I was discussing this with a trader today and wondering on some other people’s thoughts.
Money Management Using the Kelly Criterion
Supposedly with a win ratio of 60:40 and a RR of 1:1, we should make each trade with 20% risk (could reduce to 10% to allow for errors).
It seems common knowledge that traders use around 2%. Why the discrepancy between that and what the Kelly criterion says is the best risk amount?
I think you may be confusing win ratio with risk.
The probability of running into 5 losses in a row is about 1.5%. Psychologically, you may hit the 4th loss and be unable to place your 5th and final trade knowing full well you have a 40% chance to lose it all. That’s hefty risk if you plan to trade Forex long-term.
The reason that I advise trading a small position is because most new traders haven’t built up a trade log with which they can accurately assess their trading performance. The two statistics you need to calculate K are: Win Rate and Average R, both of which require a large enough n for them to be accurate statistics.
I personally am a fan of the Kelly Criterion. It is a great method that allows you to maximize your risk based on your probabiliity of success. In fact, it is mathematically impossible to bust an account if you have a winning edge and utilizing this tool.
The difficulty is defining win rates. You would need a sizeable sample size to determine that. And even then be sure that it is something that continues to be updated as your account size raises and decreases.
Nice job in bringing this up.
good topic
here is a link to the applet that made the chart in that investopedia link.
Random Equity Curve Simulator of a trading system. Learn it before you trade
No.
This is incorrect as you resize your bet size on each turn. For example, if you lost 10 trades in a row at 10%, you would still have around 30% of your account left. Obviously, not great psychologically but that’s part of what we’re going to discuss.
1,000 -10% = 900
900 - 10% = 810
810 - 10% = 729
etc.
That’s the worry, I know in the article it says your last 5-60 trades but that’s not a good enough sample size.
Also, it would need to be for 1 specific way of trading wouldn’t it?
It’s no good if as part of your samples you do a bit of scalping, a bit of trades targeting 40-50 pips, some swing trades lasting days/weeks with much higher RR’s let alone the day to day biases/mistakes that we live with as traders.
I realise that is why some use the half kelly principle.
However, the common knowledge of 2% risk is far different from a 60% system with 1:1, which Kelly says should be traded at 20% risk. I did know of a trader who used to risk 20% on their system from time to time.
Most people in trading seem to used fixed risk trades.
There are some problems with it and I need to do a bit more research on Google.
It seems to be common for stock investors but maybe that’s just because of when it was published. You can put 10% capital into an equity but the risk of the stock going to zero is very small. However, it was originally designed for horse racing.
some optimisations include: Position Sizing Methods / Money Management Models
You don’t even need the Kelly Criterion if you’re going to using a fixed fractional equity model.
How do you determine your ideal trade size?
I’m just interested to know why traders accept 2% risk as the “common” trade size when mathematically it may not be ideal at all.
Of course psychology comes into it at some point…
Interesting read and applet.
Says I should use 33% per trade for my current system, but of course also says not to go higher than 20%-25%.
Think I’ll demo this for a while and see how it turns out. The pyschology involved would really hit with a loss, and knowing me as I do, I’m not sure I could take more than a couple of losses in a row with such a system.
Still, pushing up to 5% would be very do-able, so will see how it rolls.
Cheers for the links!
Yes, I agree fully. And I’m a fan of the Kelly formula as well. Someday when i have the time I’m going to open a small account and trade with Kelly money management.
Very cool formula. Thanks for posting it up!
I’ve been under the school of thought for a while that as the account grows, so does my willingness to risk a bit more. Not in the compounding sense, but from the perspective of using house money.
If my original investment is safe, why not? I use one set of entry rules, so it makes it easy to quantify results. All that fluctuates is profits. Some nice, some teeny, but very consistent.
Thanks again! Great find:)
It should be pointed out that absolutely none of the assumptions on which the kelly criteria are based actually hold true with respect to the distribution of returns from the typical trading system, so I’d advise against using it out of the box.
Such as what out of interest? Could you list a few examples so we could have a look…cheers
This is worth a read for those interested: Blogger: Quantitative Trading - Post a Comment
A small quote: [B]The Kelly’s formula you quoted is for betting with binary outcomes only. In the case of continuous finance, please refer to the formula I wrote in the main body of my article which comes from the paper by Prof. Ed Thorpe cited at the end of that article. In the continuous finance case, f can most definitely be greater than 1, otherwise it would be quite useless to the hedge fund community.[/B]
Kelly assumes sequential bets that are independent (later work generalizes to bets that have sufficient independence). That may be a good model for some gambling games, but generally does not apply in investing and other forms of risk-taking.
WIKIPEDIA for adjusted models: Edward O. Thorp - Wikipedia, the free encyclopedia
Note the problems with correlated bets and sequential bets.
Those taking multiple trades per day all running concurrently and on the same pair should have a thorough read.
The Kelly criteria, assumes the following:
[ul]
[li]All losses are exactly the same size
[/li][li]All Gains are exactly the same size
[/li][li]The distribution of losses and wins is perfectly stationary
[/li][/ul]
None of these assumptions are generally true under realistic trading conditions, so whilst its simple enough to evaluate an optimum historic value, doing so in real time becomes impossible. The real problem is that returns are extremely sensitive to even temporary changes in the distribution of returns.
I suppose that if you can look at it in terms of buying a lottery ticket then using something like optimal F is probably well worth a punt on a small account. I’ve only played with these ideas in backtesting but though the drawdowns are horrendous, you usually find that in actual dollar terms, that account equity always remains above what you’d achieve by using less aggressive position sizing.
I don’t agree with that. It calculates the average loss and the average win. Furthermore, it’s something that should continually be adjusted with each trade to maximize the potential.
I know an awful lot of mathematicians who would disagree with you, but good luck with it 
You’ll have to do a google search as I cannot recall where I found it and not really care to look at the moment, but someone calculated the efficiency of kelly betting when you place sequential bets.
It lets you know how much you should risk per trade if say you plan on making three trades at the same time. This in turn provides Kelly betting with a conservative angle.
Maybe tonight, I’ll dig up that calculation and add it to this thread.
Could you share your findings? This isn’t a pissing match, mate. This is a discussion of the Kelly Criterion. And if you have information that could assist in the knowledge of the topic, then it would be an added value to this thread.