STATISTICAL SIGNIFICANCE

[barman]

A question for the mathematically inclined...........I have been working on a method of eliminating false favorites and identifying live favorites which has shown great results in an admittedly small sample. The key HTR factor is Ped=1, along with exclusion factors for layoff, finish in last race, today's post position and class.

So, this week I decided to put my money where my brain is and began betting the method with real money in the real world. I use a money management system which varies my bet size based on previous wins or losses, and the results for the week, up to my last bet (24 win bets & 24 place bets) are as follows:

Win %: 66.6%
Average Winning Price: $5.16
ROI: 73.6%

Place %: 70.8%
Average Place Price: $3.25
ROI: 29.7%

My question is: At what point do I conclude that these results are valid and reliable for future investment? Are there actual formulas that can be used to do this? Thanks for the help.

Barman

[Donnie]

Barman--
I would think you will need to test it against a larger sample size. Short of you posting the parameters here, Ken would probably run it thru his db of races and tell you what future expectations may hold. How much have you tested it before you went live? Did you test it over different "seasons"....I do mean Spring/Summer/Fall/Winter? Larger tracks vs. smaller tracks?
Some people have been known to split their db's into two "segments". One for testing. One for "moving forward". If you draw good results from the "moving forward" races, then you may have a nice spot play. But always use a large sample size. Those are some nice numbers, but that win % is gonna hafta stay up there with the average mutuel it's kicking out. The 70% place would possibly lead one to be thinking about keying these horses in the exotics!
Hope it pans out for you! Let us know how you do with it!
BOL!

[overdog]

Hi Barman

I would be willing to tackle the question but need to know a lot more...namely specifics:

How many races, qualifiers, actual winners etc? How many distances?

Then, is it broken down by track or all N. America?

overdog

aka Fraser

[njcurveball]

Although I am probably in the minority here, I would say to keep going with what you have.

Keep an eye on the losing streaks, especially at this very high percentage. Also keep an eye on 2nd and 3rd place finishes.

Simply put, when you are winning at 66%, you have a 33% chance of losing or 1/3. So losing 2 in a row is 1/3 * 1/3 or 11.1%. 3 in a row is 1/27 and so on. You can see that you should have a very low probability of losing streaks.

If you find you are losing 4 or 5 in a row, then it is time to regroup, since you probably don't have a true 66% system.

Also, there are simple ratios depending on odds. At an average of 3-2, you should win more than you place and show. But if you find a ratio of 3-1, (3 wins and only 1 finish where they are 2nd or 3rd), you are on a good steak, but probably not a statistically valid one. This is the "red flag" I see in your stats.

One more thing. If you are using ranking factors than make sure they have a linear relationship. #1 wins more races than #2, etc.

After this start if you "protect the downside", you should be way ahead even if it doesn't pan out. I would be cautious since I have yet to see a long term system that can generate that high a win% and combine it with that high an ROI.

congrats on your success,
Jim

[km]

Barman, if it's a specific spot play with defined parameters, list it here, i'll run a test for the last 365-days and let you know it's long term value.

If you don't want to give away your play, i can guess at it with PED=1 combined w/ such a high win rate, and it would probably be positive on future outcomes; such as combining with K110+.

PED=1 + PED 450+ have many similar applications with positive results, but high win rates are not the goal; bombs are the profit makers.

[barman]

As an update........

I may not have made myself perfectly clear in my original post, but I am only betting the post time favorite if all criteria are met.......since my original post I have had two more qualifiers, both of which won at short prices.

The longest losing streak in the win slot has been 2 races, and the longest losing streak in the place hole has been 1 race, this out of a sample of 26 bets this week.

I hate redboarders, and this was not the intent of my post.....as an example there are 3 more possible bets this evening, the only one of which I perceive to be a possible favorite, and therefore bettable.......Sam Houston Race 7 #3. The results for today are 7 bets...5 winners...1 place...1show.

I think that I will take your advice and continue to bet but proceed cautiously until I have a larger sample to verify my results. Thanks for all the positive comments and helpful hints.

Barman

[NC Tony]

Ken,

In discussions I had with Mall, I set up a series of spot play parameters. I have figured out how to check the results of that specific days spot plays. Is there a way anyone of us can do what you volunteered to do for Barman (ie run a query against the past 180 or 365 days lets say?) using the robot only? Will you consider developing that capability in the robot to handle something like going forward? Otherwise you are either forcing me to upgrade my Microsoft Access skills, or pay Dehere to do it for me. (LOL)


Also, just for kicks does the Odds button work at all in the program?

Thanks

NC Tony

[Donnie]

Barman...if those are your run-outs then I would suggest a parlay, either 2 or 3 race with a "take-down" goal each day. Say you wanna make $200 a day....$50 bets would get you there in 2 bets in many cases, and all you invest is $50. If you're patient, and it continues to go forward, it's better than money in the bank!

[Rick]

The Odds button works with ATRpro.

ATRpro has to be running to save the odds to file so that HTR2 can read them.

[km]

BARMAN
i may not have all your parameters correct, but i ran the following for 365-day; purse $10,000+

Tote Favorite
PED = 1

Plays: 7269
Wins: 2615 (36%)
ROI: 0.81

This improves an "all burger" chalk play (33%, 0.78) but doesn't make any money in the long run. Anything to do with favorites cannot sustain, but obviously can get hot for certain periods with such a high overall win rate. You would do much better by focusing on K110+ or HF"s that are chalk, many of which are also PED=1 but will win 40% and more with 0.90 or better ROI.

NC TONY
90% of the tests you see from me on here or in the Newsletter can be done with the Robot. The others are programmed in my own db app, but most can be duplicated with Access and the export.

I keep 12 months in a folder called "365". This contains the Racefile+Results+Chart (and now the Trainer text files) for the last 12 complete calendar months. At the end of this month, i will completely delete all the November 2005 files from "365" folder and then copy the entire set of November 2006 it its place. This gives me a rotating 1yr sample to test at all times. (the results above were for Nov 1, 2005 thru Oct 31, 2006).

If looking for substantial long run results and solid trends =

I used to run 2yr tests, but no more, horse racing is extremely dynamic now and shorter samples are more relevant. But it is also important, as Donnie mentioned, to get a full 1yr rotation to make sure that all the age/class cycles that occur througout the year are included.

You can't bank on the "all burger" (365-day test) for the most part, but you can get excellent trends and critical information to begin further research. It's a head start and part of the essential knowledge base that a winning handicapper needs.

Looking at an individual track, or even a specific meet is critical to feedback and realtime profit though. We exist in an era of such rapid fluidity in horse race outcomes and stats. The Robot is an important tool - we don't want to get complacent and lazy because there are plenty of other players digging deep in the same mine, and they'll bury us if we fall behind.

[NC Tony]

Ken,

Thanks for the reply. I have access and have some fundamental knowledge on how to use it, but am very good with excel etc so I don't think it will be an insurmountable task, to learn how to run the queries. What are the file extensions to distinguish between the entries, results,charts etc as I think I might try to dabble and play around with the data.

Thanks a lot!!

NC Tony

[MikeDee]

The easiest way to get started with access is with the access user guide and the template DB. They are on the web site and you can download them.

[hurrikane]

hmmm..

interesting play. I am assuming the 78% ROI = 1.78

if that is the case the first thing I would do, if winning 2 out of 3 win bets is get rid of the place bet and put all the money on win. This would maximize your return.

chances are if you are betting the fav you are not going to hurt the odds.

just a simple mans opinion. good luck going forward.

[tbrown]

Having a 36% play that returns poorly is still nice to know for pics, where the price is on several races, not one.

[barman]

For all those that are interested here are the possible plays for today:

tTRK nRACE tPGM tDATE
BM 6 5 11/25/2006
CRC 12 5 11/25/2006
DED 1 2 11/25/2006
DED 3 7 11/25/2006
DED 6 6 11/25/2006
FL 5 2 11/25/2006
FL 6 6 11/25/2006
HAW 7 7 11/25/2006
HOO 5 3 11/25/2006
HOO 8 5 11/25/2006
HOO 10 7 11/25/2006
LA 3 1 11/25/2006
PHA 1 4 11/25/2006
PHA 4 7 11/25/2006
PHA 5 7 11/25/2006
PHA 7 6 11/25/2006
RP 2 3 11/25/2006
TDN 5 5 11/25/2006
TDN 9 8 11/25/2006
TDN 13 8 11/25/2006
TUP 2 5 11/25/2006
WO 2 8 11/25/2006
WO 5 8 11/25/2006

Remember, these horses are only bet if they are the post time favorites. I apologize for the redboard on the first bet (a winner at PHA), but my son was asleep in the room in which I keep my computer. Best wishes to all.

Barman

[km]

Tony, all our files are notated with letters to distinguish them

B = racefile
C = charts
S = trainer stats

The extension on the "C" and "S" files is .txt

1120Bxx.CD = race file for CD Nov 20; xx = race numbers that day, each race has an individual file.

After downloading the Results file, the Scratches, Prices and 1-2-3 Finishers are embedded into the orginal racefile. So you don't need to keep downloaded Result ("Q") file.

CD1120C.TXT = charts for Churchill Nov 20

CD1120S.TXT = trainer text file for Hollywood Nov 20

[Donnie]

Barman--
Looks like a number of your horses ended up either scratched or not the posttime favorite. Are you seeing the prices depressed over the weekends, the way they were this past weekend? Odds-on horses are a tough way to make money, but you said that the AVG Mutuel prior to your last post was $5.16. It had to be much lower this weekend as nearly every winner was odds-on.
BOL!
-Donnie

[AwolAtHTR]

Mon 26 Nov 6

hi BarMan,

I want to bet your spot play starting Wed 29 Nov 2006.

in other words, I think your method sounds GREAT. With a solid set of history, the Exotics could also be bet. However, I believe that if other HTR players would also do REAL bets then we would find that some bets are different because SOMETIMES the chart favorite is defined by the last dollars into the pot. Of course those last dollars will always be unknown. So the player with the best answer for '..who will be the favorite?..' will likely have the bets that make money.

Well, AWOL says watch the Exacta money and you can predict who will be the favorite when the tote board is sending confusing messages about who will be the favorite.

The Exacta money will SOMETIMES tell you which of two six dollar (2:1) favorites will be the ranked favorite for the chart. When there are three chart favorites, the decision is more difficult (ie, flakey). Almost ALL clear favorites (will pay is under five dollars) are easy to find using the Exacta money.

Drop me a note if you want more about the Exacta money or I could post here.

Looking forward to giving this shot.

Duane

[Donnie]

Welcome back AWOL!

I don't know if Barman is gonna tip his hand.
But if you are serious about betting chalk or near favorites, then I'll steer you in the direction of an Ebay auction I have going on right now that you may be interested in. I added a couple of restrictions and it chugs right along. It did pick 2 $22 horses within the last 3-4 days tho as well, so it's not all chalk. But I have the results in the bottom of the auction. You can see what it did over the past 3 days at CRC. I don't know that I would of played that last race today at CRC...probably a chaos race in HTR...haven't looked.

http://cgi.ebay.com/ws/eBayISAPI.dll?ViewItem&ih=015&sspagename=STRK%3AMESE%3AIT&viewitem=&item=250053286007&rd=1&rd=1

(I don't mind being a shill for my stuff! After all it is Xmas...better to "give" than to receive, right? Not to mention the young one wants a PS3!!!):D

[barman]

Donnie,

Based on suggestions by you and other helpful members I did a little digging to find out if the results were being affected by the day of the week.

You were right on.....the majority of the big numbers were all gained on Mon-Thurs, with the winning % only slightly depressed on Fri-Sun... but the mutuels were miserable on the weekends.

So, although I do not possess a large database, I checked all currently running tracks for the last month, and I find the results to be slightly encouraging, because the ROI still remained positive at 1.25 for all days/all tracks.

Then I broke it down into individual tracks, and much to my amazement the results were incredibly aberrant, with results ranging from an ROI of .43 (ZIA) to 2.01 (TUP). Therefore, I concluded that I should restrict my bets to only the positive return tracks, of which I found 10 currently active. Based on that premise I only placed two bets today, both at FL, both of which won, with an ROI of 2.68.

So, without being picky or ungrateful, here is my question...Would you go with the all-burger to pump up the rebate, or would you simply increase the bet size and zero in on the more likely, positive scenario? All member comments are gratefully appreciated.

Barman

[Bob]

I added a couple of restrictions and it chugs right along. D

Hey Donnie! I helped out a little, placing a couple bids, those PS3's aren't cheap, hehehe!

Now it is at a price that, before I go further: Is the method fairly easily programmable into an update query in Access, using our data? I say fairly easily as I am just learning Access, thanks to MikeD's help.

Thanks in advance. Happy Holidays Donnie.

Best,
Bob

[Donnie]

Bob-
I use the HTR data instead of the Beyers that they use. Still works great. And there is one other piece of info in HTR I swap out, but again, there is no loss of accuracy. I have been thinking about programming it into access, but I will need someone to do the module for me, because I am not quite there yet with programming VBA...getting closer all the time. There are very few decisions that need to be made! Anyone with basic handicapping skills can manage thru them effectively. It takes about 2 minutes per race. Last night I used it for the MNR races for the PRM contest. Picked 2 winners and a 2nd place out there, out of 3 races. Those weird numbers from MNR stuck out like a sore thumb....was strange seeing a horse win and get a cramer of 10!!

Thanks for your bids! My kid appreciates it! (BTW....what is your ID there if I may ask? I can be rooting for you as it comes down to crunch time!! :D )

[km]

Donnie

If this book's method or system has specific parameters, list them for me and i'll program it for you into a custom Robot option. You can pop up the plays every day and test it as well. If it shows promise, i'll keep it permanently or you can post the plays on the bbs.

But if it has vague rules like most older systems -- "...accept a 3yr that ran poorly last time if it switches to the meets top rider ..." -- no thanks!

[Donnie]

And I put a reserve on it of $50.....I have had some great success with it, but never imagined it would go for what it is at right now and we still have a day and a half to go!!

No vague rules at all...3 very simple rules (or should I say calculations). If you can manage a pocket calculator, I would imagine that all you need is a data source (even a track program has everything you need in it) and for about 2 minutes a race, you "classify and rank" the horses. Many times when it misses the winner, the top 2 end up being the 2nd and 3rd place horses! This could be a show better's dream!

Ken....I will be in touch with you by email. The calcs are so easy, a person can memorize them within a matter of a couple races. Wish I had a dozen of them to sell!! LOL!!

[njcurveball]

I get letters like this weekly. LOL!

3 simple rules, do them in your head, use a pocket calculator, etc.

I hope you guys know how this sounds. :D

[njcurveball]

OH MY!

I just looked at the current price! I see your point now! I would contact the Author and get the Printing Presses rolling again!

[Donnie]

Jim--
I assume you looked at the results on that auction......
Those are not made up. And no, I did not hunt around till I found a track that would kick out those types of returns...I just decided to do CRC cause the weather is crappy everywhere else and I figured I could get 2 straight days of fast tracks there. I tweaked one rule that results in the throw out races....I require every horse in the race must run 3 races. If even one has not, pass.

Suddenly I feel dirty. Do all system sellers feel this way?:eek:

I am sure it is a feeling that will pass though!

[Donnie]

Sorry to have hijacked your thread Barman....

Back to your question.....

Why can't you do both? Pump up the bet sizes on the "good looking thangs" and run the bets thru the rebate houses? What are their rules on odds-on....no rebates? So on some you lose the rebates, on others you collect the rebates. Consistency and high hit rate are mandatory on that type of play! If you see that the weekends are "softer" stay away from them. Those are your "days off".

Others may disagree, but that is the approach I would take!

[Donnie]

Hey Bob---
Thanks for bidding that manual up!! I appreciate it! What I always try to figure out is where the snipers are sitting...you know the people who bid with 3 seconds left! That price has to have many more people wondering if it is legit!
I think I owe you a drink now next July in Vegas!!!

[AwolAtHTR]

well Donnie

I took liberty to start the thread: http://www.homebased2.com/forums/showthread.php?t=5959

as you said, let us get back to Barman's spot play

duane