Cashish 

You Go 'dog!

Thanks, but it is a long, long way to go!

kevinkdog 

I will post something on discontinuing a little later today. In the meantime, do you have a link where FT71 talks about next trade being 50/50? I'm having trouble understanding how that is true for strategies like option selling (where you might have a 90% win percentage) or trend following (I have one strat that wins only about 25% of the time). Thanks

Hi there,

If I recall correctly he talks about it in this webinar:
Webinar: FuturesTrader [AUTOLINK]FT71[/AUTOLINK] on Risk, Sizing, [AUTOLINK]Scaling[/AUTOLINK] and Trading Probabilities

He has mentioned it in several webinars. Essentially his point is that the outcome of the very next trade is 50/50. Yes, you may have a historical win percentage of X. But you dont know ahead of time where in the distribution of trades your wins and losses will be. All you know is that over time, your win percentage will likely be around X percent. So he's not referring to your overall win percentage, he is talking about the very next trade.

Give it a listen/watch, he explains it far better

kevinkdog 

OK, so you want to know if your new strategy is performing to expectations, right? This is nice, but it doesn't convey a lot of information, especially early on. To get more insight, add 2 lines for the +/- standard deviation curves: Upper curve: n*avg + sqrt(n)*(std dev)*X Avg curve: n*avg Lower curve: n*avg - sqrt(n)*(std dev)*X What this tells you is that roughly 68% (X=1) or 95% (X=2) of the time, your equity curve should be within the upper or lower bands. That is a pretty good range of performance. If it is outside of those bands, maybe there is something going on with your strategy. Here is what my Combine system looks like with X=2 bands:

Hi Kevin, this goes back to a post of yours from a week ago. I am sometimes a bit slow (and not a King of Statistics )
Anyway, as I was trying to digest the meaning of the bands, this is what I got confused about. When you cut off the 'too lucky' and 'too unlucky' tails of the bell curve at 2 standard deviations from the mean, you exclude 2.5% worth of very lucky samples and another 2.5% of very unlucky samples, keeping the least extreme 95% in between. Do you agree that what you call "Lower 5%" is actually the curve below which you've got the worst 2.5%? (and what you have above the "Upper 95%" curve is actually the best 2.5%)

A related question: what was actually the expectancy and the standard deviation that you had at the end of your testing? I estimate the expectancy (average trade net profit) was around $71-$72 - but how much was the standard deviation of the walk-forward sample?

Trovatore 

Hi Kevin, this goes back to a post of yours from a week ago. I am sometimes a bit slow (and not a King of Statistics ) Anyway, as I was trying to digest the meaning of the bands, this is what I got confused about. When you cut off the 'too lucky' and 'too unlucky' tails of the bell curve at 2 standard deviations from the mean, you exclude 2.5% worth of very lucky samples and another 2.5% of very unlucky samples, keeping the least extreme 95% in between. Do you agree that what you call "Lower 5%" is actually the curve below which you've got the worst 2.5%? (and what you have above the "Upper 95%" curve is actually the best 2.5%)

Yes you are correct. My choice of labels was misleading, at best. Here is a chart that shows the correct distribtuion graphically.

Trovatore 

A related question: what was actually the expectancy and the standard deviation that you had at the end of your testing? I estimate the expectancy (average trade net profit) was around $71-$72 - but how much was the standard deviation of the walk-forward sample?

Since I am using 3 different strategies here, it is easiest to express the expectancy based on a trading day...

Average Trading Day: $71.42
Std Dev: $425.17

Average Losing Day: -$294.31

If you consider expectancy as average trade, it is $71.42

If you consider expectancy as avg trade / avg loss, it is .243


Another thing to point out is that the distribution of trades is not a normal (bell) shaped curve, which can be a big problem, depending on the analysis you do.

In the attachment below, you will see an article that I wrote for the August 2010 issue of SFO Magazine (not defunct - thanks Russell Wassendorf Sr.!)


If you don't want to read the article, here are the main points:

1) Establish a quitting point or stopping point for any strategy you begin trading - BEFORE you start trading it.

2) Write it down, share it with trading partners, etc.

3) Follow It!


From what I've seen, and from my own personal experience, I can tell you that this is much harder than it seems. In fact, most people don't do this at all, and instead make a knee jerk, emotional, heat of the moment decision to stop trading a system. Of course, that is the worst time to make any kind of decision.

That's a great idea, thanks!

kevinkdog 

Another thing to point out is that the distribution of trades is not a normal (bell) shaped curve, which can be a big problem, depending on the analysis you do.

I guess that's where Monte Carlo simulation comes on stage as a handy tool? I mean you don't need to be a professor of statistics but you can still make good sense of the results, like how likely is it that you will have at least X dollars of profit after N trades, or no more than Y% drawdown.

Trovatore 

I guess that's where Monte Carlo simulation comes on stage as a handy tool? I mean you don't need to be a professor of statistics but you can still make good sense of the results, like how likely is it that you will have at least X dollars of profit after N trades, or no more than Y% drawdown.

Yes, that is a very good way to put it.

kevinkdog 

Since I am using 3 different strategies here, it is easiest to express the expectancy based on a trading day... Average Trading Day: $71.42 Std Dev: $425.17

The ratio of the expectancy (average trading day) over the standard deviation is 0.17 in this case.
It's interesting to look at this value from the perspective of Van Tharp's guidelines to system quality. In his book Super Trader, he suggests that a system with a ratio of 0.16-0.19 is 'poor but tradable' (this sounds a bit harsh but that's not my purpose). He considers ratios from 0.2 upwards 'average', 0.25+ 'good', 0.3+ 'excellent', 0.5+ 'superb' and 0.7+ 'Holy Grail'.

This is how I can relate Dr. Tharp's classification to your Combine experience: you have a good reason to think that your system has an edge (it's tradable). It has a positive average trading day, which is a good basis to expect profits in the long run. However the aspect that makes it less than good (in his harsh wording: poor) is that the standard deviation is relatively high. If you could get the same average day with half the standard deviation (bringing the ratio to an excellent 0.34), your 95% range breadth would also be halved. One consequence is that you would need far less trades to see the lower band cross the zero line: exactly the criteria you mentioned as so important to give green light for trading with real money.
As I calculated assuming a normal distribution, it takes almost 100 trades (days) with the ratio of 0.17 for the worst 5% to reach slightly positive profit. With a ratio of 0.34, this 'worst 5% breakeven' would take 24 trades (days). This is quite a difference, and especially so in the context of the Combine, where the first is beyond the end of the Combine period, whereas the second means you would have to be very very unlucky to finish with a loss.

I wonder what is a fair conclusion regarding the decision whether you should trade this strategy with real money. A possible answer: if you are patient enough, you could. However, if you have (or can develop) a strategy which has a better average/StDev ratio, then it's better to use that one and not your current Combine strategy. Though even then, it depends how highly you value the diversification aspect.