Differences in Real-Time Performance

[Bob Fulks <bfulks@xxxxxxxxxxxx>]

I recently was testing a trading system (with TradeStation 4.0, build 
23) using tick-based charts (1-tick, 3-tick, 9-tick, etc).

I run it on real time data for a week and get certain trades.

I then turn the system status off, then turn it back on so that it 
recalculates the trades as with historical testing. The trades are 
somewhat different.

The system uses only "next bar at market" entries and exits (no 
stops) so the cause is not related to the bouncing tick issue, the 
built-in stops, etc., and trades last several bars.

The only reason I can postulate is that the ticks get stored in the 
data base in a different order than they were received in real time. 
(There are many ticks per minute and so lots of ticks have the same 
time stamp when viewed in the "Edit Tick" window in the server.)

I suspect this has been discussed many times but I could not locate a 
discussion that applied to this case.

Can anyone explain why this occurs?

Thanks

Bob Fulks