So don't be too tied to history? Right.
And yet, all your rules are based on history. Is there a contradiction there? No, because a good trend-following system will keep you in the market until there is evidence that the trend
has changed. If you had been doing your historical research on soybeans in 1972, you would have concluded that any
time soybeans advance by 50 cents, you might as well get out, because the market had never moved up or down by
significantly more than that. Obviously, that was the wrong conclusion because it went up another $8. A good trend-
following system, however, would have kept you in for most of the move.
So you don't want to draw boundaries from history over market behavior? Right. The correct approach is to say: This structure means up, and this structure means up no more, but
never that this structure means up this much and no more.
When you trade a system, do you go with the version of the system that tested out best for the past, or do other factors command consideration? One of the toughest problems in deciding how to trade is whether you just go with what is optimal for the
data base, or whether you start from some other premise. You might deliberately trade something other than the
optimal parameter set [version of the system with the best past performance] because you think the future is going
to be unlike the past in a specific way. By definition, any other parameter set is going to have a poorer past
performance than the optimal set. But if the difference in performance is only 10 percent, it might well be worth that
10 percent difference if you believe the suboptimal set, as measured by past data, will fit the future better.
You have gone from being a very small trader to a very large trader, especially now that you are managing outside money. Do you find that order size gets in the way? Does it become substantially more
