Lets not forget that we can literally only bet 1% of our capital on system 2 to go bust. So one major factor in evaluating which system is better, is the bet size. If we plot the win/loss percentages for systems 1 and 2 into excel and create a simulation of 1000 trades. We'll easily see that system 1 far outperform system 2, just by the fact that you can place a larger percent of your capital on each bet without risking too much.
What I see is your unjustified assumption that only actual trading can produce performance analytics. So obviously you don't believe in backtesting, which anybody, not just large institutions, can do. We agree to disagree.
Yes, the Kelly fraction is an excellent performance metric, which is why I included it in my own composite performance metric. http://www.elitetrader.com/vb/showthread.php?threadid=270199
I was going to add a comment on Kelly in my post, but I decided not to. Basically, using Kelly here would be meaningless, no matter what it tells you, 1% of capital will be your max bet on system 2. As an example: If you allow for a max 10% loss on your capital. You can only bet 0,10% of your total capital on system 2. On system 1, however, you can bet 10% of your total capital. The edge you get from being able to bet 10% as opposed to 0,10% far outweighs the edge you have in expectancy from system 2. You'd need an expectancy of 20+- on system 2 to even compare to system 1.
Basically what Kelly tells you is that betting 1% of your capital would be disastrous, given that the Kelly fraction is less than a quarter percent. Using your logic, a braver trader willing to risk losing up to 30% of his capital should risk 0.3% per trade, but knowledge of Kelly would inform him that he is overtrading. You can ignore Kelly in your trading but that's not the optimal play to make.
I'm not trying to discuss whatever is the optimal bet. I'm saying that system 1 is by far better than system 2, because of the big difference in how much we can bet on each system.