Algorithms that execute Buy or Sell orders based on daily closing prices may be a bit optimistic.
The first problem (at least in Equity Securities) is that Market Makers will often close up or down a price with very low volume to balance their position. If you subscribe to intra-day trading data, you will see this in the Time and Sales Log (sometimes identified with a special acronym like "NBDLT").
The second problem with closing prices occurs when there is a large bid or offer volume available. Even though a price may have been executed, your algorithm may not necessarily have been a benefactor.
Establishing a boundary upon which your program clearly crosses is more accurate then assuming it has participated in a price simply because a security has traded there. For example - a boundary condition may be, "Buy at $x, if $close < $x". In this case, you can be more certain that the routine will have been properly executed since the security traded below your Boundary Condition.
