Dabbling in the stock market

[lemur]

from http://seekingalpha.com/article/99747-the-dangers-of-timing-the-market

"
I wrote a post earlier this year, Focus on the the Long-Term, in which it was noted missing just a handful of the market's best days in a given year can really penalize returns. If an investor missed just 40 of the biggest up days in the market over the last 20 years (1987-2007), their return would have totaled 3.98% versus remaining fully invested and achieving an average annualized return of 11.82%.

The market research firm DALBAR went one step further and looked at the returns of mutual fund investors over the 20-year period, 1986-2006, and reported the average market timer return was -2%. During this same time period, the S&P 500 Index returned 12%.
"

Selection bias pure and simple.....

What if you missed the 40 worst down days.... apply those numbers

20 yr period- more selection bias. Far too short a time period. If you bought in 1929 it took 26 yrs to get back to break even etc etc. similar argument applies to the 1970's bear market. If you bought the Nikkei in the late eighties, you would still be waiting.

Average market timer ... thats the old stick one foot in cold water and one foot in hot water and on average you get little change argument. Meaningless cliche stuff.

 

[z106]

An interesting thread. The age old argument of trading vs buy and hold.

For my tuppence worth - while running the risk of seemingly sitting on the fence - I think both arguments are correct.


Lemur is referring to trading and is referring to having a system back tested that produces a positive expectancy which he refers to as his 'edge'.
His reference to the casino was an example of the where the casino has an edge (i.e. positive expectancy vs the punter).
ANother example might be if you go into a bookies and ask him for odds on heads or harps in flicking a coin he certainly won't give you even money but more than likely 5/6 or 10/11 or something.
That is his edge.

What pro traders to is try to create an edge similar to this.

Lemure - for the record - I think you are being WAY too paranoid in refusing to even give an example of a set up just for illustration purposes given the obvious confusion by some punters on this thread to your explanations.

Anyway - as an example - some typical rules a trader may employ to give him his edge might be along the lines of:
a) Never bet more than 1% of your total available pot on any one bet
b) Never have more than 20% of your total available pot invested at any one time
c) Only buy if a 10 week moving average crosses over a 40 week moving average.
d) Never buy unless the 200 day moving average is rising.
e) a max leverage factor of 2

Basicaly there could be many many rules which once tested will display a positive epectancy over historical data (Back testing is a bit of a science in itself too mind you as there are good and bad ways to back test).

Once some traders realise the system they have selected displays a historic positive expectancy they will then use it going forward.

Buy and hold also offers good returns if properly diversified (as opposed to the top 10 in the iseq as mentioned earlier).
Buy and hold will never offer the potential riches of trading though.

That said - being a trader requires a lot of discipline - which in fact is the hardest part of being a trader i reckon.
Coming up with the system with teh edge is only half the battle.
It's unquestionably largely a psycological battle in my book.

FOr anyone who is interested you should read market wizards and market wizards 2 - some good stuff in there.

Basically - I think buy and hold vs trading depends on the type of personality you have.
Different strokes for different folks - hence agreeing with both arguments.

To the OP - unless you have a good strategy worked out in advance - regardless on buy amd hold or trading - you will near certainly lose money.

 

[lemur]

An interesting thread. The age old argument of trading vs buy and hold.


Lemure - for the record - I think you are being WAY too paranoid in refusing to even give an example of a set up just for illustration purposes given the obvious confusion by some punters on this thread to your explanations.

Anyway - as an example - some typical rules a trader may employ to give him his edge might be along the lines of:
a) Never bet more than 1% of your total available pot on any one bet
b) Never have more than 20% of your total available pot invested at any one time
c) Only buy if a 10 week moving average crosses over a 40 week moving average.
d) Never buy unless the 200 day moving average is rising.
e) a max leverage factor of 2

Fair enough comment qwerty.

However those rules don't constitute an edge. An edge is something you calculate after running a sequence of say 20 trades using the same set up rules. My favorite set up right now has an edge of 25%. So in other words, for every $100 I risk, I make $125 on average.

A well known set up example would be: the gap play set up on US equities. If you apply a set of filters, the gap has a 70-80% chance of half filling the gap which is an excellent edge. The stop is at a risk reward ratio of 1. You can then keep compounding the gains to get a very high rate of return with this set up.

The trick is to apply the right filters and then concentrate on good execution.

 

[z106]

Fair enough comment qwerty.

However those rules don't constitute an edge. An edge is something you calculate after running a sequence of say 20 trades using the same set up rules

Given that statement i am assuming you don't understand me correctly.
(Unless the fact that the examples i gave are a mixture of money-management rules as well as some entry signals. Maybe i shoudl have been clearer with my rules of set ups and buy and sell signals only as opposed to throwing money management rules in there too)

Obviously i am not saying those specific rules give an edge. But these rules could be an edge once backtestedand a positive expectancy is displayed.

My main point is that many successful trading strategies use similar rules(excluding the miney-management rules) to give themselves an edge.
This edge is determined by backtesting the rules to prove the positive expectancy.
The rules i outlined above were off the top of my head for illustration purposes.

In fact - any type of rule can be employed once it shows an edge.
It could be something like buy a certain index in september and sell in may.
THat may equally be a successful trading strategy.

The main general point is that you are looking for value for money bets
i.e. where the reward is greater than the risk
E.g. getting 6/4 on the flick of a coin.

qwerty - I would say that a,b and e sound like bankroll management rules that are espoused for games like poker, in order to ensure that you don't go broke, if you have an unlucky streak.

In answer to SPC100 comment aboive you are right in that a trader must also use proper money management in conjunction with this successfully backtested trading strategy. (Ok- you are right in that strictly speaking admittedly not their edge)
But without the proper money management then the edge is useless.
These are all equally important parts of the jigsaw for coming up with a successful trading strategies.
Casinos and bookies also employ money management with their edge by having maximum allowable bets - which they combne with giving the punter bad-value bets only (i.e. their edge) - which basically is why the booie/casino never loses.


WIthout question traders use exactly the same general tecniques on the market as bookies/casinos use for horses or roulette or whatever else they allow bets on.

Lemur - going by your posts you seem to now what you are talking about.
That 25% edge is pretty impressive.

How long have you been trading by the way?

Also - out of curiosity how did you get involved in the game?
Are you a full time trader?

Any books that you would recommend?

I've read a good few aready.
Just wondering if there are any great one i have missed out on.

 

[zephyro]

Obviously i am not saying those specific rules give an edge. But these rules could be an edge once backtestedand a positive expectancy is displayed.

Do you or anyone else know of even one rule that outperforms in backtests in at least three statistically significant out-of-sample periods? The 'evidence' I've seen until now is laughably lacking in out-of-sample verification, e.g. Covel in 'Trend Following' backtesting from 1990 to 2004 using a specified portfolio and then claiming that backtesting using very similar instruments during the same period represents out-of-sample verification!

However those rules don't constitute an edge. An edge is something you calculate after running a sequence of say 20 trades using the same set up rules. My favorite set up right now has an edge of 25%. So in other words, for every $100 I risk, I make $125 on average.

Do you mean that you think this set up has an edge of 25% because that's what it has averaged over 20 recent trades? Do you realise how big a role luck can play in a sample size as insignificant as 20?

 

[z106]

Do you or anyone else know of even one rule that outperforms in backtests in at least three statistically significant out-of-sample periods? The 'evidence' I've seen until now is laughably lacking in out-of-sample verification, e.g. Covel in 'Trend Following' backtesting from 1990 to 2004 using a specified portfolio and then claiming that backtesting using very similar instruments during the same period represents out-of-sample verification!

I don't think covel was expecting people to use his backtesting strategy as the definitive way of backtesting.
To be fair to the guy he does state on numerous occassions that backtesting is a science in itself.
He doesn't attempt to go through that science in detail because trend following is the core of that book.
I think he states in no incertain terms that backtesting must be done correctly to verify an edge.

I wouldn't get too bogged down in his backtesting methodology as that misses his main point which is is that edges do exist and that trend-following is one way of establishing an edge.

Do you mean that you think this set up has an edge of 25% because that's what it has averaged over 20 recent trades? Do you realise how big a role luck can play in a sample size as insignificant as 20?

Ya - I would certainly agree that 20 is in no way a proper sample size to claim a system has a certain edge.

 

[lemur]

Do you or anyone else know of even one rule that outperforms in backtests in at least three statistically significant out-of-sample periods? The 'evidence' I've seen until now is laughably lacking in out-of-sample verification, e.g. Covel in 'Trend Following' backtesting from 1990 to 2004 using a specified portfolio and then claiming that backtesting using very similar instruments during the same period represents out-of-sample verification!



Do you mean that you think this set up has an edge of 25% because that's what it has averaged over 20 recent trades? Do you realise how big a role luck can play in a sample size as insignificant as 20?

Zephyro. I trade my set ups with real money every day on the market so 'out of sample' backtest arguments are not relevant.

Yes I have often run 'confidence interval' stuff requiring 212 samples for 95% confidence Blah blah blah but trading in real life does not work like this. The confidence stuff is relevant to a sterile engineering environment where the inputs/outputs can be controlled but not in the market which is dynamic and suffers from 'trader effects'.

Once a set up becomes well known among traders it stops working. Hence in practical terms a trader does not have the luxury of sitting around waiting for hundreds of samples to determine if he can get an edge with this method. In my experience, a sample size of twenty is adequate in real terms for an indication that the set up will improve odds.

Backtesting - I am not a big fan of this as it is very difficult to capture all of the dynamics that will happen in the trading day. For example, I once had a trade set up with 452 backtests which had a very nice 3% edge per trade on average. When I tried to put it to work it failed miserably because I used end of day prices for the tests which did not catch the wider daily variation which often ran my stops.

 

[lemur]

Lemur - going by your posts you seem to now what you are talking about.
That 25% edge is pretty impressive.

How long have you been trading by the way?

Also - out of curiosity how did you get involved in the game?
Are you a full time trader?

Any books that you would recommend?

I've read a good few aready.
Just wondering if there are any great one i have missed out on.

Been trading - first started in the late 1990's. I worked as an engineer for many years before I packed it in to become a full time trader. A risky career move for sure but it has worked out for me.

Its difficult to learn this business from books. You are welcome to call out and see me if you want me to show you some stuff for real.

 

[lemur]

lemur and qwerty do you believe stock prices are a random walk?

Stock prices are largely random with a 10% upward drift over the last 100 years. However, they are pseudo random rather than completely random because of traders effects.

 

[z106]

Zephyro. I trade my set ups with real money every day on the market so 'out of sample' backtest arguments are not relevant.

Yes I have often run 'confidence interval' stuff requiring 212 samples for 95% confidence Blah blah blah but trading in real life does not work like this. The confidence stuff is relevant to a sterile engineering environment where the inputs/outputs can be controlled but not in the market which is dynamic and suffers from 'trader effects'.

Once a set up becomes well known among traders it stops working. Hence in practical terms a trader does not have the luxury of sitting around waiting for hundreds of samples to determine if he can get an edge with this method. In my experience, a sample size of twenty is adequate in real terms for an indication that the set up will improve odds.

Backtesting - I am not a big fan of this as it is very difficult to capture all of the dynamics that will happen in the trading day. For example, I once had a trade set up with 452 backtests which had a very nice 3% edge per trade on average. When I tried to put it to work it failed miserably because I used end of day prices for the tests which did not catch the wider daily variation which often ran my stops.

Some interesting points there Lemur.

The first one re set ups having an expiry date is an interesting good point.

I think your 2nd point is an unfair point though as the error there was your backtesting was done incorrectly by you (for not including daily swings resulting in stop outs) as opposed to the whole concept of backtesting being flawed generally. (unless you take the time limits on valid set ups although presumably many setups have distant expiry dates).

Lemur - out of curiosity - what time frames do you trade?
Are you a day-trader or do you hold positions for many weeks or does it vary?

Also - out of curiosity - are you a professional trader?
DO you trade for a firm ? Or just for yourself personally in a professional capacity?

Aldo - how long does a set up last you before the market adjusts?

 

[lemur]

Some interesting points there Lemur.

The first one re set ups having an expiry date is an interesting good point.

I think your 2nd point is an unfair point though as the error there was your backtesting was done incorrectly by you (for not including daily swings resulting in stop outs) as opposed to the whole concept of backtesting being flawed generally. (unless you take the time limits on valid set ups although presumably many setups have distant expiry dates).

Lemur - out of curiosity - what time frames do you trade?
Are you a day-trader or do you hold positions for many weeks or does it vary?

Also - out of curiosity - are you a professional trader?
DO you trade for a firm ? Or just for yourself personally in a professional capacity?

Aldo - how long does a set up last you before the market adjusts?

Backtesting - v.difficult to take everything into account with it becoming over complicated & unwieldy.

Time frame - usually swing trades 1-3 days. But I change with market conditions. This is a whipsaw market so I trade it with small position sizes, wide stops and quick profits.

Set up last - depends on the set up. Can vary from one month to years. For eample, gap plays seem to have been around a long time.

Set ups are not to be confused with seasonality and cycles which have reasonably regular timeframes.

 

[nuac]

qwertyuiop - referring to your post re use of moving averages - trading on cross of 10 week and 40 week and on a 200 MA rising trend - are there any tests or other data to show these rules would work/

 

[z106]

qwertyuiop - referring to your post re use of moving averages - trading on cross of 10 week and 40 week and on a 200 MA rising trend - are there any tests or other data to show these rules would work/

I was just throwing those out as complete off the top of my head examples of typical rules that someone may have to create an edge.

I am not for a second saying those specific rules work.

That dosn't mean they wouldn't work of course- but highly unlikely.
As lemur mentioned - if the rules were that easy then everyone would be employing them which in time would make them redundant.

If you do want to backtest a specific set of rules there are a number of sites where you can do it.
One is www.wealth-lab.com.

A great forum for trading is ome called www.trade2win.com where a lot of trading questions will get answered.

 

[lemur]

can you define pseudo random for me?

Do you think academics that have decided they are a random walk are incorrect?

Pseudo random means its largely random with pockets of non-randomness caused by trader actions. I can show you lots of non-random things on a price chart.

Lets put it this way academics make lousy traders.

 

[nuac]

qwerty - thanks for the information re trading systems

 

[zephyro]

I think he states in no incertain terms that backtesting must be done correctly to verify an edge.

I wouldn't get too bogged down in his backtesting methodology as that misses his main point which is is that edges do exist and that trend-following is one way of establishing an edge.

Why doesn't he backtest correctly in that case?

You may be correct that edges do exist but have you any properly backtested verification?

 

[lemur]

Why doesn't he backtest correctly in that case?

You may be correct that edges do exist but have you any properly backtested verification?

Zephyro, I sense you are been too analytical about this. The only verification that counts is the profit & loss column. If mine is not green, I cannot pay my bills. That tends to concentrate the mind onto what works.

 

[limerickboy1]

just a note about edges.the simplest ones are usually the best ones

for example - if you used the one that says buy if the 100 day crosses above the 350 day and sell when it goes under then you would have gotten big profits on the dow jones in the last bull market. you also would have shorted the dow back in february and you would again be in profit.

FYI - trend following dosent work in the markets we have at the moment, this is great for us in the sense that most traders will adopt the belief that trend following is dead and it dosent work anymore. this is always the point at which it starts to work again. its quite possible that trend following wont work again for 12months + but when it does there will be big gains to be had.

there are loads and loads of edges out there and they are simple to follow. the problem people have is keeping to the system and their rules even when their emotions say that the system isnt working . being a successfull trader is mostly about sorting out your mind first.

 

[zephyro]

Zephyro, I sense you are been too analytical about this. The only verification that counts is the profit & loss column. If mine is not green, I cannot pay my bills. That tends to concentrate the mind onto what works.

Your approach seems to be to ignore backtesting and trade systems with real money, presumeably until they stop 'working'. I'm sure you realise there's a one in two chance that any system will record a profitable first trade no matter how daft it is. Without backtesting properly how do you have any confidence that a system has an edge and that any success you may have had isn't just luck?

 

[zephyro]

for example - if you used the one that says buy if the 100 day crosses above the 350 day and sell when it goes under then you would have gotten big profits on the dow jones in the last bull market. you also would have shorted the dow back in february and you would again be in profit.

It's very easy to pick a system that works in a particular sample. How does this system perform out of sample?