Hi,
as far as I can see, quite a few systems on this website average down, to improve their statistics.
However, some systems take extreme risks by trying to pressure for a positive outcome (instead of clearing the trade and re-entering at a later time).
For this reason (and for some others) I would propose to treat such averaging down for statistics in a different way (e.g., for win%):
Treat the multiple trades as individual trades (based on FIFO). This would be more adequate also for traders who limit the number of
instruments in autotrading… - for example some systems trade typically with <6 but on very rare occasions they average down and go up to greater >30 - something everyone in his right mind should not go along for.
By treating this as independent trades, it would imply for someone who has a upper trading limit, that the statistics would be more appropriate for him / her, as it would mean simply one follows along for the first (sub)trades, but not for later.
It would also take away some of the incentive for system developers to perform averaging down, as this is typically done for hiding an initially wrong decision.
Cheers
Klaus
I made this suggestion a couple of years ago. I agree that it would give a more realistic winning percentage statistic. Those who average down keep on adding on til the whole position is net positive giving them 1 win and 0 losses for the position when in reality the individual trades may have been half winners and half losers.
The downside is that traders who average down are readily identified under the current structure of keeping the whole position one trade until it goes flat. Just one click on the trade shows all the piling on. I think a great improvement would be to be to display the trades the way they are now but book-keep winners and losers as you suggest.
Can’t remember why Matthew was reluctant to make the change, but a number of times he’s said there are legacy problems with changing stats. Can’t go back and re-do all the stats on all the systems that have ever been at C2.
Hi Keith,
right. Sorry, I did not make it explicit. For the trade list I also prefer this "aggregated view", but for statistics it would be better…
(Perhaps both views could be offered…)
A different angle would actually would be to get the whole statistics (including drawdown and everything "under the assumption that trading would be restricted to x positions", this would also allow to identify the
problems of some systems much better… )
Klaus
Although people average down to "improve" stats, it has a negative impact on other stats (it was the main driver for introducing APD). But also, max drawdowns and other stats tend to be hurt.
EVERY potential sub should avoid systems that will have most trades of X units, but occasional trades that are 3-4-30 times as many units. It is a system that likely is very dangerous.
But some of us were pressing for statistics based on a UNIT cost, which would water down the effects of averaging down.
Personally, I also usually use the first entry and the last exit of trades for a system (to judge it) that scale or average down, ignoring all the rolling up and down in between.
One thing to note. Just because you have a lot of trades within one unit, does not mean you are averaging down. A market neutral strategy may have two positions with many trades within those two positions.
I suspect that most systems average down to protect winning% or equity curve, rather than hedge or "market neutral"
Index,
You may be correct in that, I have not closely looked at other systems that use this method.
For risk management, I personally use a market neutral strategy in part of my methodology instead of cash. Since that piece is not correlated to my other timers it can help with gains while reducing the max draw.
