Preliminary Run of Derby Computer Model

[derby1592]

Below is a first cut using my computer model for picking Derby winners. I think I included most of the likely Derby horses. The percentages will not quite add up because I included over 20 horses and had to run the simulations twice to get all the horses included. The numbers will change a lot after the post-position draw anyway so these are only preliminary.

The computer takes into account all sorts of factors with an emphasis on the Derby profile that I have described in previous posts, which emphases overall development line, current condition, breeding, 3yo preparation and running style. The percentages are generated by simulating many thousands of races and tallying up the results (referred to as \"Monte Carlo\" simulation).

Listed in order of most likely to win to least likely are the horses names followed by what the computer model calculates to be the break even odds. That is followed by the predicted percent chance of winning, place percentage, show percentage, and Top-4 percentage.

I tested this latest model on the last 7 derbies and it did very well with the highest predicted win pct horse winning 4 of the seven races and the second choice winning once (top choice finished second in that one) and the third choice winning once (top choice finished third). With the exception of Charismatic (the model did not pick him in the top 3), the actual winner has always had break-even odds (based on the model) of less than 7-1. Depending on the post position draw, that would indicate that the likely winner will be either Perfect Drift, Harlan\'s Holiday, or Saarland. The odds will obviously dictate the best play. Also, the top 2 or 3 finishers usually had break-even odds of 15-1 or less.

Enjoy and good luck to all. Use or ignore at your own risk. If you have any questions or comments feel free to post away.

Chris

P.S. I personally like Saarland a little more than the model does but I think it may simply be because I have been touting him since last September...

**************
Model Predictions:

Horse, Fair Odds, Win pct, Place pct, Show pct, Top-4 pct

Perfect Drift, 3.4, 23%, 37%, 48%, 58%
Harlan\'s Holiday, 4.5, 18%, 31%, 42%, 51%
Saarland, 5.3, 16%, 30%, 42%, 52%
Request For Parole, 8.8, 10%, 22%, 32%, 41%
Buddah, 9.9, 9%, 19%, 27%, 36%
Blue Burner, 20.1, 5%, 10%, 17%, 24%
Easy Grades, 25.1, 4%, 9%, 15%, 22%
Medaglio D\'oro, 26.2, 4%, 9%, 15%, 21%
Came Home, 26.5, 4%, 10%, 17%, 27%
Essence of Dubai, 26.7, 4%, 9%, 15%, 22%
Private Emblem, 28.3, 3%, 8%, 13%, 19%
Sunday Break, 30.9, 3%, 8%, 13%, 17%
War Emblem, 42.2, 2%, 7%, 13%, 20%
Ocean Sound, 57.7, 2%, 4%, 8%, 12%
USS Tinosa, 66.6, 1%, 5%, 10%, 17%
Windward Passage, NA, 0%, 2%, 5%, 9%
Castle Gandolfo, NA, 1%, 1%, 2%, 4%
Wild Horses, NA, 0%, 1%, 2%, 4%
Lusty Latin, NA, 1%, 3%, 5%, 8%
Johannesburg, NA, 1%, 2%, 3%, 6%
Proud Citizen, NA, 0%, 1%, 2%, 4%

[HP]

It seems weird that Essence of Dubai scores so low. He\'s the leading money winner and a winner at the distance. Also have to wonder what accounts for Perfect Drift and Saarland\'s high scores (and I like Saarland too - it\'s just not based on anything he\'s done...yet). HP

[kev]

Just like you derby1592 i\'ve done my own test running for the past derby\'s. I will be posting my order later and mine is close to yours. My top derby horses this year is SAARLAND and PERFECT DRIFT. I have RFP and D\'oro right up there for 3rd and 4th.

[Dnb820]

Chris,
First thanks for the info and if you wouldn\'t mind.
Where did Inv Ink fit your model for last year\'s Derby.

Thanks again,

DD

[J-DUB]

Interesting data, Chris. Thanks for posting. Is the win% of Perfect Drift in line with previous top choices and the 5% advantage over HH as the second choice? How do you incorporate the post positions into the formula that would make such an impact on final data-- pp win % of derbies, how a certain post flatters or inhibits a horse\'s running style\' etc. Also, do the raw TG numbers ( a 2 vs. a 5 ) come into play or just the overall developement of the line or pattern? Thanks. See ya, jdub.

[derby1592]

I will try to answer all the questions that have been asked so far. A few were on how the model would have done in past years. I summarized that at the end of the message. As you can see, it would have done well in some years and not as well in others but overall it would have performed very well as the top choice ran first or second in 5 of the 7 years.

More specifically to David’s question, as you can see, the model would have pointed to Invisible Ink (and Monarchos) as a great value play.

Jdub asked about how Perfect Drifts win% compares to previous top choices. Things could change after the post draw but right now he looks solid and at the likely post-time odds should be a great value.

HP asked about EOD and wondered why he was not rated higher. The model takes a lot of factors into account and crunches lots of numbers and it is not always easy to predict what it will spit out and why but EOD has several obvious knocks that I know the model picks up on. He ran fast early as a 2yo, which is a negative, he always reacts to efforts, which is a big negative for a Derby horse and he is coming in off a new top at a grueling distance (I am not sure a Derby-distance prep in March is the best way to prep for the Derby itself). His breeding is also a little suspect for the Derby distance despite the flashy names in his pedigree and his high yearling price. Finally, he comes into the race as one of the slower horses so he needs to jump up to win and given all the other factors that seems very unlikely. So he gets ranked in the middle of the pack with about 7 other horses (Blue Burner through War Emblem in the list) that all have very little chance to win but about a 15% chance to hit the board. One of these may very well do just that but most will be at odds that will not be worth a bet. EOD may sneak up into the playable range but I doubt it. Still Godolphin tends to send them out ready and he makes me a little nervous. However, they did not even bother to send any of their other horses to the Derby (they usually bring a couple) so you know he has not beaten much this year in his preps and the stable rider is going to fly all the way out to the US to ride the Oaks on Friday but does not think it is worth sticking around to ride EOD in the Derby on Saturday. Obviously, none of this latter stuff is factored into the model but it gives me more confidence with regards to trusting the model and leaving him off my tickets.

HP also wondered about Perfect Drift’s and Saarland’s high ranking. Both of these horses have classic, Derby-profile condition lines with a fast (but not too fast) race late in the 3yo year (6-10 range) and a consistently, forward-moving line without any real drastic jumps. Saarland is closer to the ideal with a nice winter break and a pair-up several points better than his 2yo top after some time off. Victory Gallop, Real Quiet and Silver Charm had similar lines. Saarland is a few points slower than several but his overall line, near-term pattern, breeding and running style all indicate that he could make a forward move in the Derby. Perfect drift is not quite as strong in any of these areas but his line is still good as is his breeding and running style and he is much faster than Saarland coming into the race. Of the other really fast horses coming in, he is the only one that looks likely to pair-up (based on the factors considered by the model) and a pair-up probably wins the Derby easily. Thus Perfect Drift is pegged as the most likely winner (note that the model still indicates he has a 77 percent chance of losing) and Saarland is not too far behind.

Jdub asked how post-position is factored into the model. Post position and running style are used to estimate ground loss. For example, a front-runner that draws an inside post will likely save ground. The computer simulates thousands of races and the ground loss is randomized for each race (i.e., a horse may grab the rail in one race and in the next he may be forced wide on the first turn) but, on average over many thousands of races, a front-runner that draws an inside post will save more ground than a closer from an inside post and a lot more ground than a presser who is stuck outside in post 20.

Jdub also asked if TG numbers are factored into the model. The answer is yes. They are a key part of the model. In fact, they are the starting point for all the calculations.

I hope that answers your questions. If not or if it just raises more questions, feel free to post away.

Chris

**********
Below are what would have been the model’s top 4 picks over the last 7 Derbies.

Each horse’s name is followed by their actual order of finish followed by their actual post-time odds followed by their model-calculated break-even odds. In theory, you look to play a horse with a reasonable chance of winning (I would say at least 10-20 percent) and with the tote odds significantly higher than the break-even odds (I would say at least 50 percent higher for horses at 6-1 model odds or less and at least 100 percent higher for horses over 6-1 model odds).

2001
Monarchos, 1, 10.5, 3.5
Point Given, 5, 1.8, 5.7
Invisible Ink, 2, 55, 7.2
Congaree, 3, 7.2, 7.3

2000
Fusaichi Pegasus, 1, 2.3, 4.5
The Deputy, 14, 4.6, 4.6
Wheelaway, 5, 20.8, 5.4
Aptitude, 2, 11.8, 6.5

1999
Menifee, 2, 7, 4.7
Stephen Got Even, 14, 5.1, 5.0
Adonis, 17, 18.7, 6.2
Excellent Meeting, 5, 4.8, 8.3

Charismatic 1, 31.3, 31.7 (missed that one by a mile, he was ranked tenth by the model but note how close his tote odds were to the break-even odds...)


1998
Real Quiet, 1, 8.4, 1.3
Halory Hunter, 4, 6.6, 4.6
Indian Charlie, 3, 2.7, 12.8
Victory Gallop, 2, 15, 13.5

1997
Silver Charm, 1, 4, 3.6
Captain Blodgett, 2, 3.1, 5.6
Hello, 8, 9.6, 5.9
Phantom on Tour, 6, 19.5, 6.2

1996
Unbridled Song, 5, 3.5, 5.1
Grindstone, 1, 5.9e, 6.1
Alyrob   , 8, 7.2, 8.0
Zarb\'s Magic, 13, 25.7, 8.4

1995
Timber Country, 3, 3.4, 3.7
Talkin Man, 11, 4, 3.9
Thunder Gulch, 1, 24.5, 6.7
Jumron, 4, 5.6, 8.6

[kev]

Well said about SAARLAND.........

[HP]

On Saarland, if he\'s too slow, I think maybe that rates more weight than the other factors you cite in his favor. He \'could\' move forward in the Derby, as could others. If the big race pre-Derby is something you\'re looking for, he doesn\'t cut it. HP

[TGJB]

Okay, ya got me.  When I get the time(God knows when that will be) I\'ll let you run the studies you want to run.  I agree with everything except \"coming off a new top at a grueling distance\" being a negative- lots of derby winners have.
I would add that those interested in making their own line using their own handicapping and estimated ground loss should try tru-line.  It will probably take at least half an hour to do the Derby, but when you are dealing with so many prices it helps to be able to qauntify your thinking.

[derby1592]

Cool.

I look forward to it.

Thanks

Chris