More Winners or Larger Avg. Winner?

[wally_]

Quote from inandlong:


Preference cannot be decided by a mathematical model. Preference is subjective, models are not.

That is true and in fact sometimes that means simply unjustified prejudices. It is better to have justified prejudices. One of my systems had a bad drawdown in 2001 about which I learned after I started using it. But I have chosen it based on other criteria and the fact that it peformed well in the first 3 quarters of 2002. The system is doing very well and it would do better in 2001 if I wanted to 'optimize' it, which of course would be silly.

In fact, in 2002 it had only a very small dip in equity curve on a monthly basis, but if you were to judge it by its performance in 2001, you would most likely discard it.

 

[darkhorse]

Quote from Corso482:

Assuming the two result in the same return, would more winners be superior because it result in a smoother equity curve?

The question overlooks a few important real world considerations. High winning percentage strategies carry disadvantages as well as advantages. You have to pay a price for that artificially smooth equity curve.

1) Higher win % requires smaller targets. Smaller targets require higher frequency and leverage to get a decent payout. Higher frequency and leverage require higher overhead and higher exposure. Higher overhead and exposure = higher vulnerability in times of drought. When it's on, it's on- but when that bad month hits, you run the risk of being fragged by your liabilities (imagine, for example, breaking even with the market and owing your broker 10K in commissions anyway). Of course, you can reduce your overhead and exposure by cutting back on size and frequency, but then your potential profits diminish. Live by the sword, die by the sword.

2) High win % strategies invariably have restrictive size caps because of the small target requirement- limits to how much size you can work with before your typical slippage cost becomes greater than your standard profit objective. In other words, there is a built in salary cap. This may not matter to most daytraders, but it matters a great deal to those trading millions or even billions of dollars.

(There is a way to maintain a high winning percentage while trading big money, but it requires trading core positions only a few times per year, or maybe even a few times per decade.)

High % strategies are like temperamental italian sports cars imho. They're sexy, they're fast, and they give the driver a thrill. But they're also ridiculously high maintenance, no good for long trips, and when they break down you're kind of screwed.

p.s. many of the biggest and best traders in the world have a 90/10 ratio or something close to it- i.e. 90% of profits coming from 10% of trades. This seems to be a natural result of starting small as a function of risk, and then getting more aggressive as the position unfolds in your favor. Another strike against high % strategies: premature exits = missed opportunity to take a low risk/ no risk swing at the longball.

 

[darkhorse]

p.p.s. which brings up another thought: there's no reason why you can't have a low win percentage AND a low max drawdown at the same time, given the right combination of risk management skills and pyramiding skills... pretty dumb of me to overlook this point as it nicely reflects my own approach...

 

[Fluidity]

Quote from darkhorse:

p.p.s. which brings up another thought: there's no reason why you can't have a low win percentage AND a low max drawdown at the same time, given the right combination of risk management skills and pyramiding skills... pretty dumb of me to overlook this point as it nicely reflects my own approach...

Sums my trading up very well... Low clip, big winners, high DD/Net ratio...

I think it is often a folly of traders to think that a high win % will insure a small DD, but in my experience it has always been my low win %-high/R-multiple methods that had the best DD/Net on the macro-levels...

Plus, I must say, it is a damn fine feeling when you wheater a storm of 7-8 losses, keep your head, and then make it all back with one solid hit...

 

[wally_]

I believe it's good to combine systems of different characteristics. I do agree that a higher percentage of wins is not necessarilly indicative of a superior system, but it's good to have systems that have both decent win/loss ratio (say, 1.5 or so) and decent percentage of winners (say 50%).

One of my systems has a low percentage of winners, but they are big and the equity curve has been very smooth for the last 13 months, two others however have higher frequency of winners and a decent ratio of wins to losses, and too have pretty smooth equity curves. So you can have smooth equity curves in both types of systems.

But it depends on what you really want... so back to preferences.

 

[Mr Subliminal]

Quote from acrary:

Actually they aren't the same. Here's a example using the assumptions I used from above with 3% risked per-trade instead of 1% (to show the effect a little stronger). Here's a cycle of 10 trades using each method to show that the higher win % with the same expectancy and same % of equity risk has a superior return. It's caused by using the % of equity risk method which penalizes a lower winning % in favor of a higher win %. If the order of trades were reversed or jumbled, the outcome would be the same.

I am enjoying this brain exercise as well. Your equity.gif is convincing and has me scratching my head, trying to reconcile our different results. But here is where I really need your help - consider a trader who has no edge :

% win = 50%
win size = 1

% loss = 50%
loss size = -1

expectancy = (.5 * 1) + (.5 * -1)
= 0

For a starting equity of $100,000 with 3% risked per-trade, my simulation yields $100,000 as the expected equity after 2 trades. Your method of analysis, which is attached, has transformed our supposedly breakeven trader into a losing one. Who is right?

 

[Girlpower]

Quote from Mr Subliminal:

I am enjoying this brain exercise as well. Your equity.gif is convincing and has me scratching my head, trying to reconcile our different results. But here is where I really need your help - consider a trader who has no edge :

% win = 50%
win size = 1

% loss = 50%
loss size = -1

expectancy = (.5 * 1) + (.5 * -1)
= 0

For a starting equity of $100,000 with 3% risked per-trade, my simulation yields $100,000 as the expected equity after 2 trades. Your method of analysis, which is attached, has transformed our supposedly breakeven trader into a losing one. Who is right?

Hmmmmm. now that is very interesting. I've been thinking about this phenomenon, and putting some numbers into a quick mathematical model (hastily put together I might add), but the results also seemed to vary according the the order of the winners/losers when the size is 3%.

Now I'm really puzzled. I expected the outcome to always be evens (like you said), but it wasn't. Not entirely sure that the model was correct, but in any event sometime I will need to figure out why the order was so important and find the error in the model if indeed there is one...

Rgds
Natalie

 

[Mr Subliminal]

Quote from Girlpower:

Hmmmmm. now that is very interesting. I've been thinking about this phenomenon, and putting some numbers into a quick mathematical model (hastily put together I might add), but the results also seemed to vary according the the order of the winners/losers when the size is 3%.

The order should not matter in acrary's model.

 

[acrary]

Quote from Mr Subliminal:


For a starting equity of $100,000 with 3% risked per-trade, my simulation yields $100,000 as the expected equity after 2 trades. Your method of analysis, which is attached, has transformed our supposedly breakeven trader into a losing one. Who is right?

All I was showing was that by using the % of equity as a sizing strategy, it adds a negative expectancy that has to be overcome by the expectancy of the strategy.

There was more about this in another thread and I put together a little spreadsheet that shows the more you risk per-trade the higher the drag your strategy must overcome.

http://www.elitetrader.com/vb/showthread.php?threadid=11054&perpage=6&pagenumber=1

 

[Girlpower]

Quote from Mr Subliminal:

The order should not matter in acrary's model.

Well, here's the interesting bit.

The order didn't matter when a fixed sum was risked, and the outcome was evens whether it was 50/50 on 10 trades, 100, 1000 or whatever. However, when the sum risked was a % it made quite a big difference. It does seem that the draw down from the losers outweighs the run up from the winners.

I would guess it works on the same basis as an initial 50% loss requires a 100% gain to get back to even.

I'm still checking the maths, and trying to work out the reason.

rgds
Natalie