Shaun,

I assume your talking about results from real bets/lays. This is how I do it.

I have one table that records the race result (profit/loss and commission) regardless of selections.

It has :
Race id
Profit Loss
Commission

I have another table which records the bets that are placed and which systems have selected the selection and how much they each bet on it. This is where mine gets a little tricky. So I could have 10 selections all bet $1000 on a selection but I have a maximum liability of $5K. So I record that $10K would have gone on for those systems and also store the actual size of $5k as the total bet size.

So it has:

Race Date
Race Id
Selection Id
Profit/Loss
Bet Id
Type Of Bet (Lay or Back)
System Id
Bet Size for the system
Total Bet Size on the selection

I can then use these two tables with a bit of SQL magic to correctly work out the actual bet size for the system (if it was scaled back) and what portion of the commission should be assigned to it.

You may not have the above problem if you do not have too many systems. In that case your Bet Size of the System will always equal your total bet size. I can't do that.

So back to your original question ... it needs to be option 2.

Doing back testing on what if scenarios that include how systems interact with each other is actually really hard to do. The reason being that sometimes including a losing system can actually result in more profit because it turns over and generates commission for your account offsetting any premium charge. That's getting a little advanced though for your average systems bettor. Talk to me about it if you are paying premium charge on betfair and I can help out with some tips.