So you want to trade successfully?

[Knobby22]

Ordered a copy.

I trust Techs judgement.

 

[Twiddle]

That booko site is great.

Cheers Razza.

I am nearly finished the book, it has been very educational, glad I bought it.

 

[SmellyTerror]

Shouldn't the result read out expectancy per year for example? or is it per amount of trades?

Or is it like expectancy in terms of stats like as trades tends to infinity expectancy tends to E(x)?

Dunno if the answers you got made sense to you. I’ll give a long answer where I’ll ramble all over the place and tell you stuff you already know, even though I don’t have the book. Hey, sometimes hearing the same thing in different words helps.

Expectancy is per-trade. So each trade has X expectancy, and that’s how much you’ll expect to make per trade (either in dollars or as a proportion of capital put in), on average. It's then of benefit to have a system that trades often, since that'll make you more money (allowing for expenses).

So if I've got a system with an expectancy of 10% where signals only come along once a year, I can put $10,000 in each year and expect to come out with $11,000 (on average).

If I've got a different system that has an expectancy of 5% that gives signals 100 times per year, I could put $10,000 in each year and come out with more than a million dollars. I could use a million dollars. I have enemies. So the 5% expectancy system kicks the crap out of the 10% one thanks to frequency...

But now we come back to risk of ruin. I'm putting 100% of my money on the line every trade. If that 5% expectancy system wins about 50% of the time (but wins enough compared with losses to get to the 5% expectancy), and I put all my money in every time for 100 trades, the chances that I'll lose all my money is pretty much 100% - I only have to lose once and I’m out. I’d need to flip 100 heads in a row – not going to happen. That is, I've turned a great system into one that will certainly lose everything I have, and my wife will set my testicles on fire.

That is known in the biz as a Bad Outcome.

So you want to maximise how hard you hit a good system, but at the same time make sure you don't get wiped out. Risk 101.

IMO, there are three main considerations. (These assume you’ve got a system with a positive expectancy. No point worrying about risk if it’s negative. Might as well pile your money up in your yard and light a nice bonfire).

First, the risk everyone needs to worry about is a run of bad trades. In a 50% win system you really have to be able to account for a run of a dozen or more failures. Seriously, spend an hour or two tossing a coin and record the results. It’s edumacational. See how big a run of tails you get in that time. It’ll be big. So say you want to survive a run of 15 failures. And that might be in a period where you got a few more losers than winners anyway. Suddenly you can see why people like to limit themselves to 1-2% risk per trade, tops. Who wants to watch 45% or more of their account ablate defending a perfectly good system? And that’s when you know your system WORKS and has a win rate up at 50%! A lot of very good, profitable systems don’t win anything like 50%. In fact, looking back at wanting high frequency stuff, well, those systems are usually a lot lower than 50%. Which makes the runs of bad luck go even longer.

So estimate your chances, and make sure you don’t risk too much per trade to ride out a slaughter.

The second thing for ROR is maximum position size. If you're like me and hit margin like hobo hits a bottle of cooking-sherry, you need to worry a lot about maximum position size. It's all very well to risk 1% of your capital on a trade, but if you’ve got a tight stop, you might well end up with a “correctly sized” position that’s bigger than your account.

Example: ABC trades at $20, and I can put a valid stop at $19.95 – SWEET, a 5 cent stop! I rule.
1% of my $20,000 account risked = $200 risk.
$200 divided by 5 cents risked per share = 4000 shares.
4000 shares * $20 = $80,000!

Now I can afford that with my $20,000 account, thanks to margin, and it conforms to my other rules, only 1% risk, so in I go! Nice and safe. But hang on – if ABC goes into a halt tomorrow, uh, what then? And if it comes back on the market at $10 after a crushing announcement, or doesn’t come back on at all, I’ve just lost more than my entire account. I’m losing my house. I’m sourcing burn cream. All for a trade that officially only risked $200.

What are the chances? If you do this trading thing long enough, eventually you’ll be holding the bag when something like this happens. You HAVE to be in this for the long haul – this WILL happen some day. Bank on that. And it might be your next trade. So this is where a lot of people put a limit of, say, 20% or 12.5% of total account size in any one position. Not knowing your formula, I’m guessing that’s where you got that number, since it’s a number that makes sense.

Third thing is diversity, or the total of related positions. I’ve got 1% risk per position, and less than 12.5% of total account on any one position, so I’m safe. But I’m holding 10 iron ore producers. Uh, hang on. Aren’t they going to be related? Yeah. In a lot of ways you have one enormous trade with 10% risk and 100% of your account size. So when news breaks of a new mining tax, or civil war breaks out in China, and the price of iron ore tanks, so does your account. You could be looking at a gap that, even without margin, murders you on the open.

Well the same sort of thing goes for different scales. At its broadest, look at long vs short. You may have your holdings well diversified, but if it’s all long and a plane hits a building, you’re still going to get hammered. Again, it’s an issue for everyone, but much more important for the highly leveraged. If I’m fully leveraged at 10 to 1, all long, and the markets drops 10% on the open, I’m bust. And if that seems unlikely, again, remember you’re playing for the long run.

Anything that can happen, will. Bank on it.

So don’t commit to one thing so heavily that it can kill you. Most simply, you could avoid margin and, further, keep a large part of your account in cash. More complex, perhaps more profitable, is to hedge – best done (IMO) by finding genuinely (if marginally) profitable trades to take that will also respond to a black swan event favourably (eg going short in a weaker, but related stock to one you already hold long). Could be you’ve got a great system that only works on mining stocks and only goes long. Ok, then see if you can afford to take out some insurance by shorting the index – run the numbers, see what you can afford. If you’re picking the standout stocks, could be you’ll still be profitable with a partial hedge, and you may save yourself from ruin some day.

You don't need to be too paranoid about diversity, IMO - just keep an eye on it, hedge profitably where you can, and only do more than that when you're too heavily invested in one direction (particularly if heavily leveraged, and paying particular attention to total long vs total short).

All three factors here effect your risk of ruin.

Sum up:
Risk per trade: covers you against a string of losses, to be expected in any system. You should go into a trade knowing where an appropriate stop is, and use risk to size the position for you.

Position size: covers you against a single share or instrument suffering a catastrophic move. This imposes a limit on the size of individual positions.

Diversity: covers you against a black swan event. This imposes a limit on what kind of positions you can take.

Does that help?

Disclosure: I don't actually know anything.

 

[opulence]

Wow cheers for that reply. You talk a lot of sense kind sir.

In regard to your ABC example of the 80% drawdown and ROR. Does the same apply for something like forex that doesn't really gap around the place like stocks can given the liquidity and 24 hour market?


Also, I would imagine if I could trade forex without margin then I would be rich enough not to need to trade it.

How do you hedge risk in the forex? Does market risk exist or is it a zero sum game?

Once again thanks for that, really cleared up whats been bothering me.

 

[SmellyTerror]

Welcome.

Yeah, a lot of that goes out the window with Forex - and though there will be moves too fast to hit your stops, or even come close to hitting your stops , you don't need to worry too much about the total position size / diversity stuff. In fact, it's probably best to focus on one pair.

It does depend a bit on timeframe, though. Overnight positions can do some nutty stuff, and you've always got the weekly hop. My (horribly limited) experience with Forex is in scalping, which is as close to pure position sizing as you can get. IE: how much do I stand to lose per trade? My system stands to lose 10 pips on any entry, so my position size is pretty much the same all day, every trade. It only changes when my account size does.

One bit of advice for day trading, if that's what you're looking at, that Trembling Hand says a lot, is to keep a daily stop loss. Have a DAILY stop of, say, 2%. That makes you start small, go careful, and then increase to full size positions only after you're comfortable and have a bit of a buffer. Prevents you from going all-in on a day when your head isn't in the game and then getting into a frustration-fuelled losing cycle.

 

[Wysiwyg]

Expectancy is per-trade. So each trade has X expectancy, and that’s how much you’ll expect to make per trade (either in dollars or as a proportion of capital put in), on average.

I have yet to understand expectancy. I understand profitability over time but can't grasp the concept of "expecting" a preconceived amount every trade.
Sure I want to end up with more than I opened the trade with, but can a figure be put on that?
I am supposing a profit target of 10% above buy price would be a positive expectancy and a stop loss of 10% below buy price would be a negative expectancy, per trade? How can expectancy be arrived at?

 

[opulence]

I have yet to understand expectancy. I understand profitability over time but can't grasp the concept of "expecting" a preconceived amount every trade.
Sure I want to end up with more than I opened the trade with, but can a figure be put on that?
I am supposing a profit target of 10% above buy price would be a positive expectancy and a stop loss of 10% below buy price would be a negative expectancy, per trade? How can expectancy be arrived at?

Its more of an averages thing. On a single trade you are relying on your strategy to give you the best chance of achieving the profit target. However, things don't always go to plan and as time goes on you assume that some trades will trade very profitably and most will trade with small losses. When averaged out you can achieve your expectancy. Or mathematically:

Expectancy = (P(Win)xAverage Win) - (P(Loss)xAverage Loss)

 

[Wysiwyg]

Its more of an averages thing. On a single trade you are relying on your strategy to give you the best chance of achieving the profit target. However, things don't always go to plan and as time goes on you assume that some trades will trade very profitably and most will trade with small losses. When averaged out you can achieve your expectancy. Or mathematically:

Expectancy = (P(Win)xAverage Win) - (P(Loss)xAverage Loss)

Okay thanks. A different number will be arrived at with every change of tested data (time period) but if the number remains positive then it is considered one has a positive expectancy. Another way of showing profits exceed or not exceed losses over a specified time.

Thanks for your explanation.

 

[Wysiwyg]

For example in one test I have an expectancy of $837.82 per trade but advance forward one month and that number reduces to $678.95. To me quantifying expectancy lives in the unreal world of trading. It can only be positive or negative during the time it is measured and is forever changing.

 

[opulence]

Another way of showing profits exceed or not exceed losses over a specified time.

Thanks for your explanation.

Exactly right. As SmellyTerror said. If you do the calc and it comes up negative you might as well use your money as darb paper.

 

[opulence]

For example in one test I have an expectancy of $837.82 per trade but advance forward one month and that number reduces to $678.95. To me quantifying expectancy lives in the unreal world of trading. It can only be positive or negative during the time it is measured and is forever changing.

Expectancy like any other statistic approaches its true value as the number of entries in the data set approaches infinity. So the more data you have the closer your expectancy will be to its actual value. As with all average based calculations there will always be variance day to day etc.

 

[SmellyTerror]

Expectancy like any other statistic approaches its true value as the number of entries in the data set approaches infinity.

Yeah: that.

Test for as long a period as possible. The trick, of course, is that systems can have good periods and bad periods. Some systems stop working one year, then start working again a couple of years later. Some systems only "turn on" during some conditions - but then you've got to be careful in backtesting to not over-optimise. You can easily make something that'll give you insane returns over history, that doesn't do a thing into the future.

My system might be "BUY BPH at 0.020 and SELL at 0.400 - use a LOT of leverage", well that system tests pretty well. Might just be a wee bit over-optimised.

But if you look at a system that has operated, or been fairly tested, over a period of several years, you can get a good idea of expetancy. Of course, it's only EXPECTED returns. Could be your system is already broken and you just haven't noticed yet.

...all the more reason not to position-size yourself to the point that you'll get murdered finding that out.

For more on fair backtesting and expectancy, I'd recommend Howard Bandy's stuff (on this forum, his website, and / or his books).

 

[tech/a]

Just a few more practical points to add to the excellent information presented from that smelly guy and his well healed mate.

The Stinking--Rich work well together!!

I think it important to be able to trade in both directions/long and short.
I also think it important to understand the nature of the system your building.
(1) Long term less winning % greater Reward to risk
(2) Short term higher winning % less reward to risk---frequency

When designing a system understand the conditions you are designing it for.
(1) All conditions bull and bear
(2) Bull only
(3) Bear only.
Many have unrealistic expectations of all systems working in all market conditions and speak of robustness.

I note not a great deal given to forward testing of a system/market condition nor perhaps (Particularly in stocks) another bourse.
Not so for say Currencies on Indexes.
Currencies currencies---index index--Yes.

 

[gav]

Many have unrealistic expectations of all systems working in all market conditions and speak of robustness.

Do you mean knowing when to turn your system on/off? Like having a "system stop" (which Penfold mentioned in his book), or a filter that will turn off your "long only" system in a bear market?

I note not a great deal given to forward testing of a system/market condition nor perhaps (Particularly in stocks) another bourse.
Not so for say Currencies on Indexes.
Currencies currencies---index index--Yes.

Could you please elaborate?

Cheers

 

[tech/a]

Do you mean knowing when to turn your system on/off? Like having a "system stop" (which Penfold mentioned in his book), or a filter that will turn off your "long only" system in a bear market?

No I mean expectations.
Most develop long only systems (initially) and are disappointed when they don't perform in bear market conditions--you speak of additions to a system which may give it a better overall performance---your expectation is that it wont preform in market conditions which don't suit it---correctly I might add.

Could you please elaborate?

Cheers

Indexes behave differently to currencies.
Its been mentioned here by others and from my own observations they have their own life form.
Quicker and more volatile than Indexes---generally.

 

[Temjin]

Indexes behave differently to currencies.
Its been mentioned here by others and from my own observations they have their own life form.
Quicker and more volatile than Indexes---generally.

It's no wonder when 70%+ of the volume traded in the equity indexes are from high frequency institutional traders who love to flash trade in milliseconds range. Where as, the penetration from the HFTs in the currencies/currencies market is still fairly low to insignificant due to the vast liquidity and transparency of the market. (There was a report done from zero hedge on this, but don't know where it is now)

Very different markets indeed.

 

[sparfarkle]

Gday all,
Just wanted to try and put some perspective on the expectancy debate,bear with me if you can.
A simple backtest of CSL from 15/6/94 to present, using a slow stochastic 8 indicator, buying on turnup from below 15 and selling on a turndown from 98 returns 111 trades,87 good and 24 bad.
Average profit per winner 6.94%,average loss per loser 7.05%.
So to determine the mathematical expectancy (m.e.);

m.e.= (1+profit ratio)x win ratio-1
Where profit ratio = ave win * ave loss, and win rate = nos of wins * total nos of trades.
So 6.94 * 7.05 = 0.984 +1 = 1.984. 87 *111 = 0.784.

m.e.=(1+0.984)x 0.784 = 1.555 -1 = 0.55me
0.4% commission included in original calcs.

To apply this to monetary values,assume $20,000 position for all 111 trades,giving an average loss per losing trade of $1410.This then becomes your risk per trade;

So 1410 x 0.55me (Van Tharpe) = $775.5 expected average return per trade.

Apply this to the average nos of trades per annum i.e. 111*16 = 7.

So 7 x $775.5 = $5428.5 ave expected return per annum.

Any comments appreciated.By the way,fantastic post earlier smellyterror,i printed it off it contained so many simple truths that are so easily forgotten.

Cheers,
Sparfarkle

 

[wayneL]

Indexes behave differently to currencies.
Its been mentioned here by others and from my own observations they have their own life form.
Quicker and more volatile than Indexes---generally.

250 period statistical volatility:

S&P 500 - ~18.5%

EUR/USD - ~10.5%

Numbers don't lie.

 

[skyQuake]

250 period statistical volatility:

S&P 500 - ~18.5%

EUR/USD - ~10.5%

Numbers don't lie.

Thats cause during the AUs session NOTHING moves lol.

Also did that include the flash crash? Could be skewed

 

[wayneL]

Thats cause during the AUs session NOTHING moves lol.

Also did that include the flash crash? Could be skewed

That's 24 hour sessions, nothing to do with AU session whatsoever.

Flash crash affected vols in both, but peak 250 stat vols for interest sake(circa Aug 2009)

S&P 500 - ~47%

EUR/USD - ~16.5%