Fully automated futures trading

[isotope1]

You're free to believe the "lengthy backtest." I believe the realized post-2008 trendfollowing results.

Oh I don't believe it at all. I think a more realistic Sharpe is probably 0.5 going forwards. Strict people might say that's not statistically significant; but would you really want to risk poor returns for a lifetime to find out for certain?

I originally came to all of this as a way to improve my AIM investing (I read lots of company reports, went to visit companies at their AGMs etc), but I had poor overall performance as I lacked a system, despite being right about many investments. I've fixed that now, and despite a blow up last month (Conviviality Plc, talk about losing 100% in 2 days).

Here are the main things I've learnt:

• It's important to have a system and stick to it. When it comes to equities, it's best to have equal £ value in each stock, as they have the same expected vol and return, as they are highly correlated. Many funds have holdings 10x the size of others. That makes no sense to me. Rebalance every so often.
• Futures are fun but you can have your ass handed to you. February 6th was a great example of that. I don't think I slept in 3 days.
• A paraphrased quote from "Advances in Financial Machine Learning". In the early days, you could find gold yourself with a shovel. Now, you need teams of people, specialists and millions of $ of equipment to mine gold. Some aspects of quant trading are like this too. Simple strategies degrade, and finding the gold is harder.
• A model is only that; a crude approximation of reality. Never believe a backtest.
• With trading, HMRC takes a cut. Postponing that cut for as long as possible makes a real difference.
• The #1 way to lose money is to be overconfident. Nobody ever went broke taking it easy and taking a profit.

Perhaps the last one is one that I read from Warren Buffett, which is ~ "Don't risk what you can't afford to lose for what you don't need." Personally, I like to sleep well at night.

 

[truetype]

GAT will say not statistically significant!

Of course 10 years of daily returns data are significant.

 

[traider]

Of course 10 years of daily returns data are significant.

Why do you believe trend following is different from the past?
GAT says the strategy is not overcrowded yet, still has more capacity.
Why choose 2008 as the structural break? What happened? Volcker rule?

 

[traider]

Oh I don't believe it at all. I think a more realistic Sharpe is probably 0.5 going forwards. Strict people might say that's not statistically significant; but would you really want to risk poor returns for a lifetime to find out for certain?

I originally came to all of this as a way to improve my AIM investing (I read lots of company reports, went to visit companies at their AGMs etc), but I had poor overall performance as I lacked a system, despite being right about many investments. I've fixed that now, and despite a blow up last month (Conviviality Plc, talk about losing 100% in 2 days).

Here are the main things I've learnt:

• It's important to have a system and stick to it. When it comes to equities, it's best to have equal £ value in each stock, as they have the same expected vol and return, as they are highly correlated. Many funds have holdings 10x the size of others. That makes no sense to me. Rebalance every so often.
• Futures are fun but you can have your ass handed to you. February 6th was a great example of that. I don't think I slept in 3 days.
• A paraphrased quote from "Advances in Financial Machine Learning". In the early days, you could find gold yourself with a shovel. Now, you need teams of people, specialists and millions of $ of equipment to mine gold. Some aspects of quant trading are like this too. Simple strategies degrade, and finding the gold is harder.
• A model is only that; a crude approximation of reality. Never believe a backtest.
• With trading, HMRC takes a cut. Postponing that cut for as long as possible makes a real difference.
• The #1 way to lose money is to be overconfident. Nobody ever went broke taking it easy and taking a profit.

Perhaps the last one is one that I read from Warren Buffett, which is ~ "Don't risk what you can't afford to lose for what you don't need." Personally, I like to sleep well at night.

I talked to a fund manager before, he runs a long short strat. Surprisingly he put all his net worth into 1 stock. Guess which one it is?

AAPL

He invested long time ago so now he just does his job for the fun of it.
Luck is more important in this age when alpha is scarce...
Don't take it too hard man.
Have also seen people make and lose millions in a day. It distorts one's world views quite a bit...

 

[truetype]

GAT says the strategy is not overcrowded yet, still has more capacity.
Why choose 2008 as the structural break? What happened? Volcker rule?

"Crowded" is a question of degree. Certainly there's more money in explicit trend and closet trend (risk parity, macro, Meb Faber) than ever before. As to "what happened in 2008" -- I hope you're joking. Or maybe you're a 21 year old?

 

[HobbyTrading]

Of course 10 years of daily returns data are significant.

The available data is not the determining factor. The number of trades which took place during this period is. This will determine whether you have a large enough sample base to conclude something which is statistically significant.

 

[traider]

"Crowded" is a question of degree. Certainly there's more money in explicit trend and closet trend (risk parity, macro, Meb Faber) than ever before. As to "what happened in 2008" -- I hope you're joking. Or maybe you're a 21 year old?

I'm aware of the financial crisis. Just wondering if there is some fundamental reason as to why things have permanently changed.

 

[globalarbtrader]

Of course 10 years of daily returns data are significant.

The available data is not the determining factor. The number of trades which took place during this period is. This will determine whether you have a large enough sample base to conclude something which is statistically significant.

Fascinating discussion here. Just my 2p worth - using higher frequency data does not significantly improve statistical significance because both noise and Sharpe Ratio change at the same rate of square root(T); you get a slight improvement because the critical value for the T-statistic falls with more observations but this decreases rapidly and once you are at monthly data using daily isn't going to give you any advantage. And looking at trading data probably doesn't help either for the same reasons. 10 years would only be enough evidence if the realised SR was a lot higher than 0.5, and higher than 1.0 for that matter.

Personally I assume my SR is either 0.5 (to determine my risk target at half kelly, 25%); or 0.0 since I assume zero return from my trading account when calculating my household budget (lucky because last year was pretty flat again; I'll be posting a full annual update on my blog shortly).

GAT

 

[isotope1]

I talked to a fund manager before, he runs a long short strat. Surprisingly he put all his net worth into 1 stock. Guess which one it is?

AAPL

I sold all my apple shares (£7000 worth) just before the release of the iPhone. That's worth ~£250k today! oops! I also had a bitcoin mining rig, and sold all my BTC at $300.

I'm sure trend following will continue to work for generations to come, but in the futures markets I posit the performance may degrade because of HFT arbs increasing the correlation of related contracts, thereby reducing the diversification available. I think a real risk with this formulation of TF is that it provides the illusion of diversification, when in reality that diversification may have been arbed away by faster traders in the last 20 years.

 

[Napoleon@Dynamite]

The theory of the degradation of trend following is further boosted, I think, by the equity curves of various iterations of the strategy over the last few years. They have a pronounced and long-lasting levelling-off, not seen in my backtests back to the 70s. I would say it is an unprecedented degradation in performance since then. Do you all think this is significant or just a random, although rather large, bump in the road?