|
|
05-24-2019, 09:45 PM
|
#256
|
Authorized Advertiser
Join Date: Dec 2014
Location: Oakland, Ca
Posts: 7,953
|
"Regression analysis" as it applies to selecting the winner of a horse race--
how is that not essentially just a fancy term for "back-fitting"?
Last edited by NorCalGreg; 05-24-2019 at 09:46 PM.
|
|
|
05-25-2019, 11:21 AM
|
#257
|
Registered User
Join Date: Mar 2005
Location: Queens, NY
Posts: 20,671
|
Quote:
Originally Posted by NorCalGreg
"Regression analysis" as it applies to selecting the winner of a horse race--
how is that not essentially just a fancy term for "back-fitting"?
|
Good point.
I guess it does a better of finding the right weight for each factor than trial and error.
__________________
"Unlearning is the highest form of learning"
|
|
|
05-25-2019, 12:17 PM
|
#258
|
Registered User
Join Date: Dec 2001
Location: JCapper Platinum: Kind of like Deep Blue... but for horses.
Posts: 5,305
|
Wikipedia - Regression analysis:
https://en.wikipedia.org/wiki/Regression_analysis
Quote:
In statistical modeling, regression analysis is a set of statistical processes for estimating the relationships among variables.
|
Imo, the sentence at the very top of the Wikipedia page on regression analysis says a lot.
-jp
.
__________________
Team JCapper: 2011 PAIHL Regular Season ROI Leader after 15 weeks
www.JCapper.com
|
|
|
05-25-2019, 12:23 PM
|
#259
|
crusty old guy
Join Date: Aug 2003
Location: Snarkytown USA
Posts: 3,944
|
Any type of analysis using past performance data is "back-fitting" to one degree or another, including old school pen-and-paper handicapping. There are many different kinds of regression techniques; choosing the one most appropriate to handicapping or a specific analysis type would obviously be of primary importance. That and using different samples for finding factor weights and then testing the algorithm(s) going "forward".
__________________
"Don't believe everything that you read on the Internet." -- Abraham Lincoln
|
|
|
06-25-2019, 08:59 PM
|
#260
|
Veteran
Join Date: Jul 2004
Location: totonto
Posts: 618
|
|
|
|
07-14-2019, 12:25 PM
|
#261
|
Veteran
Join Date: Jun 2002
Location: near Philadelphia
Posts: 4,560
|
There is a way to determine lengths per second.
There is also a way, as many on here know, how to determine feet per second.
A happy medium could be found doing both type of analysis, imo.
|
|
|
08-26-2019, 01:00 PM
|
#262
|
Registered User
Join Date: Apr 2014
Location: Flint Hills
Posts: 474
|
Apologies for interrupting the cycle here, but after going through all the many pages on this thread I feel compelled to say I REALLY miss Raybo and his contributions.
I'm a combo style. I use my computer as a starting point to construct my multi-factor pen and paper playsheet.
__________________
"Better to do little well than more poorly." Appy
|
|
|
08-29-2019, 07:54 AM
|
#263
|
Registered Loser
Join Date: Apr 2002
Posts: 2,633
|
Goodbye software
I can't recall the exact date,around 1998 maybe, is when I started using software in my handicapping. Probably one of the worst decisions I ever made. Before that time ,doing it my own way through simple deductions, i had much more success. Lately,I've decided to delete all of my software and go back to my old ways. With the state of horse racing in this country,it's probably too late, but it's the way I feel more comfortable now. Now, I skip races. Using software, i was afraid that I might miss a winning selection and bet more races, races I never would bet on my own. Goodbye software.
|
|
|
09-01-2019, 06:39 PM
|
#264
|
Registered User
Join Date: Jul 2019
Posts: 12
|
Quote:
Originally Posted by jk3521
With the state of horse racing in this country,it's probably too late, but it's the way I feel more comfortable now. Now, I skip races. Using software, i was afraid that I might miss a winning selection and bet more races, races I never would bet on my own. Goodbye software.
|
Why do you say that? Sorry if this is a stupid question.
|
|
|
09-02-2019, 09:53 PM
|
#265
|
Veteran
Join Date: Sep 2003
Location: NEW YORK CITY
Posts: 3,670
|
Quote:
Originally Posted by jk3521
I can't recall the exact date,around 1998 maybe, is when I started using software in my handicapping. Probably one of the worst decisions I ever made. Before that time ,doing it my own way through simple deductions, i had much more success. Lately,I've decided to delete all of my software and go back to my old ways. With the state of horse racing in this country,it's probably too late, but it's the way I feel more comfortable now. Now, I skip races. Using software, i was afraid that I might miss a winning selection and bet more races, races I never would bet on my own. Goodbye software.
|
Wait till you buy that Racing Form that you were so used to....
|
|
|
09-13-2019, 08:43 PM
|
#266
|
Registered User
Join Date: Oct 2002
Posts: 149
|
Back-fitting is pure evil.
The vast majority of computer handicappers do it because they get a buzz off it. Two buzzes in fact.
The first buzz is the unbelievable success: a ridiculous strike rate and ROI such as 40% and 150% respectively. The second buzz is adjusting the method to make those results even better either if, or not big losses have occurred.
The truth is: back-fitting is shooting an arrow at a tree and painting a target around where the arrow landed. This is so wrong and yet so many handicappers believe it is the path to riches.
This has gone on for decades and computer handicapping tools have exacerbated the problem.
If you want to avoid back-fitting then only include a filter / method / system rule if it can be justified. Why is it that A Trainer is good in claiming sprints, why is it that B Jockey is good with favourites in the summer months? Think about it first and justify it else reject it.
And make sure you have adequate sample sizes and definitely sandbox the method and use stats to check if the results (pre and post system) are likely to be due to chance or not.
|
|
|
09-13-2019, 08:51 PM
|
#267
|
Veteran
Join Date: Sep 2003
Location: NEW YORK CITY
Posts: 3,670
|
Quote:
Originally Posted by flatstats
Back-fitting is pure evil.
The vast majority of computer handicappers do it because they get a buzz off it. Two buzzes in fact.
The first buzz is the unbelievable success: a ridiculous strike rate and ROI such as 40% and 150% respectively. The second buzz is adjusting the method to make those results even better either if, or not big losses have occurred.
The truth is: back-fitting is shooting an arrow at a tree and painting a target around where the arrow landed. This is so wrong and yet so many handicappers believe it is the path to riches.
This has gone on for decades and computer handicapping tools have exacerbated the problem.
If you want to avoid back-fitting then only include a filter / method / system rule if it can be justified. Why is it that A Trainer is good in claiming sprints, why is it that B Jockey is good with favourites in the summer months? Think about it first and justify it else reject it.
And make sure you have adequate sample sizes and definitely sandbox the method and use stats to check if the results (pre and post system) are likely to be due to chance or not.
|
Let me ask you this....
Suppose you have a set of 100 factors.....and 3 years back data
Say you use Excel's Correlation & Regression Analysis.....
Break data by 2016 2017 2018......
and the same top 10 factors are significant for each year....
What would you do for 2019?
Would that be back-fitting or deductive reasoning ?
Last edited by mikesal57; 09-13-2019 at 08:53 PM.
|
|
|
09-13-2019, 09:10 PM
|
#268
|
Registered User
Join Date: Oct 2002
Posts: 149
|
I would sandbox it until 2020 or 2021. I would wait until the number of *live* runners (and the odds of those runners) was sufficient to decide if it was worth following or not. That process would help in not being dazzled by iron pyrite.
|
|
|
09-13-2019, 09:24 PM
|
#269
|
Veteran
Join Date: Sep 2003
Location: NEW YORK CITY
Posts: 3,670
|
Quote:
Originally Posted by flatstats
I would sandbox it until 2020 or 2021. I would wait until the number of *live* runners (and the odds of those runners) was sufficient to decide if it was worth following or not. That process would help in not being dazzled by iron pyrite.
|
I could be dead by then
|
|
|
09-26-2019, 02:01 AM
|
#270
|
Registered User
Join Date: May 2004
Location: New Hampshire
Posts: 14,526
|
Quote:
Originally Posted by mikesal57
Let me ask you this....
Suppose you have a set of 100 factors.....and 3 years back data
Say you use Excel's Correlation & Regression Analysis.....
Break data by 2016 2017 2018......
and the same top 10 factors are significant for each year....
What would you do for 2019?
Would that be back-fitting or deductive reasoning ?
|
Regression Analysis is overrated, reading the form like a trainer and knowing moves other trainers in the circuit you play will payoff when you know when to ignore the computer at the right times. The real problem is value and being able to say this is not the one (race).
|
|
|
|
|
Posting Rules
|
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts
HTML code is Off
|
|
|
|
|