Tell me where this line of reasoning breaks down…. it’s just a thought experiment I did a moment ago.. I’ll take you through it, in the order I thought of things.
I’ve heard that a typical dummy trader (see here for examples if you have no idea what that means) has a 40 to 45% win rate. That means, if I take the other side of those trades, I should have a 55 to 60% win rate, as long as I always take profits at 1 R (where the dummy trader would have been stopped out). Luckily, with that kind of win rate, 1 R per winning trade is all I need to have a positive expectancy.
Of course, 1 R at 55% win rate isn’t going to amount to a whole lot of money unless I can find a whole lot of anti-dummy trades to take. But, what about this…
What if I set up two trading accounts. And, I always simultaneously entered my dummy trade and my anti-dummy trades? Then, when the dummy trade is stopped out, my anti-dummy trade mostly cancels out the loss (slippage and commissions make it not perfect). When the dummy trade works, 1 R of profit is cancelled out because of the loss I take on the anti-dummy trade. Again, it won’t be perfect due to the times the winner turns too fast, and you win less than 1 R on the dummy trade. But, overall, shouldn’t I come out a little richer (by like 10 R to 20 R per 100 trades) this way?
Even better than a little more profit, is the fact that my drawdowns should be pretty darn shallow. By definition, losing streaks in my dummy trades have equivalent winning streaks in my anti-dummy trades. Right?
Come to think of it, doesn’t it stand to reason that this scheme improves the profitability and consistency of any system that wins less than 50%? The lower the win rate, the more it helps. I’ve read that there are some trend following systems out there with like 20% or less win rates… in that case taking the anti-trades too would mean an extra 60 R per 100 trades, or so. Seems like that would make the drawdowns on those trend-following systems much more bearable, right?
I think I read about FX traders doing something like this, getting long and short the same position in two accounts… can’t recall where at the moment, though. Do any stock traders out there do it?
August 20th, 2006 at 4:28 am
Aren’t you assuming that every time a trade isn’t a winner that it’s a 1R loss? I’ve been thinking lately about winning percentage and how there might be too much emphasis on it.
In the long run, since the majority of your trades will be between -1R and 1R, those anti-dummy trades will simply cancel out your dummy trades. For trades over 1R, like you said, you’ll just take away 1R of the gains.
I think you’d need to see a distribution of your R to see whether this makes sense.
I used to trade a system with a 30% win rate with an expectancy of close to 1.0. I think this might have worked nicely with a system like that - i.e. one that relies on a small number of trades with very high R multiples to be profitable.
Your post is great food for thought!
August 20th, 2006 at 4:49 am
@Dave, no matter what size the losers are, adding the anti-system will have the shallow drawdown benefit. Of course, if there are enough small losers in the main system, then the anti-system might have a negative expectancy. In that case, you pay for the drawdown protection with smaller gains over time. That may not be as desireable, depending on your appetite for risk and reward.
The cancelling out effect is actually the point, really! I contend that if your win rate is 50% or less, and the majority of your trades are between -1 R and 1 R, then they mostly cancel themselves out anyway. I think a lot of accounts are like this. They owe all their real gains to a few big winners, while the rest of their trades are just up and down noise. Adding in the anti-system just nullifies all the small trades more completely and immediately. Makes the pain of a 5 to 7 loss streak much less.
Like you pointed out, dummy trades are maybe not the ideal system for this approach, since they have a win rate close to 50%. That’s just how I started thinking about it. But, low win rate systems (at first glance anyway) appear to me to be much stronger with the anti-trades in place.
August 20th, 2006 at 6:10 am
FX traders sometimes take both sides of a single pair for the same amount. The expectancy is that the pair will trend for some period of time but since the trader doesn’t know which way the trend will go if he takes both sides one side will automatically bust out while the other side continues to run.
There is no reason why that wouldn’t work in a trending stock but the trend would have to be strong enough and long enough to overcome overhead expenses.
Options traders have all sorts of methods for doing this using stock and puts/calls.
August 20th, 2006 at 6:22 am
@John: yeah, thinking about it more… I think the big overhead for stocks isn’t expenses, but rather that you’d have to maintain 2 accounts. It would theoretically always be more profitable to just pour twice as much money into whichever system had the higher expectancy. (assumes your biggest drawndown doesn’t kill you, of course). So, from that standpoint, implementing the anti-system always hurts your profit potential, while reducing your downside risk. Standard tradeoff, I suppose. I knew there must be _something_ I was missing!
August 21st, 2006 at 5:56 am
I would think that some of the 55% “loser” dummy trades are due to a whipsaw price action, so some of your anti-dummy trades would get whipsawed out too. You’d end up with a 45% win rate each way, or maybe worse. It would depend on the expectancy of each after that to see if it made money…
August 21st, 2006 at 4:16 pm
@ExEngineer: I agree. The cancelling out stuff is just approximate. The important thing is whether bost systems can have a positive expectancy (or at least the anti-system’s expectancy shouldn’t be so negative that it doesn’t justify the benefits).
August 22nd, 2006 at 4:51 am
The net effect should be that you’ve moved your entry by +1R on the Dummy trades and paid twice the commissions(assuming your AntiDummy trades have their stop placed at -1R).
Assumptions:
Two entries - one in either direction.
Positions exited if it moves 1R against us for either method. AntiDummy profits taken at +1R.
Scenario 1) AntiDummy Success
The trade moves in the direction of the AntiDummy trade. The Dummy trade is stopped out for -1R and the AntiDummy trade exits for +1R. Net effect 0R - (commissions).
Scenario 2) Dummy Succeses
The trade moves in the direction of the Dummy trade. The AntiDummy trade is stopped out for -1R. Your Dummy trade continues to make whatever it makes. Note that for the move first 1R move in favor of the Dummy trade, you netted 0R. The net effect is that you’re only seeing a gain once the Dummy has passed +1R. Any point between +1R and -1R is a wash - (commissions).
You’d be better off only taking the Dummy trades and entering at +1R above where the original Dummy would have entered. Hopefully that makes sense.
- MikeB
August 22nd, 2006 at 4:54 am
Blah.. proofreading.
“Note that for the move first 1R move in favor of the Dummy trade”
should read:
“Note that for the first 1R move in favor of the Dummy trade”
- MikeB
August 22nd, 2006 at 5:15 am
@MikeB: I had thought about that before, and rejected it. But, right now, I can’t recall why. What’s interesting, is that is seems to point out another condition for the strategy to be a success: Trades that go to 1 R must continue up more often than they fall back down. This is kinda also what ExEngineer was saying about the whipsaw effect hurting the composite strategy.
If you enter above +1 R, and you win, great. If you enter above +1 R, and you put a stop at your original entry point, then you lose at most 1R. So, compared to the original trade, it would appear you are taking the same 1 R risk for less reward than the original trade had. The only way this trade is more attractive than the original trade, is if 1R winners have a higher further win rate than the original trades did.
August 22nd, 2006 at 10:25 am
MikeB is right, the net effect can be accomplished without 2 accounts by waiting until the dummy entry is ahead by 1R before putting on the position. You are correct that at this point, you are still risking 1R (and with less reward), but the same is true when you are taking both sides of the trade in different accounts because your anti-trade just stopped out for a 1R loss, cancelled out by the gain in the open trade, but the stop is still 1R away in the open trade. Seems better to just skip the anti-trade and use the best entry.
August 22nd, 2006 at 10:54 am
If I had inta-day data I’d run some tests. If both systems have a positive expectancy, I’m still thinking anti-trading would outperform 1R trading (though I’m realizing that it’s probably not by very much, and you’d have to tie up twice as much capital to do it).
Great input, everyone! I appreciate it.
September 29th, 2006 at 10:09 pm
I noticed that nobody commented on your second question about trend following systems with a low win percentage, I think that was the exact concept that Linda Raschke took advantage of with her Turtle Soup trade.
September 29th, 2006 at 10:16 pm
@Rob: thanks!