# Win/Loss per unit?

What does win/loss per unit represent?

Thanks.

I first thought it is your cumu \$ return (e.g. \$4106 in your case), divided by the total number of shares traded (8900 in your case, not counting open positions). However, that doesn’t give the \$0.63 in the table, but a much lower estimate. One possibility is that C2 calculates it as the (unweighted) average across the trades. Thus for each trade you take the profit or loss (in dollars) and divide it by the amount of shares. If you then take the (unweighted) average of this across all trades, you get the P/L per unit. In that case, I get \$0.63 for your system.

If that’s indeed the way it’s computed it’s difficult to interpret or use if the trade size is not constant. E.g. consider a system with 2 trades, one with 100 shares and a profit of \$10 per share, and one with 1000 shares and a loss of \$1 per share. The “P/L per unit” would be (10 - 1) / 2 = \$ 4.50, while the actual total profit would be zero.

"If that’s indeed the way it’s computed it’s difficult to interpret or use if the trade size is not constant."

Where were you a few months ago when I wanted to make that point in an almost infinite discussion?

Yes, this must be how it is computed. See also the (much shorter) discussion of me and Dustin a few weeks ago.

Oops, I must have missed those threads. Anyway, it never hurts that these points are made again every once in a while.

But, If one chooses to trade the above system with only 1 lot (100 shares) for each signal, his P/L would indeed be \$4.50. I prefer this because except when on a 1 lot basis, the statistics would also depend on the order (sequence) of the trades and therefore would be misleading.

But the P/L per unit isn’t even valid for the case that you trade 1 lot each time. As Dustin pointed out a few weeks ago. See that thread for his reasons.

It is valid because we must keep in mind that this is paper trading and not real trading. We have no way to forecast when this 1-lot would be traded in the real world.