Track surfaces: fastest to slowest

[GameTheory]

Based on my personal track-equalization procedure, and looking at fast dirt tracks only from Jan 1, 2003 until mid-March 2004, I get the following rankings from fastest to slowest.

Now I'm not talking strictly about raw times here, but relative surface speed (the faster tracks are easier to run over -- the same animal in the same condition running the same distance should run a faster time over a higher-ranked track). So if you were comparing times from a faster track to a slower one, you'd need to add a few ticks to the faster track time to make it equivalent to the slower track time.

I didn't separate distances, just averaged everything together. And of course certain tracks like Mountaineer alter their surface suddenly at certain times of the year, so they really should have two rankings. Given those caveats, anything look majorly out of whack? (Is Charles Town really that slow?)


1) EMD
2) FNO
3) PLN
4) YAV
5) LA
6) TUP
7) SR
8) GG
9) SAC
10) SOL
11) BM
12) STK
13) BMF
14) SRP
15) ALB
16) ARP
17) CNL
18) MED
19) WDS
20) SUN
21) SA
22) LS
23) AQU
24) HOL
25) RUI
26) BEL
27) BEU
28) DMR
29) STP
30) MAN
31) GP
32) FPX
33) CD
34) HOU
35) AP
36) HST
37) TIM
38) FL
39) TDN
40) RET
41) SAR
42) DEL
43) OP
44) RP
45) BRD
46) MTH
47) KEE
48) CBY
49) FE
50) LAD
51) HOO
52) PRM
53) FG
54) EVD
55) LRL
56) FMT
57) PEN
58) PHA
59) IND
60) AQU (inner)
61) PIM
62) ELP
63) TP
64) PM
65) FP
66) TAM
67) RD
68) FON
69) BOI
70) MNR
71) WO
72) HAW
73) LBG
74) ASD
75) CRC
76) CLS
77) NP
78) DED
79) SUF
80) GLD
81) CT
82) LNN

[JackS]

I have done about the same thing using most of the major tracks and a few of the mid and minor. My adjustments are about 2-3 years old now, so I really need to start over. My adjustment uses CRC as a base to compare all tracks because this is one of the very slowest at least among the majors. I have even took this concept one step further and adjusted to a single number for all distances to be applied to pace lines (pace call and final time) run at most any track . One caveat is turf. Amazing as it may seem, CRC is also one of the fastest turf venues (at least here on the east coast) and most definitly in the state of Florida. I would recommend that everyone do this even if it's for final time only. Comparing final time from track to track is virtually worthless without a par adjustment. For emphisis and to remind everyone, adjustments can be wild when comparing the slowest tracks to the fastest.

[sjk]

Here is a list using 6 furlong pars:

FNO
YAV
EMD
SAC
ALB
TUP
SR
SOL
GG
PLN
MED
SUN
ARP
IND
HST
CNL
BM
WRD
BRD
STK
LS
PRM
SA
GS
FE
CBY
MTH
STP
OTC
HOL
BEU
RUI
RP
PHA
SH
PEN
DMR
FPX
MNR
NP
LAD
HIA
LRL
WO
TDN
HOO
EVD
AQU
ASD
FL
TAM
BEL
CD
GP
FG
PM
DEL
TP
KEE
AP
OP
HAW
SAR
RET
ELP
AQI
SPT
PIM
FMT
LNN
FP
WDS
RD
CLS
RKM
CRC
FON
SUF
HPO
GLD

CT does not appear on the list since they do not run 6 furlongs; they are the slowest track on the 7 furlong list.

Lists will differ based on the time frame from which the data was gathered. I quick look shows considerable overlap between the lists.

[GameTheory]

At 6 furlongs, the spread in times (that I consider equivalent) from the fastest track to the slowest is something like 5 seconds, a huge difference.

My numbers are all distance equalized, so nearly all fast track races are considered and averaged together. Then I'm ranking by the average amount of adjustment required to bring a race in line with my universal standard.

[JackS]

sjk- Overlaps would be unavoidable due to the ever-changing circumstances such as weather, maintainance and who knows what else. Auto updates via computer should keep most tracks in perspective. For non computer users- changes should be minor and anyones homemade chart can be accepted as" close to accurate".

[JackS]

GT- I've got adjustments that are more then 6 seconds in cetain routes comparing Calder to some tracks on the west coast. This should be enough to convince everyone how important these adjustments really are.

[Zman179]

Originally posted by GameTheory
(Is Charles Town really that slow?)


The surface itself isn't super slow, it's the fact that the turns aren't banked that makes the track seem like a quagmire. That'll change when CT closes for a couple months later this year for track renovations (and a possible extension to a 7/8ths).

[sq764]

I love the beginning of the Delaware Park meet when you get Suffolk shippers coming back and racing here. A lot of people discard them for their slower times, but quickly see how much faster DP is..

[kenwoodallpromos]

Sorry, you may already know my comments- with few exceptions for a few very slow tracks, you cannot use any 1 number, speed, par, variant, or ranking for any track unless you want to bet an EP when there is a negative rail bias or they dig up the track like 03 Ky Derby or 03 2nd week of DM. Or bet a closer on Wed and after the day is over figure out it was set very fast for the week! Those produce "bad streaks". // You can very generally expect a certain speed most of the time, but you have to be able to tell when the track is true fast, normal, or slow. // If you do not wish to use my method, at least look for the height of dirt and dust the hooves cause in the stretch and on the turf courses to be kicked up!! The fastest par-rated tracks are often the ones that will suddenly be slower all week or on a weekend day- NY, CA, TX, MD. Some of the biggest biases show up on smaller tracks with less capable crews to correct it, like TB downs this meet or when they try to correct an insanely fast track at one of the Ca fairs and reverse it too much. (SR 02)

[JackS]

Ken- IMO, pars are meant to put you in the "ball park" versus predicting the final outcome of a race. A few ticks up or down this week compared to last week would only tell us that the track has either gotten faster or slower or that the horses actually competing this week are faster or slower than a week ago. Decisions have to be made using the odds board and other Hcp factors. If anyone is aware of drastic changes to the speed of the track, a revised Par becomes reasonable and should be made..

[sjk]

GT,

The reason I use the 6 furlong pars is that all but a hand full of tracks run 6 furlongs and I think that gives a pretty good gauge as to track speed.

I have been puzzling as to how I might equalize track speed over distances when the mix of distances varies greatly from track to track. How do you do this?

[JackS]

sjk- This is another one of those adjustments that may have a few differing opinions. I've been using Cramers(?) 6.5sec per 1/16 mile and 13.2seconds per furlong. Thus- a 110.0 6f time adjusted to 6.5f becomes 116.5....7f becomes 123.2. I use these numbers universally but do understand that closer scrutny of each individual track could produce a more accurate adjustment. When attempting sprint to route adjustments, these two standards can't be used. Edit- These standards also can't be used for route to route adj.

[GameTheory]

My process is kind of screwy.

I have four adjustment factors:

distance-to-distance
reverse distance-to-distance
track-to-track
daily variant


For each track/surface, I choose a standard distance, which is the one with the most numerous races at a particular track. At most tracks this is 6 furlongs, but some tracks it might be something else. All other distances at this track will have an adjustment that converts times at the original distance to the standard distance. (This is the distance-to-distance adjustment, or D2D.)

So each track has a standard distance. One of these is then chosen to be the universal standard track/distance. I use PHA 6f as my universal standard. I chose PHA mainly because it runs all year long, and time of the year has shown itself to be an important factor in my predicted times/pars (I don't really make traditional pars, see below). Plus PHA is a mid level track that has what I call "class overlap" with pretty much any other track. In other words, I can find races with somewhat like conditions (in terms of class level -- purse / claiming price) between PHA and any other track. Comparing races of like conditions is how I get my adjustments. So every track gets an adjustment from its own standard distance to the universal standard PHA 6f. (This is the track-to-track adjustment, or T2T.)

Then there is the reverse D2D, which is a distance adjustment from PHA 6f to PHA *whatever*, the whatever being the original distance of the race. The reverse D2D is used when I want to look at a track-equalized time, but not a distance-equalized one. So if I have a 1 mile race at Santa Anita, I would convert it to 6 furlongs Santa Anita first (the SA track standard), then to PHA 6f (the universal standard), then finally to PHA 1 mile (which is why I call it the reverse, it is the undoing of the original distance to distance adjustment, although not perfectly because it makes it a PHA time instead of whatever the original track was). Now I've got a 1 mile time that I can compare to any other 1 mile time from anywhere else, assuming they all undergo the same process. (If PHA doesn't run a certain distance, then I interpolate from the distances they do run.)

The daily variant is calculated at the level of the original track, of course, and is used along with the other adjustments.

All of these adjustments are multipliers. I simply multiply them by the race time to adjust it. If I leave out the reverse D2D, then I've got a distance & track equalized time (for speed figures), if I apply the reverse D2D then I've got a track equalized time (for velocity ratings, energy, etc.).

So how do I get the actual adjustments? They're all strictly based on condition pars (yes, I know, very flawed). There are 15 variables I use to numerically describe the conditions of the race -- purse, claiming price, non-winners restrictions, race type, etc. I exclude 2 yr old races (and 3yr olds early in the year), maiden races, and a few other things. I make samples like this for each track/surface/distance and use the sample to train a predictor (basically non-linear regression) that can predict times for this track/surface/distance given the conditions. So this predictor is a substitute for tables of par times. The neat thing is I can input a race condition that hasn't actually occured before (exactly) in the sample and it can still output an accurate predicted time. So I build these predictors for all the different distances for a particular track/surface. The one with the biggest underlying sample of races becomes the standard (more samples = more stable = more accurate). So to create the D2D adjustment to the standard I run the conditions in the sample through the predictors of both the original distance and the standard distance, and take the average ratio (I use the geometric mean) between the two predictions as the adjustment. And I only compare races where there is class overlap between the two samples. Then I can throw the predictors away except for the one for the standard distance. (Once I have the adjustments, I no longer need the predictors for the non-standard distances.)

A similar procedure between the various track standards and the universal standard gives me my track to track adjustment.

And so on and so forth...

[sjk]

Thanks for the explanation. I think I see what you are saying.

I am still not clear on how to compare LA with Pha since they have no common distances.

[JackS]

sjk- GT's method is probably the premium answer and should be used if possible. The simpler way is to adjust one of these tracks to the other by using simular time factors mentioned above. After the adjustment has been made, apply the pars. GT's suggestion of usiing one universal track to make opinions for all tracks is superior to trying to make coninual track to track adjustments. The Universal Track can even be a track that doesn't exist. It could be a track that you devise and a track that all other tracks will be compared against.

[GameTheory]

Really? None?

Anyway, I don't require the distances to be common. If the standard distance at one track is 8f, then it simply gets converted to PHA 6f no differently than if it were the same distance. The adjustment is just much bigger, that's all. All I require is that I have some race conditions that are somewhat similar -- then I just take the average ratio between the times and that's the adjustment. So I have one and only one adjustment to convert from one track to another -- and the second track is always PHA 6f.

Once they are all adjusted to the same standard, I can compare them. So if I've got a 8f time at DEL that I want to compare to a 6.5f time at HOL, it goes like this:

DEL 8f => DEL 6f (DEL track standard) => PHA 6f

HOL 6.5f => HOL 8.5f (HOL track standard) => PHA 6f

So now I've got two PHA 6F times I can compare. Of course, I generally would convert them to an integer speed figure instead of looking at 6 furlong-scaled times, but the point is they are equalized (in theory). As I said, the adjustments are all multipliers, so it would be something like this:

DEL 8F time = 98.56

* D2D (Del 8F => Del 6f) = 98.56 * .71635 = 70.60

* T2T (Del 6f => Pha 6f) = 70.60 * .9861 = 69.62


If I wanted to see an adjusted time at the original distance, then I'd apply the reverse D2D:

* RD2D = 69.62 * 1.406 = 97.89

I'd also apply a daily variant, which is always something like .99 or 1.01 (in reality, these numbers go out to 8 decimal places or so). But without that, we would say that 98.56 at DEL for 1 mile is equivalent to 97.79 for 1 mile at PHA.

The way I came up with the rankings that begin this thread is simply to multiply all 4 adjustments together for all the races in the date range at each track and take the average. Multiplying all the adjustments together always gives you a number around 1.0 -- the slowest tracks being about .95 and the fastest being about 1.03 -- which represents their fastness or slowness compared to PHA.

[GameTheory]

BTW --

I'm not trying to say this is the greatest method -- it has many flaws. It is just what I do, and it works ok. The big plus for me is that it is totally automated. I just tell the computer "make adjustments & variants please"....

[Tom]

A problem I have run into in comparing track to track is the class of the horses running the distances. Finger Lakes used to run 6 furlong Gr3 stake each year and the likes of Groovy,and that filly that won the BC sprint set records here. 6 furlong 3year best times were way out of wack at 6 because of one race a year. So when I compare FLt o Bel, I use 6 furlongs to get my track to track. But when I compare FL to Pen, I use 5.5F because the same basic grade of horses run that distance and the comparison was better.

[Que]

GT,

For what it's worth, for comparison purposes, below is my list:

PLN
EMD
FNO
YAV
TUP
GG
SUN
LSB
SA
BM
HOL
ELC
SOL
MED
ARP
ALB
SR
STP
SRP
YD
DUN
KIN
HST
FPX
IND
DMR
SAR
WYO
HPO
MAN
BMF
FER
SJ
RIL
BEL
KAM
CD
BEU
CBY
BRD
AP
RUI
LS
CNL
FE
MTH
STK
PRM
WDS
SAC
GP
OTC
LRL
LA
HOU
RET
TAM
WO
RKM
OP
ELP
GPR
PM
AQU
MC
PHA
TP
LAD
ELK
MOF
FL
KEE
PIM
PEN
DEL
NMP
TDN
HOO
SON
EVD
RP
TIM
LET
HAW
NP
RD
FG
CRC
UN
MNR
MAF
FMT
ASD
MD
MIL
GRP
LNN
SAF
DG
FON
FP
MDA
CLS
KSP
BRN
DED
ATO
SUF
SUD
GCF
CT
EUR
GLD
GIL
GF
YKT
ANF
MPM
FS
CPW
TIL

I've got Charles Town being pretty slow, but not as slow as Flagstaff, Chippewa Downs, or Tillamook County Fair :).

Que.