Looks to me like your basic approach is pretty sound, avoiding backfitting and testing over various timeframes. A ball park figure of 1000 bets is what I would use as a minimum testing area but there are guys here with more statistics knowledge than I who will advise you better with reference to your strike rate etc.
How are you testing? I read you are inputing the data so I assume you are testing on a computer in some way. If so there are CSV files (data separated with commas) of results on a few sites which you may be able to download and put into your system for testing.
Sounds like a good system so far, though how you derived it based on the results of 1 month's races I've no idea. Best of luck anyway.

KV