. Had he been using numbers it would have been much bigger. That is what surprised me so much.

I'm sure somebody could do it, but that isn't something that interests me much.

Adding to yesterday, another surprising bit of info. The slower the pace, the bigger the average margin of victory. The faster the pace, the smaller it gets. Conventional wisdom has been that slow paces bunch them at the wire, but the data doesn't show that. This was true regardless of surface and distance with one small exception, turf sprints are nearly even for average and slow paced races.

However, one thing did go as expected. When comparing turf to dirt, as the pace slows, the value of a length definitely changes with pace. The faster the pace, the closer the values become on the different surfaces. The slower the pace, the more they spread apart.

Good post CJ.
I think this illustrates how important linking Speed Figures, Pace Figures and Trip Handicapping is when it comes to evaluating races.
This is especially the case in turf races. The slightest bit for difficulty (wide trip, waiting for a hole, not ideal pace setup etc) can virtually eliminate a horse from hitting the board in a turf race.

Figures are fantastic tools but they only tell a piece of what happened and they need to be used with good trip notes and charts to get the full picture.

Final time figures for turf races are not part of my arsenal....just too much variation with multiple turf courses and rail settings at many tracks.
I prefer to do more of a fractional analysis (esp the 3fr) and therefore adjust lengths gained or lost during certain segments of the race (1/4m, 3/16m).

This is along the lines of what I'm thinking, final time doesn't tell the story in a lot of turf races, and even on dirt as well at times. There are better ways to measure races and they can still be measured numerically.

It can get tricky with slow paces. For example, a race at 9f with a crawling pace won't really tell us how a horse will do in a race at a similar distance with a fast pace. The pace is effectively changing the distance of the race many times.

Yep, definitely true. I always laugh when people label guys on this board as "figure guys", like we are blindly betting numbers. I'd be living under a bridge by now if I did that.

Hopefully the latest chart I posted adds some value too. There is always new stuff to learn in this game, and old "truths" that can be challenged.

Conversely, I use final times and traditional pace-n-speed figs for turf sprints under 6f.

Now, what to do with those 6f and 7f turf races at Bel is a thing that has eluded me for years. I tend to treat the 6f races as sprints and the 7f ones as routes, but I will admit that my numbers at these two specific distances need some tweaking, to say the least.

CJ - how do you treat 7f turf races?

That is what I'm trying to figure out!

I don't think the 7f races play as routes though based on the data I've seen. Even 7.5f races, which in the US are all two turn races on turf, play more like sprints that routes, probably in large part because of the proximity of the first turn.

However, even in sprints, as I noted, there is still a noticable difference in the value of beaten lengths on dirt and turf. That surprised me.

Good stuff.

Are you looking at just winning margins or margins between all horses?

They could produce mildly different results.

Many big winners are good horses that got loose on the lead in a relatively slow pace and drew off from the rest of the field and others are dominant speeds that torched the rest of the field in an average pace for them that was too fast for the others. Those are kind of unique situations relative to the typical race that could distort something.

You also might want to remove maiden races from the data to see how that impacts the results. I think my data in maiden sprints (at least on dirt) showed that there are more dominant "big" winners in maiden races than in the typical race because you run into some real killers at that level that will eventually move way up. That's more of a class issue than a race development and average beaten length issue.

Just winning margins. I think that is a more stable set to use than including other margins.

I've been doing this a long time, I know about all the distortions. I am greater than 99% certain that those things aren't changing the results with sample sizes as large as I'm using.

Yes the rail settings will make a difference in turf racing, but in comparing dirt racing to turf racing there are much more to consider.
The time of all races whether dirt or turf are primarily affected by the following 4 resistances: 1) Aerodynamic Drag, 2) Surface Winds, 3) Surface COF, and 4) Track Geometry/Turn Impact.
The fist two will affect the horse's dirt and turf speed performance similarly, but the last two have very different impacts on the the horse' s speed performance.

I know. We've discussed this some.

I just suspect that slow paces producing larger margins is probably the result of fresh faster horses already on the lead drawing off from the rest of field into the stretch as opposed to something related to times.

It is all related, including field size. Smaller races are less competitive gate to wire, thus larger margins, less contested pace, etc. When you look at the factors individually, the big picture starts to make a lot more sense.

I've discovered a few things along the way. For example, at TimeformUS, we should probably not have "Favors Horses On / Near the Early Lead" for races where field size is high, because the pace is rarely slow in those races no matter how it looks on paper.

This is one of the reasons I use velocities (average rate of speed in fps) instead of pace or speed figures. I also use variable beaten lengths based on the leader's speed. That way the value of a beaten length varies according to how fast the pace for that segment was run. Once the segmental velocities are calculated I use my own total velocity formula for calculating the whole race's velocity. It's all based on Sartin's work but with my own modifications. I decided to do it my way so that I don't end up doing exactly what a lot of others who uses velocities are doing.

Here was a good example from Race 2 today at Gulfstream. I'll give last turf race TimeformUS Final Time Figures, then what they would have been if I were counting beaten lengths more. I'm excluding horses that didn't match up with the best figures:

1 (5-1) Race 97, Horse 89, doubling beaten length factor (was 5 lengths back) would have given 81

2 (7-2) Race 96, Horse 89, adjusted to 82
4 (5-1) Race 94, Horse 84, adjusted to 74
5(11-1) Race 94, Horse 86, adjusted to 78
6(25-1) Race 99, Horse 90, adjusted 85 (race was a sprint, so adjustment not same)
7 (5-1) Race 94, Horse 92, adjusted to 90
8 (2-1) Race 86, Horse 86, adjusted to 86

The 7 had the highest figure going in anyway, but others looked close, his 92 was followed by a 90 and two 89s. But when the adjustment I've been talking about was made, the horse goes from a two point edge and three horses within three points to a four point edge over two other horses. He won easily at 5-1.

This is exactly what I was talking about in the beginning, overrating horses that weren't really competitive in turf races.

That advice is definitely worth the price of admission. ;)

That's where I am heading. There's no escaping the fundamental factors, but if you are doing what everyone else is doing, it's not going to be easy to find value no matter how good you are and how valid the ideas. If you use something that's measuring the same things in a new and equally valid way or find holes in generally accepted thinking, you have a better chance of finding value. I want a different model of thinking about reality. I have ideas, but testing them is a b$tch.

there is an obvious difference between surfaces, that is easy to prove.
but i don't believe anybody should say this many points equals that many lengths, regardless of surface.
you would obviously have different time standards for those different surfaces.
maybe you need to work in proportions rather than this many points equals that many lengths.?
running 69 seconds seconds when the winner ran 68(and the standard) on turf is nowhere near as good as perhaps running 70 when the race is won in 71(and time standard) on dirt.

proportionate tends to take care of most things that vary like surface, going, and other things.
maybe not perfectly but far better than using points equal lengths.

One other thing to toss around (related to what I said previously about large margins and loose front runners in slow paced races) is that closers tend to have their moves timed to the horses in front of them rather that blowing fields away. So if a surface is tilted towards closers, you might get smaller average winning margins for that reason.

CJ's data is suggesting that rather than pace being the dominant factor in smaller margins on turf, it might be running style. And if I am correct, the wider margins in slow paced dirt races might be the more frequent loose leaders blowing fields away. That happens less often on turf.

I'll try it tomorrow and eliminate races where the margin is >5. I doubt it is going to have much different though to be honest. These are huge samples and the outliers even out for the most part.

the dominant factor is going to be because dirt is more tiring than turf, thus everything will logically be magnified on dirt.

a long time ago when i was figuring japan for some people, it was blatantly obvious that the difference between male and female horse was different depending on surface

about 5 metres per 1000 metres average on dirt but only 3 on turf.

that will apply across the board no matter what tests you run, if you know how to run them.

it will apply to the class relationships, in that whatever the difference on average between class 'a' and class 'b' on turf it will be more on dirt.

it will be the main reason why margins are magniifed on dirt by comparison turf.

etc
etc
etc

it is also the main reason why raybo is correct doing what he does(if i read it right).

Then why don't we see more closers or horses with stamina win on the dirt?

if you get to the front when you are not working too hard and when you plenty left in the tank, then it becomes harder for others to run you down at the business end.

dirt by its very nature has to favour front runners more than turf does.

This is true coupled with race distance because a horse can only expend so much energy.

The majority of NA dirt races compared to NA turf races are sprints which are good for the front-running style horse.

Yeah, but races at the same distance favor speed more on dirt than on turf, and the margins are also greater between the horses.

I find this all very interesting, especially this point.

I might not have interpreted Classhandicapper's post correctly, but I believe his comment within his post of "it might be running style" answers the question about the margin difference on dirt versus turf.

In the end, I learned a lot from reading Beyer, Quirin, Quinn, etc. But they all sort of do things the same way. Quinn broke turf races out and rated them completely differently realizing the differences on turf, but I didn't find his method worked very well. It was much too crude in my opinion.

I've hypothesizing after this past week of studying things that it is the speed charts used that cause problems with turf. A horse that runs a time of 1:36 at a mile while losing by 5 lengths (his time, not the winner) should be rated the same as a horse that wins a race in 1:36. That is how most people do it, but it doesn't work on turf.

I think it is the speed charts that are to blame. They were just adapted by most of the pioneers of figure making from dirt to be used on turf, and I think they clearly underestimate the value of time on turf. If a beaten length in a turf route should have double the value of a beaten length in a dirt route, it only makes sense that the amount of time that beaten length represents should also have double the value of the corresponding amount of time on whatever speed charts or formulas are used to assign ratings to final times.

I have some pretty strong ideas on how to do this, going to test them out the next several weeks while racing is in the slowest part of the year for the sport.