slippage on rollovers

Discussions about the testing and simulation of mechanical trading systems using historical data and other methods. Trading Blox Customers should post Trading Blox specific questions in the Customer Support forum.
Post Reply
tigertaco
Full Member
Full Member
Posts: 23
Joined: Fri Feb 27, 2009 1:50 am

slippage on rollovers

Post by tigertaco »

Hi,

I'm testing a system assuming all buy entries happening at day's high and sells at day's low. But, if I do the same brutality on every rollover it kills the system which is not too surprising perhaps. My question is what's a good assumption for slippage at rollover points?

Separate question. I'm discovering (as everybody else on this forum probably already knows) that slippage can kill a moderately good TF system. I wonder what are the alternatives. Is limit order at the risk of missing a trade a good alternative (especially if trading a lot of markets and insisting on good entry points once signal is generated)?

t.

sluggo
Roundtable Knight
Roundtable Knight
Posts: 2987
Joined: Fri Jun 11, 2004 2:50 pm

Post by sluggo »

Instead of entering at market (MOO or MOC or something) you could enter at price P using a limit order. You could keep trying to enter at price P, every day, for some number "D" of days after getting the entry signal. If price never comes back and touches your limit price P, you could say to yourself "This is a hell of a strong trade, I want to get in no matter what!" and just enter at market on day number D+1. Or, you could say, "to heck with this trade, I'm not chasing it any more" and stop paying attention until the current trade (that you missed) is over.

P might be the Close of the day that you got the signal, or the Low of that day, or two ticks below the Open of the next day (i.e. "zero slippage AND two ticks of extra profit"), or some function of ATR, etc.

This idea adds another parameter to the system (D, the number of days you keep trying to enter on a Limit order, before giving up) which may or may not make you nervous about nonrobustness, overfitting, impure thoughts, and so on.

tigertaco
Full Member
Full Member
Posts: 23
Joined: Fri Feb 27, 2009 1:50 am

Post by tigertaco »

Thanks, sluggo. That's a good idea even with an extra parameter D. As for overfitting, it seems that the primary danger is underfitting. Once that danger is taken care of, some sort of stability of results wrto to parameter step may help with overfitting.

t.

Post Reply