Citadel, Rentec, D.E. Shaw, William Miller all use the Kelly Criterion or half Kelly. If you understand the mathematics, the results are indisputable. It is critical that once you have a model that correctly prices assets, that when you bet, that you bet using mathematical money management to attain either: 1) very nice equity curves, or 2) the optimal way to grow your money on repeated trials with a positive expectancy. nitro
What I find fascinating is how many times the Kelly formula has been completely mistated, thus "justifying" the Chicken-Little screaming that using it will blow up your account.
Yes, it is often best to just go to the source: http://www.bjmath.com/bjmath/kelly/kelly.pdf William Poundstone wrote a very nice book that is a must read for every trader: http://www.amazon.com/Fortunes-Form...2468032?ie=UTF8&s=books&qid=1187571813&sr=8-1 I am a fan of 1/2 Kelly. nitro
I'm sorry for confusion and misunderstanding. After reading this forum for awhile I accidently found acrary's posts. I read almost all of them. The guy is a genius. Now I understand the difference between market character and real edge, and how to test ant track it. This is really what was missing in understanding of trading systems. Now as a newbie with only 2 years of experience, looking back I can point a few big mistakes newbies make: * thinking that all you need is to find a trading method and then you won't have to think about anything anymore, just trade it * relying too much on backtest results, and even forward tests * relying on a single trading system, so that when it fails, the whole trading process fails. Instead, use at least 3 trading systems with uncorrelated results (i.e. diversification across systems) * misunderstanding the edge. (Big thanks for acrary here!) When edge is defined and formalised, it can be tracked by simple statistical methods, so when a given edge diminishes, you stop trading it. * taking too much risk (this one is too common with newbies). Especially when one sees maximum historical drawdown of N points, he can think that it is very unlikely that the future drawdowns exceed 2N points. This is very wrong, markets change, and suddenly you get 3N drawdown. Now if you based your money management on the premise that you won't get drawdowns bigger than 2N and tune up your risk-per-trade accordingly, you are killed. Simple as that. Do not ever base your money management on historical drawdowns only, because markets change, volatility changes, and predicting future drawdowns is the same as predicting future price movement - futile. If you do, then know that someone that doesn't will take your money. * reading too much books about trading. Most of them are total BS. Not only that, but lots of them are written very very poorly. Avoid the likes of Larry Williams, especially stuff like his BS money management formulas and day-of-the-week/month optimisations. Read Alexander Elder, Mark Fisher, Toby Crabel, but just for overall understanding, don't trade their trading methods because I'm almost certain that you will fail. Think with your head, find your own edges and keep them in secret. I never heard of anyone that became a millionaire while trading a system described in a publicly known book about trading. * thinking that trading is easier than any other job. Wrong, trading requires a lot of intellectual and mental efforts. If you cannot handle it, stay in your 9-to-5 cubicle, it will be alot easier on your mental health and provide stable income. * overtrading. Thinking that the more trades will result in bigger compound result. Which could be true in ideal world, but... They choose daytrading, where comission/spread is quite big compared to profits and losses. The volatility and direction of intraday price is very unpredictable. It's certainly not suitable for inexperienced traders. Brokers like daytraders and advertise daytrading. I'm sure there are many more things, but I guess these are the biggest and most common mistakes of the newbies. Well, this post won't change anything, just reading it won't prevent newbies from making these mistakes. It's very hard to believe until you get hit with reality check. But that's okay, newbies just add more profits to already profitable traders.
Quote from nitro Yes, it is often best to just go to the source: http://www.bjmath.com/bjmath/kelly/kelly.pdf William Poundstone wrote a very nice book that is a must read for every trader: http://www.amazon.com/Fortunes-Form...2468032?ie=UTF8&s=books&qid=1187571813&sr=8-1 I am a fan of 1/2 Kelly. nitro, Hello, how's life been treating you? How's that young lady of yours doing? I'm feeling betrayed I hought you were a full fledged fan of mine? So which half do you prefer now, left or right, front or back, top or bottom, inside or outside? I'm feeling very torn in multiple directions as you can see.:eek: Hope all is going well. Take care, Good Trading, and God Bless! Kelly
Hi to all, Firstly, I can declare that I did some statistic researches about this problem: The NxATR(10 for example) StopLoss method. How to adjust properly the N parameter. I know how to do that, there is a math-statistic relationship and I discovered that every time frame has its own N. Again , for the same time frame , every instrument has its own N , extremely stable over long long time. Something like the human personality. Now , what I propose you is this: I'd like to have an ExcelSpreadSheet, or any other program who implemented the VanTharp Random Entry System. In exchange, I'll disclose with you these my extensive research results. The reason why is this: 1. I want to understant exactly the VTharp system .I read the chapter in his book, but.. it is not very clear: How large the initial risk, aso 2. I want to run this experiment using my scientifical approachment in choosing properly the N-parameter. I tottaly agree with you in saying that the key is money management ( managing the exit also...) so this is the reason why I'm aware for the fact that my results worths. Thaks in advance, Nafty
Thanks, gbos. This post and the example you gave here emphasize the importance of using the exact formula to determine the Kelly fraction. The approximation (p*W - (1-p)*L)/(W*L) grossly overestimates the true value in this case. Ironically the wrong but widely posted approximation p - (1-p)/(W/L) is actually safer and more accurate in this case. All of which points out a serious flaw with every piece of trading software I know about. Nothing but a spreadsheet with an optimizer routine can calculate true Kelly on the fly.
Yes, let's all pretend the three paragraphs I posted after that had nothing to do with the topic. Whatever your problem with me is, get over it, newbie. You're just a gnat buzzing around my head. And now you're on my ignore list.