Hi Everyone,

Have just finished a test of Arlington Park for the top four picks of the profiler.

Testing Procedures:

First build an untuned database for the track you are testing for all the data files you have available (and continue to add any new files downloaded). Once completed, then connect to the database once Equisim has been opened and click use simulation biases off.

Next before opening any race cards, go to Profiler database>profiler preferences and click on the configuration tab and set the default setting (by placing a check mark in the following fields) to track, dist. Cat., surface, no off trks, and class. Also I have checked the box use date range and have set the dates to the time period of the first calendar month of our data (in my case this is 8/1 to 8/31). For testing newly opened tracks, one must be careful early in the meet to realize limited data is somewhat questionable in profiler analysis.

Now I open the next day's card, which is September 1st, run all simulations and then profile all races (using the simulation and profiler options in simulation and profiler database dropdown menus). Now that I have the export html function, I can export the profiler information using single file, brief mode and sort on profiler value (check these boxes) for our testing.

I decided to track the success of the top 4-profiler picks. Using the browser portion of Spat to determine each race winner and payoff can easily complete this.

After completion of the day's races and prior to next day testing, reset the data range in profiler preferences by adding the last day tested. This way the profiler will be testing on all the prior race cards as each day is tested.

From preliminary testing of my own and Nathan's latest post on profiler testing, it appears the profiler maybe the key to long term profits.

Results:

Choice # of bets wins Win % Return ROI Avg P/O Largest
Top pick 294 84 28.6 519.60 .88 6.19 23.60
2nd pick 294 61 20.7 488.60 .83 8.01 21.00
3rd pick 294 48 16.3 645.80 1.10 13.45 84.60
4th pick 294 28 9.5 374.20 .64 13.36 28.60
Total 1176 221 18.8 2028.20 .86 9.18 84.60

Note: Testing was from 9/1/02 through 10/14/02. ROI breakeven is 1.00

Will post my thoughts on the results sometime tonight, need to head to work now.

Hi Everyone,

Here are my thoughts or conclusions regarding the profiler testing John and I have just finished.

1) Based on average field size of Arlington (7.88) and Mountaineer (9.34), at random any pick should win at the following rates respectively: 12.69% and 10.71%. Thus playing the fourth pick should be eliminated from consideration because it performs worse than random. ROI of the fourth pick was also well below the takeout rate.

2) Based on the win % of the top 3 profiler picks, an odds line could be derived and used as a cutoff. This should improve ROI and eliminate many under lays. However, realize that win percentage will certainly decrease because odds do have a direct correlation to actual win rate.

3) Other filter setting may improve the performance also, but is very time consuming to test at this time.

4) Nathan's pace line selection method is strong, but picking ones own pace lines may improve the results.

5) It may also be possible to improve results by playing the two longest of the top three selections, again another thing that could be certainly tested.

6) Adding other handicapping factors, such as, prime power rating, any current biases, or other factors utilized by the handicapper should also result in elimination of some false plays since this test is playing every race.

In closing, I am sure many of you will come up with other ideas to improve these results. But all and all, I feel the profiler may lead to the pot of gold at the end of the rainbow.

Bill

Hi guy's i use the profiler unlike jim i use the top 4 horse.For last week bris contest 7 of the 10 winner's were in top 4 which is not bad most time it is 50% at any track.Then use horse fraction\ split and converted fraction this help to find the winner in those four horse.Oh ya the filter's i use are race shape\spd index\surface\class.Granted it is loose but i feel i want a general idea this work's for me and hope help's someone.Have a great day

BigHit -- I'm surprised that the top 4 Profiler horses have only predicted the winner 50% of the time for you. My analysis shows the top four Profiler value horses to produce 70+% winners typically. But my analysis (so far) has not included Shape or Class filters. It has also required a minimum of 20 races in a profile before I consider the race in my records.

In the contest last week I had 3 wins, 3 seconds, and 2 thirds. My selections were derived either from the top simulator, top profiler, or by looking at these things, and seeing poor performance, realizing that the winner was coming from somewhere else. The previous week I ran 2nd in the first 3 races. Contests are tough.

-Nathan

In last week's Bris contest, Profiler had the winner among the top 4 in 8 out of 10 races. The interesting thing is that it had the winner in the top 4 in 5 races where I did not use the "Track" filter because I don't play those tracks or didn't have enough races in the db. It also had the winner in the top 4 in 4 of the 5 races where I did have data. I think that shows how strong Profiler is.

Hi yes your right with the filter i use comonly 70% is normal and low rarely go's below 50%.And it did have 7 of 10 winner's that day plus two of three longshot's one $15.00 and the other $19 and chg.