I assume you mean you have too many candidates for your available capital.
This is a large field of study - I can't really do this justice in a few sentences.
You should run some monte carlo tests against a random selection of entries - make sure your system is robust enough to handle an alternative sequence of trades.
If you have a trading system that generates more signals than available capital with a large universe then you could conisder a smaller universe (but of course, you need to backtest it). For example, if you are using Russell 3000 as your candidate list, you could try Russell 1000 instead, or S&P 500. Make sure you do not introduce survivorship bias.
Alternatively, investigate ranking mechanisms (i.e. a signal score) and also backtest that. Ensure that such ranking is not forward looking.
There's a few items for thought for you, but developing a robust system requires a lot more coverage than just these points.