Weighted Average

[jay68802]

Do you use this for handicapping? Is there any rules you follow to come up with the different weights? I have been messing around with it for years, but never really used a weighted average in any form. Just curious.

[FakeNameChanged]

Originally Posted by jay68802

Do you use this for handicapping? Is there any rules you follow to come up with the different weights? I have been messing around with it for years, but never really used a weighted average in any form. Just curious.

On a homegrown program, let's say you have 10 factors contributing to an overall number in a route. I may weight speed rating at .20 of total, class at .25, recent form at .20, weight change(or total) at .10, etc. So looking at the first three mentioned, they might make up .65 of the total ratings. And within speed ,form or class, I have different weights for last race vs. prev. races. Inside the class rating, I may have 3 different factors contributing to that number and they have different weights. Also in speed, weighting for last race may make up .20 of the total, average of 3 good races at .50, speed improvement at .20, etc. Of course this is all done in Excel.

Like the recent thread on Bris power numbers, sometimes the weightings look wrong. For example, sometimes the last race numbers seem to dominate after looking at the winner, and other times, it's average of last three or last three good races. Or recent form or class is dominating, so it's an exercise in tweaking. Or maybe I didn't understand your question, lol.

[jay68802]

"so it's an exercise in tweaking. Or maybe I didn't understand your question, lol."

You understood the question, so I started tweaking some numbers. As a test I did 9 races, all on turf, for Saturday. Not really interested in how the top picks are going to do in these races. ( I am but that is not where the weighted averages were used). Am really interested in how the tosses perform, or hopefully, do not perform. The races are listed as Picks / Contenders / Tosses.
The picks and contenders are listed in order, as you read them. Will adjust for Scratches in the AM.

Saratoga

Race #1: / /

Race #8: / /

Race #10: / /

Ellis Park

Race #4: / /

Race #8: / /

Monmouth

Race #9: / /

Arlington

Race #7: / /

Indiana

Race 3: / /

Woodbine

Race #9: / /

Total runners 75 / Total tosses 38 or 51% of the runners.

[jay68802]

Correction for Ellis Race #4:
Picks / Contenders / Tosses
/ /

[FakeNameChanged]

Jay, How did your tosses' tweaking work? All turf races, you're a glutton for punishment.

[GMB@BP]

Originally Posted by jay68802

Do you use this for handicapping? Is there any rules you follow to come up with the different weights? I have been messing around with it for years, but never really used a weighted average in any form. Just curious.

I do think there is some merits to this style of handicapping. I know Stats Race Lens allows you to adjust factors but it feels somewhat limited. I wanted to play around with it a bit more but it always felt like no matter what I tweaked the program would still end up with the same selections as the default.

[jay68802]

Picks / Contenders / Tosses

2 / 2 / 5


Two of the tosses that won were Chad Brown runners.

Sitting here tonight looking at what I did and where it went wrong. Used some real basic weights in the areas I did (6,5,4,3,2,1) as a start. Tweaking time. As for only doing turf routes. I looked at my records and turf routes are a weak area. With the increase in turf races I felt that needed to improve.

[classhandicapper]

Originally Posted by jay68802

Do you use this for handicapping? Is there any rules you follow to come up with the different weights? I have been messing around with it for years, but never really used a weighted average in any form. Just curious.

I played around with using weighted averages with speed figures, but I stopped after it showed less promise than I was hoping. A weighted average will probably be able to outperform some simple rules like last race, best of last 2, most recent relevant race etc... but long term I think a subjective analysis of the's horse's races, development, and trips will do better than any formula.

[Aerocraft67]

Isn't this doing manually what a regression would do for you? Not that there's anything wrong with that. In know your question wasn't about formal quantitative modeling. But why not just start with the morning line or some off-the-shelf power rankings? Will manually choosing weights really produce anything novel that a more individual qualitative assessment of the race wouldn't?

Maybe focusing on tosses yields more productive results. Or just use a simple set of weights as a first cut to sort things out.

Thing is, a big dimension of the game is to find the non-obvious, which models would seem to steer you away from. I guess you could build models and weights for specific anomalies. But that's more like filtering for angles, and also not really the thrust of your question.

I struggle with seeking the non-obvious without ignoring the obvious, and re-doing work that's already done aptly for me. Applying manual weights to factors seems to violate the latter problem, unless it's a simple screen. But even then, I worry about tossing the truly clever picks.

I know I'm not breaking any new ground here. Just thinking about my own game in the context of your question.

Put another way, I guess I'm just pointing to this:

Originally Posted by classhandicapper

...a subjective analysis of the's horse's races, development, and trips will do better than any formula.

[coachv30]

The other element here would be assigning the weights to each of your categories / factors. Would they change based on distance, surface, race conditions, etc...?

If you assign a high weight to late speed, is it constant across the board or more for turf routes...etc.

[jay68802]

Originally Posted by coachv30

The other element here would be assigning the weights to each of your categories / factors. Would they change based on distance, surface, race conditions, etc...?

If you assign a high weight to late speed, is it constant across the board or more for turf routes...etc.

The programs I use are 4 different spreadsheets, they all use the same factors. It was easier just to copy the original, instead of starting from scratch. One is for Dirt Sprints, one is for Turf Sprints, one for Dirt Routes, and one for Turf routes. In your example of using late pace or speed, I found from my first try that using a factor that you use to pick contenders, should not be used for picking tosses. I think the reason is probably simple, horses that you are looking for are improving and in those areas (LP) you need to be looking for that improvement and not a reason to toss a horse. As for different weights for different types of races, I do not know yet. Only have been looking at Turf routes but it will be interesting to see the differences. Doing the opening card at Kentucky Downs right now, will post the Picks / Contenders / Tosses when I am done.

[jay68802]

A few for Kentucky Downs.

Same as last time Picks / Contenders / Tosses

Race #4: / /

Race #5: / /

Race #7: / /

Race #8: / /

Race #9: / /

Race #10: / / The rest of them.

Time to have fun with the Ugly Bartender.

[jay68802]

Originally Posted by jay68802

A few for Kentucky Downs.

Same as last time Picks / Contenders / Tosses

Race #4: / /

Race #5: / /

Race #7: / /

Race #8: / /

Race #9: / /

Race #10: / / The rest of them.

Time to have fun with the Ugly Bartender.

Updated for scratches.

[coachv30]

Originally Posted by jay68802

Updated for scratches.

I put together a weighted spreadsheet and did KD Race 1..

2,3,10 / 4,6,9,11 / 1,7,8,12,13

Same format- Picks / Contenders/ Scratches

[coachv30]

Originally Posted by coachv30

I put together a weighted spreadsheet and did KD Race 1..

2,3,10 / 4,6,9,11 / 1,7,8,12,13

Same format- Picks / Contenders/ Scratches