Powerful System soon to be seen at Our-PM: Midnight's LuLu system!

[midnight]

(Note: In case anybody just fell off the boat, the following is a joke and not meant to be taken seriously)

Yes, that's right. I'm going to market a system at Our-PM, which I proudly refer to as "Midnight's LuLu system." This system spots more longshots than you could possibly believe! It will pick you winner after winner and net you a fantastic ROI! It's easy to use. In fact, you only have to remember two rules! The system would be a bargain at $1,000, but I'm going to sell it for a mere $45.

Since I know that you guys are all honest, I'm going to post the system right here for everybody to see. I trust all of you to send me $45 once you've gotten rich from it.

Here is ths system:

1) Ignore all steeplechase races of a mile or less in distance, races with 20 or more first time starters, races with purses of over $20,000,000, and races carded exclusively for horses 10 years old and up.

2) Play any horse whose name begins with "Lu".

And that's it! Simple, yet effective.

What's that? You say you want proof.

I'm not going to post a full workout, but here's a 45+ day workout, based on my HTR database of all North American tracks except Fonner Park and Blue Ribbon Downs from January 10th through February 25th.


Plays: 242
Winners: 39
Win Pct: 16.1%
ROI: 1.14


Yes sir: a 15% profit on your money! Now don't forget to send me the $45 once you get rich. My address is:

Our PM
1234 Sucker Lane
Anyfool, MN 55427

The purpose of all that? To show you that any system---even an obviously random one---can hit reasonable winners and a good ROI over an extended period. I'm sure that if an HTR user with a bigger database than mine gave this "system" a more extended workout, it'd show a nice fat loss.

[formula_2002]

Midnight. You forgot about the free cornet lesson and uniforms for the kids.

Joe M

[hurrikane]

you system guys are all alike.

2yrs
2739 bets

5478 bet
3479 won
0.63 ROI

you know..if you had said horses whos names start with 'Lucky' you would have been on to something.

Wait 'till Larry see's this..you're in for a reaming. :D :D

[hcap]

Midnight

qoute
"to show you that any system---even an obviously random one---can hit reasonable winners and a good ROI over an extended period."

I totally agree.

When I first started, I used lotus 123 for dos v4. I downloaded 3 part bris files and converted them to the single file format.
I soon discovered that I could make all sorts of "composite" factors. I proceeded to combine all sorts of stuff-whatever struck my fancy looking for the hot one.
Probably made over 50 strange new factors. If I had the jocks middle name, I betya I could have combined that with whether or not the horse took a dump (if I had that info), and come up with something-as long as I had a few more factors to judiciously throw in.

I could always find something that seemed to work. At least on past data.

If your dilligent enough, and are willing to mess around long enough trying anything, you to can find the perfect system.
System creators and sellers rarely try new data or if they do they find a few situations where their system appears to work.
In overmanipulating back data and then searching for the perfect spot, they are super guilty of backfitting.

[formula_2002]

Hcap, back fitting is an easy matter, and yes I can produce hundreds of "systems" buy back fitting.

However, I know of no other way to attempt to produce a profitable system without first back fitting.

The method to determine a profitable system is by comparing expected events to actual events and looking for statistical significance. Back fitting is part of the process...

Any other thoughts?

Joe M

[hcap]

Joe

No question, looking at past data is the only way. Even the successful non-computer guys have to learn from past stuff and develop some sort of "rules"

My comments were in support of Midnight's observation, and I was trying to illustrate how most "systems" come to be, particularly the ones that are usually for sale!

By using typically, completely non logical (even logical) composites, anyone can strike it rich on past races. By mathematically combining various factors, or querying on lots of individual factors on a large sample you can cherry pick your way to eroneous conclusions.

Statisitical tests can help, but the only way I have found is to try your conclusions on new data.

And then the game changes, and what may work on new stuff has a tendency to go stale.

The only methods that worked that I have found on very large samples, tend to be near break even.
Beats the track take but very difficult to profit by without a large bankroll.

Smaller samples whether they be track specific, race type specific, etc., are easier to deal with but also have to be tested on new data.

Now this is computer database stuff and not using human judgement, except for the initial setups.

Ironically, as our computers and software have become more powerful, the trap of over analysing past stuff has become easier.


My comment was also an aside to "Our-PM", and perhaps also Tom C.

Truth in advertising would be appreciated