Choosing Position Size - 1%, 2%, 3%, 4...?

How do you choose your position size percentage?

Plenty of dudes will say “Never risk more than 2% per trade.” or “Thou shalt not risk more than 2%

But how do you choose 2%? What if 2% is TOO big?!?!?

My thoughts:

Look at a back-test for a given entry system and see how many trades it will produce over a given period. Imagine that all of those trades were losers. If the period produced 20 trades and you lost 2% for each trade, you would have lost 33% of your account over that period (2% losses compounded 20 times).

For daytraders, if you have a system generating 20 trades a week, risking 2% each in a nasty string of losses would lose you 33% of your account in just one week.

For long-term traders, if you have a system generating 20 trades in ten years, risking 2% each in a nasty string of losses would lose you 3.95% each year.

I hear a lot of people throwing out percentages without ever talking about how they get those percentages. Everyone is talking about how to calculate the right number of lots to put on 2% risk without talking about whether 2% is a lot or a little.

Anyone have any good methods for choosing that percentage?

3 Likes

It all depends on your strategy, account size ,SL and RR ratio. The more accurate is your trading(low drawdown) the more risk per trade you can take. But everybody is different, 2% is just basic rule of thumb.

Can you be more specific? How do you quantify accuracy and adjust your percentage risk accordingly?

I want to make sure my drawdowns don’t draw down into my beer money.

Here’s an interesting fact. It takes just as many winning trades to recover from a losing streak reguardless of risk as a percentage. Personally (except when I’m “scalping”) I trade fixed lots equal to my account unleveraged. That means a 200 pip movement against me equals two perecent of my account. With SL usually set between 60 to 80 pips it leaves plenty of room to manage the trade and allows a basket of pairs to be traded.

Pretty simple

Quantifying accuracy comes from either backtesting of forward testing your trading system and risk management processes. That’s where you get the numbers that thesecret mentioned and then you can adjust your positions to whatever your risk tolerance is. Like bobbillbrowne said, it’s pretty simple, but there’s work you gotta do first before you can get to setting proper position sizes.

1000 minus 30% = 700
700 plus 30% = 910
700 plus 43% = 1000

If you lose 30%, you must gain 43% to get back to even. Unless… Are you saying you do not bring your position size down during a drawdown? But that would mean you are taking on bigger and bigger percentage risks as you go into drawdown. Right?

1 Like

Be specific.

Example: “This system lost 4 trades and won 16 and landed on 10% up with a 6% drawdown with a fixed position size. Therefore I will double that position size to get a 12% drawdown and a 20% gain over the same trades.” Isn’t that just curve fitting? As soon as you do that you end up with 7 losers in a row and a 39% drawdown.

Na bro and you know how the maths work lol. Now I dont have the spreadsheet in front of me so I’m going off memory. If you risk 2% and lost 20 trades in a row your drawdown would be like 35% approx. It would then take 21 winning trades in a row to recover. Now if your risk was 5% drawdown would be around 70%. It still takes 21 wins in a row to recover. In reality however we all know that its easy to loss 20 in a row. Its a lot harder to get that 21 nett winners back but.

Editting note. Based on a 1:1 R:R

Oh, yes. I am with you. OK. There is some slight loss but you are right. Roughly 21 wins in a row would cover 20 losses in a row if the same % was risked each trade.

And this gets to the OTHER side of the question. Am I risking ENOUGH? In other words, how high can I run the percentage before I will risk running into a drawdown so deep that I can’t continue the position sizing algorithm? And THAT is what I am getting at.

Over the last year I heavily overtraded in order to play with position sizing percentages and was SHOCKED to see how high I can go. I think I have a good methodology for making my decisions. But I would love to see what other forex dudes do. EXACTLY.

Personnally, I dont place much weight on % risk or R:R ratio these days. My belief is these things are myths exploited by marketers, brokers and IB to get people trading their small accounts. I believe one must work out expectancy of ones sytems as an unleveraged system and work backwards to get correct trade size. This involves a lot of hard work either backtesting or forward testing to collect data then analyzing that data. I’m currently at work but when get home tonight I’ll explore this concept some more for criticing.

This why it’s good to have a daily loss limit don’t you think if you lose 20 trades that’s enough for the day 20 trades a day is considered hft, but really though proper money management risking 2% of 1k for example is \$20 if you have a daily loss limit of \$100 and have enough discipline not to exceed that amount you should be fin especially if your splitting that \$100 dollar loss in multiple trades and increasing your odds in profiting , you a robot or a pro trader will set there and lose 20x in a row…THE definition of insanity is reapting the same thing over and over and expecting different results wouldn’t you say that’s close to insanity there ?,

If one sufferred 20 losses in a row then I would stop trading the system altogether. The point was to illustrate that it doesn’t matter what %risk you choose, after “n” losses in a row it takes “n +1” trades to win it back. The problem being what is the correct %risk for your strategy. And when I get home tonight I’ll give my method based on expectancy.

I don’t like the loss limit idea. Suppose I hit my loss limit and quit trading for the rest of the day (or week, or month, or whatever) only to miss the trade that would have put me back in black.

Also, I hear people arbitrarily choose drawdown limits: “I will only accept a drawdown of x%.” But that arbitrary choice may actually limit their profitability just as much as it limits their risk. Plus, once they have hit it what will they do? All they can do is draw another line in the sand.

So what drawdown am I willing to tolerate? The maximum drawdown I can take without preventing me from continuing my position sizing algorithm (how I would quantitatively define “blown up”). (Give me thoughts if you have a different definition or ideas).

I closed 13 trades in the last 30 days. My trade frequency with my current system should be roughly 200 trades per year (this would be the grand total within all pairs). Suppose I bet 2% on every trade and lost 1R on all 200 trades (worst case scenario, yes extremely unlikely). I would have a 98.24% drawdown for the year. Suppose an 82% drawdown would prevent me from carrying on my position sizing algorithm. I would need to reduce my risk per trade to 85 basis points to arrive at an 82% drawdown after 200 losses in a row.

So here are my steps thus far:

1. Quantitatively define maximum drawdown (highest it can go while continuing position sizing algo).
2. Determine number of trades expected over given period with given system.
3. Calculate % risk per trade to prevent a drawdown above maximum with solid losses for period.

I have a few ideas about where to go from there but I want to get feedback at this point.

Thanks for taking time to go over some of the nerdiest shix ever with a dude that ate a lot of mushrooms and drank a lot of beer down at the lake this summer.

I get that. I am with you now.

And yes I typically hear about the profit expectancy of the system in relation to these considerations (that is in Van Tharp’s work). My objection to that method is that the expectancy is curve fitted to a given period. This essentially guarantees that choosing a percent risk per trade based on that expectancy will either produce a greater than expected drawdown (but possibly a greater than expected profit) or a less than optimal performance (but with a lower drawdown).

I think I may have a better method that will allow me to not only risk more or less in an absolute sense as the series of trades progresses (as do fixed fractional position sizing algos), but also in a percentage of equity sense.

I do not have a strict percentage MM plan when I trade.
First of all, I adjust very much of my risk according to what what kind of strategy I use and what instrument I trade.

Risk is always my second question when I trade. If I get a setup my first question is - where will I know that I am wrong.
That gives me a figure of pips that a trade will move against me before I take it of the table.
The second question is risk. How much am I willing to pay to see if my trade work out or not. (NOT %, but how much money)
Now I have both numbers - pips to loose, and how much money I am comfortable to loose if that’s a loosing trade.
That gives me my position size.
I know many uses percentage as a “yardstick” but I don’t.

I ask myself in the morning how much am I willing to loose today before I walk away…and that usually ends up on 5% of my total accounts.
How I manage to spend/risk those 5% during the day varies according to what I trade, and what setups.

MM is just as personal as trading style, find what works for you and so on… It has to fit your trading style somehow, but what MM works for your next door trader, might not work for you caused by several reasons - you have different strategies, different amount of money to trade with AND our mind respond differently to risk and loss.

I understand the question - but that is just as easy to answer as how long is a rope.

I think that is the most common answer. And it may indeed be the BEST answer. I am wondering if there are any traders here that get a risk percentage from a systematically defined process.

There is a good article in Active Trader about this question but it doesn’t answer the question. It only points out the the effect of changes in risk percentage in a back test, it just offers food for thought. I should send you guys a copy of the article… hmmm…

Well bro, don’t think I can learn you anything but I dare say you can expand my knowledge so lets see if I can provide a different approach.

So how can expectancy be used to calculate risk. Well I’m no teacher but I’ll apply the logic to two systems, one I’m trading, one just progressed into the live forward testing. So first to calculate expectancy we need some hard data. Two ways to obtain that, exhaustive back testing or exhausted forward testing. Of cause nothing beats forward testing in a live account but many don’t make it to this stage.

So those who know me know that I have a once a day tick trade strategy on the EURUSD. Last year It yielded 68 winners @ 10 pips and 52 losers @ 10 pips. This year its evolved and now I have secured 15 wins for 205 pips and 6 losses for 47 pips. Totals 141 trades, 83 winners - 885 pips, 58 loses at 567 pips. So I had 59% winners at average of 10.6 pips, 41% losers at 9.7 pips. Knowing pip value is important as it all starts with knowing what your system can deliver un-leveraged.

So the formula for expectancy is

(% winners * avg win) - (% losers * avg loss)

In my case

(0.59 * 10.60) - (0.41 * 9.7) = 2.3 pips.

So over the course of my trading year I’ll make 200 trades at expectancy of 2.3 pips can expect a return of 460 pips. So on a un-leveraged account of \$10 000 USD traded at 0.1 lots yields a return of \$460. Less my trading cost (comm at 70c in/out) \$140 and net profit of \$320 or 3.2%. Man that’s just not worth the head !@#\$. So I apply leverage to help out. Industry standards suggest that 5:1 is the max we should use but my ballz are slightly bigger so I’ll use 10:1. So now I’m trading with one lot. End result nett return \$3200 or 32%. Now that’s worth my time. But just as leverage multiplies my earnings it also does the same for my loses. Lets explore that.

Now as a stand alone trade this represent a \$97 loss, just less than 1%. Nothing to sad there. But what about a losing streak, what sort of draw down can I expect. Now there’s a fairly simply formula we can use to predict losing streaks. I forget where I got it but it seems to work.

CL = ((log n / -(log p)) +1

where CL = Max consecutive loses
n = Trades in sample size

So over the course of a years trading, n =200, p = 0.41 so

cl = (log 200 / - log 0.41) +1
cl = 6.94

I can therefor expect to loss 7 trades in a row once a year. Total just under 7% quite manageable.

But we all know trading doesn’t quite work that way. What if I’m hit with a 1 in 10 year event, what to expect then. Now over 10 years I’ll make 2000 trades so plug that into our formula and I’l expect 10 consecutive losing trades. Now the markets a ***** so this one in 10 year event undoubtedly will be followed by my annual event and I’ll suffer 16 consecutive losing trades. A draw down just under 16%.
Now I believe that’s quite manageable. I actually use 20:1 leverage with this system so can expect a 32% draw down at any point.

Now for my second system. This has gone through many hours of back testing by myself and a third party. Results from last year where 63 trades, 42 winners (66.7%) @ avg 105 pips, 21 loses (33.3%) @ avg 90 pips. Lets plug those figures in un-leveraged.

Expectancy = (0.667 * 105) - (0.333 * 90)
= 40 pips

So over a year my expected return is 67 trades * 40 pips = 2680 pips or 26.8%. If I applied 4:1 leverage my return now tops 100% a year, man that sounds good.

So lets look at loses. At 4:1 leverage a single trade risk 3.6% of my account. My once a year event would then calculate as

(log 67 / - log 0.333) + 1 = 5 consecutive loses

Thus expecting a draw down of 18%, not bad. My ten year event should be 7 consecutive loses therefore my black swan event can be expected at 12 consecutive loses or exposing me to a draw down of 43%.
My ballz are big but they’re not quite that big. Best I reduce my risk to 2:1 leverage. Still going to return 54% for the year but now I only risk 1.8% and can expect a 21.5% draw down.

Hope that makes a bit of sense bro. Of course there’s a few more variables one must include like currency conversions ( I have an AUD account, these calculations are based on an USD account) or if your trading the system in a basket. But I hope you get the jest of it.

Keen to heard your thoughts particularly where errors might occur

1 Like

Totally with you. I get entirely what you are doing.

My first trouble in doing what you are doing: I do not have profit targets, just trailing stops. So whether I am looking at a back test, or live trade history, my profit results are purely period specific. With that, I can only calculate a range for potential losses.

In my case,

(% winners * a period specific value offering no future expectancy) - (% losers * average loss) = a period specific value offering no future expectancy

So really all I can do is implement a % risk that will keep me out of a blow up. I can’t actually put a number down as an expectant profit percentage or pip value or anything.

Exactly. Suppose a trader like me that does not have profit targets does a backtest with percentage risks of 1%, 2%, and 3% and gets the following:

Suppose that a drawdown of 20% to \$800 would be sufficient enough to disable him from trading his system. So naturally, after seeing these backtest results he will not choose to trade a 3% per trade risk because 3% put him in a +20% drawdown in the backtest. He may choose to trade 2%, or perhaps be fearful enough to choose 1%.

Then, lets say he trades in real time and gets the following results:

Because he chose to go the whole period at 1% or 2%, he made less profit than if he had chosen 3% which would have not actually killed him with this run of trades in real time that did not produce the +20% drawdown. Most traders would say: “Hey I will take the performance I get if I am not risking a blow up.” (And I am one of them). I think this is simple enough and most traders get this pretty easily.

Lets now imagine an added option. Suppose the trader sets up a rule that says:

“Start with 2% and risk 50 more basis points on the next trade if you win or 50 less if you lose. Maximum risk is 3%, minimum is 1%.”

That variable percentage risk algo compares thus:

The red line shows the equity curve of the variable rule algo taking only a slightly greater drawdown than the 1% fixed algo and it almost performs as well as the 3% fixed algo. Am I dreaming? Should I just trade 1% and shut up? I have a few different ideas about variable percentage risk rules that I think could make a serious impact and I think some of the bigger market players use variable percentage risk algos. Anyone have any info on that? Anyone ever try this?

Am sounding too sober? More beer? More acid?

P.S., imagine that my cruddy graph shows the 3% not quite going under \$800. I added the variable algo to the backtest and not the “real” set of trades. But you get the idea. I should make up a much more detailed set of images with the data so everyone can see the data and how it all plays out exactly.

I went in search for the article this morning but did not find it. They pulled it from the shelf. It was either a January or December issue.