|
JCapper Message Board JCapper 101
|
| Home |
Register
Log In |
| By | Building a UPR | |||||||||||
| Cole 4/18/2008 4:38:29 PM | I am fairly new to JCapper and this is my first attempt to build a UPR. My question is how do I build a UPR and utilize more than 1 factor in the UPR. I can build a UPR but how do I add factors using the Wizard, ie. Save, Next, NEW or what? Thanks PS. I probably need step-by-step directions.
| |||||||||||
| jeff 4/18/2008 7:33:28 PM | I'll present the steps for building a VERY SIMPLE two factor UPR for route races. To add more factors you simply run data for another factor and create a set of table entries for that factor. IMPORTANT - before we get started: make a backup copy of the following file: c:\2004\JCapper.mdb That file contains the Default GroupName table entries that drive JPR. Never under any circumstances (unless YOU want to change JPR to something other than what I gave you) make table entries to the DEFAULT GroupName. Step 1 - Run some Data Window studies. As an example I've run all route races through my calendar year 07 database. It looks something like this:
And if I break that out by two factors, it looks something like this:
Step 2 - Launch the IVTable Wizard. From the Main Module click System Settings followed by Impact Values Table Wizard. Step 3 - GroupName - Decide on a GroupName for your UPR. In your case, if you want to create a new UPR maybe something like COLEUPR1 would work. If you want to create your first UserFactor then use USERFACTOR1. I'll illustrate how to create a new UPR. You would use the same procedure to create a new UserFactor - the only difference would be the GroupName used. Click the NEW button on the Impact Values Table Wizard. Key COLEUPR1 into the GroupName textbox. Step 4 - table entries for a Factor Select a factor from the factors drop down. In this case we'll use COMPOUNDLATE since that's one of the factors I chose to run. Click the Active checkbox. Behavior - Key a 0 (that's a numeric zero not the letter O for the Oscar my last girlfriend deserved when she convinced me it was time to get the kitchen redone) into the behavior textbox. 0 means rank and I'm using that because that's how the data is presented in the sample I ran. Weight - Pick a numeric weight for your factor. Ultimately factor weight represents the amount of importance YOU are assigning to any one factor in your overall factor mix. But let's keep things simple here and just enter 40 into the textbox for weight. Track - Just like you would for a UDM you could enter track code(s) if you want your UPR to require specific tracks. Example: HOL-SAX-OSA-DMR etc. But in this case just leave the track textbox blank. AvoidTrack - Just like you would for a UDM you could enter track code(s) if you want your UPR to avoid specific tracks. Example: HOL-SAX-OSA-DMR etc. But in this case just leave the track textbox blank. Surface - Enter a * (or wildcard character) for surface... unless you want your UPR to only react to a specific surface - in which case you'd use: D* all dirt D main or outer dirt course only d inner dirt course only T main or outer turf course only t inner turf course only A articifial surface only MinDist - I'm illustrating a route only UPR so key a numeric 8 character into the MinDist textbox... as in 8 which is the number of fulrongs in a 1 mile race. Note: If you were creating a UPR for sprint races you might want to use -999 as your min dist. MaxDist - I'm illustrating a route only UPR so key 999 into the MaxDist textbox... as in we don't care how far they run. Note: If you were creating a UPR for sprint races you might want to use 7.9 as your max dist. Class Descriptor - If you want your UPR to react to specific Class Descriptors you would enter Class Descriptor codes just like you would for a UDM (C-M-A-G, etc.) But in this case we'll have the UPR react to all Class Descriptors so just leave it blank. High Score - LOOK at your data. In this case the top impact value for CompoundLate rank 1 is 1.5901 so key that into the HighScore textbox. DefaultImpact - Again, LOOK at your data. In this case the impact value for CompoundLate rank 12 is .2349 so key that into the DefaultImpact textbox. The interface will use that as a default IV for any value you fail to specify when you key in individual IVs. Matched Pairs - The process for entering matched pair data is the same whether it is for rank 1 or rank 20. You are entering a number into the RANGE field which describes a data range. You follow that up by entering the impact value for that range into the IMPACT field. Example: For CompoundLate rank 1 you would enter the following matched pair data:
For CompoundLate rank 2 you would enter the following matched pair data:
Do this for as many matched pairs (ranges) as you have valid data for. I generally do it for ranks 1-12 and leave it at that. Generally the deeper you go the stronger/more accurate your UPR will be. I get far better results going 12 deep than I got going 8 deep when I first started out. Until you know better use the Impact Values right from your Data Window output. Once you have more experience you can use your human powers of observation to interpolate a little bit... for example if rank 9 has a higher IV than rank 10 which is a likely occurance when you don't have enough data. After you have entered matched pair data for ranks 1-12 look it over for typos. Then hit the SAVE button when you are satisfied that you have keyed the correct numbers. At this point the table entries for the first factor in your UPR is complete. Note: This post is getting rather long so I'll end here and then come back and add more as time allows. -jp .
| |||||||||||
| jeff 4/18/2008 7:43:59 PM | Adding a factor to a UPR Now it's time to add the second factor to your UPR. Click the NEW button on the Impact Values Table Wizard. Key COLEUPR1 into the GroupName textbox. Step 4 - table entries for a Factor Select a factor from the factors drop down. In this case we'll use COMPOUNDE1 since that's the other factor I chose to run. Click the Active checkbox. Behavior - Key a 0 (that's a numeric zero not the letter O for the Oscar my last girlfriend deserved when she convinced me it was time to get the kitchen redone) into the behavior textbox. 0 means rank and I'm using that because that's how the data is presented in the sample I ran. Weight - Pick a numeric weight for your second factor. Ultimately factor weight represents the amount of importance YOU are assigning to any one factor in your overall factor mix. But let's keep things simple here and just enter 60 into the textbox for weight. Notice that 40 was the weight used for late. And now we're using 60 for early. What I'm expressing here is something along the lines of: --quote: "I think late represents 40 pct of the total handicapping puzzle and I think early represents 60 pct."--end quote We can go back in and edit these numbers afterwards. But for now our UPR is 40 pct late and 60 pct early. Track - Just like you would for a UDM you could enter track code(s) if you want your UPR to require specific tracks. Example: HOL-SAX-OSA-DMR etc. But in this case just leave the track textbox blank. AvoidTrack - Just like you would for a UDM you could enter track code(s) if you want your UPR to avoid specific tracks. Example: HOL-SAX-OSA-DMR etc. But in this case just leave the track textbox blank. Surface - Enter a * (or wildcard character) for surface... unless you want your UPR to only react to a specific surface - in which case you'd use: D* all dirt D main or outer dirt course only d inner dirt course only T main or outer turf course only t inner turf course only A articifial surface only MinDist - I'm illustrating a route only UPR so key a numeric 8 character into the MinDist textbox... as in 8 which is the number of fulrongs in a 1 mile race. Note: If you were creating a UPR for sprint races you might want to use -999 as your min dist. MaxDist - I'm illustrating a route only UPR so key 999 into the MaxDist textbox... as in we don't care how far they run. Note: If you were creating a UPR for sprint races you might want to use 7.9 as your max dist. Class Descriptor - If you want your UPR to react to specific Class Descriptors you would enter Class Descriptor codes just like you would for a UDM (C-M-A-G, etc.) But in this case we'll have the UPR react to all Class Descriptors so just leave it blank. High Score - LOOK at your data. In this case the top impact value for CompoundE1 rank 1 is 1.4340 so key that into the HighScore textbox. DefaultImpact - Again, LOOK at your data. In this case the impact value for CompoundE1 rank 12 is .3153 so key that into the DefaultImpact textbox. The interface will use that as a default IV for any value you fail to specify when you key in individual IVs. Matched Pairs - The process for entering matched pair data is the same whether it is for rank 1 or rank 20. You are entering a number into the RANGE field which describes a data range. You follow that up by entering the impact value for that range into the IMPACT field. Example: For CompoundE1 rank 1 you would enter the following matched pair data:
For CompoundE1 rank 2 you would enter the following matched pair data:
Do this for as many matched pairs (ranges) as you have valid data for. I generally do it for ranks 1-12 and leave it at that. Generally the deeper you go the stronger/more accurate your UPR will be. I get far better results going 12 deep than I got going 8 deep when I first started out. Until you know better use the Impact Values right from your Data Window output. Once you have more experience you can use your human powers of observation to interpolate a little bit when rank 9 has a higher IV than rank 10 which is a likely occurance when you don't have enough data. After you have entered matched pair data for ranks 1-12 look it over for typos. Then hit the SAVE button when you are satisfied that you have keyed the correct numbers. At this point the table entries for the second factor in your UPR is complete. To add more factors just keep repeating this step. Of course as you add more factors you may want to re-weight them according to how important you think they are as part of the whole. -jp .
| |||||||||||
| jeff 4/18/2008 7:53:13 PM | Ok. That's the table entry process for creating a very simple 2 factor UPR. Next it's time to evaluate the UPR itself. Any time that you create a new UPR or make changes to an existing UPR, the next step is running a Database Build From Scratch on a target folder. After running a Database Build From Scratch on a target folder, THEN you can use the Data Window to evaluate the performance of the UPR. This is where the FUN begins. After evaluating the UPR try changing something about it... start with the factor weights. Pick ONE factor and increase the weight. Then run a Database Build From Scratch and look at it again. If your UPR improved take note of what you did. Maybe you want to increase the weight for that one factor even more. If your change weakened the UPR from what you originally had, maybe you want to drop the weight for that one factor from what you origianlly thought. It can be a long drawn out process. But if you keep at it you can create some VERY powerful ratings. And the really cool thing about this is that YOU are the ONLY one on the planet that has access to what you create. -jp .
| |||||||||||
| Cole 4/19/2008 6:12:46 AM | Jeff, Thank You very, very much. I sure do appreciate it. You have been more than helpful to me. Bill
|
| Reply |
|
Copyright © 2018 JCapper Software back to the JCapper Message Board www.JCapper.com |