Sample size for assessing results

[speedtree]

I've been using Patternform for some time now as part of my criteria for assessing potential DOBS. Thanks for all your work, it's a fantastic resource.

I have a question maybe someone can help me with. Lets say I have completed 1000 DOBS over NH races ranging in price from 2.02 to 100 and the strike rate is 60%. I have been looking for areas to improve the SR, but am unsure about the sample size for making a judgement. For example, horses priced 6-7 are only performing at 49%, but there is only a sample of 90. I would assume that this would correct itself over time as the sample gets larger, but at what point is a sample size large enough that we can be confident of our results being close to a true/accurate representation and not just an example of random variance?

[mikesonoffjohn]

Dont think you can truly ever have a sample size big enough i find things that worked in 2017 dont always follow on to next season . To get a truer reflection on results so far i would categorize into Novice Hurdle / novice handicap / handicap etc and then have a look , Big meetings also tend to throw statistics in the air as i assume every horse is trying . I used to paper test systems (non Patternform) that i was privy to via a firm who would offer them to bettors including Football and also every Telephone tipster and they all produce Wavy lines when on graph paper taking up way too much time , My conclusion is Years rather than size and break it down as far as you can . Try to avoid "Tweaking" things i fell for that a lot

[speedtree]

Thanks for the reply, much appreciated.