Perhaps I can be slightly clearer in my comment. Betting on a coin with a 60% probability of coming up heads, you should bet 20% of your bank IF you want to maximize returns. This isn’t my analysis, this is Kelly. I understand, for many reasons, a trader may not want to maximize their returns. For example, they might perceive that the risk of ruin is too high, or not like the drawdown, or not be comfortable with the bet size. Bet sizing is a theme that’s been interesting to me for many years. Largely because I haven’t got a great answer. How can it be that some traders are very comfortable taking huge bets? While for others, that comfort is only achieved when the bet size is very small? I fall into the latter group, and in much of my trading don’t risk more than 0.5%. Do I subconsciously know that my entries are only marginal at 50.1%? So how do traders taking a big bet get comfortable? How did Soros know how much risk he needed with his British Pound trade? Or how did Buffett know his bet size on Coke? I know there are many factors in play, but they knew to be aggressive. And they knew it was not a standard 2% bet. Sure, a trader can always bet a maximum 2% and get by (probably very well, if their signals are at 60% probability!). But 2% is not the right risk for this probability of bet, if you want to maximize your capital growth, and Soros and Buffet are not ‘getting by’.
that is one of the basic reasons why some very profitable traders never become billionaires. or in other words: that's why eternal compounding is a fake argument to use.
Do you have a better approach than Kelly? I would be very interested to hear. It’s a nice problem to have! Being faced with a 60% probability bet and not knowing how big to get. Most of the time I am faced with marginal bets and use small sizing as a default. Even with my ‘compelling’ trades I don’t feel like increasing the size by much. That tells me my confidence is still low. I have a good mate who is a professional sports gambler and says many of his friends (and rivals) use some sort of fractional Kelly. He uses half Kelly as a starting point. I am more inclined to start with quarter Kelly. I have no analysis/testing to back this up, but half Kelly would ‘seem’ to reduce the risk of ruin a lot, while quarter Kelly ‘feels’ like its erring on the side of caution. Again, no hard figures here, but at quarter Kelly its somewhat exploiting the great rare opportunity that’s being presented, but still allows me to be comfortable.
No I don't. But as a mom and pop small trader, this is my understanding: 1. In investing, the right side of the Kelly fraction often has a steep drop off in returns, so if your Kelly computation is off, you may be on the right side without knowing it and get wipe out. 2. Unlike gambling, the edge and odds are moving targets and so the Kelly fraction is also a moving target. 3. In gaming, one can achieve terminal value with a large enough number of bets quickly. No so in investing/trading where your number of trades are usually not large enough to achieve statistical terminal value, or provide the correct Kelly fraction computation. 4. So, according to Ed Thorpe, you get 3/4 of the performance of Kelly by betting 1/2 Kelly thus greatly increase your margin of safety. 5. Because of the limited number of trades possible, most traders use 1/4 Kelly or less to further avoid the risk of ruin. A further comment on Buffett and Soros: Their bets tended to be one off, so, as in the Bank of England bet, if the edge/odds are considered overwhelming (approaching 1) then an all in bet is not unreasonable. Don't know if I make any sense.
Thanks IronChef. I can‘t say I disagree with any of that. It still seems to me that bet size is the much smaller part of the problem; the larger part being how accurately you can estimate the probability. If you can’t/don’t get near the right number, then the default answer is to err on the side of caution, quarter Kelly or less. But what if you can get to a pretty reliable number? For example, what does Ren Tech use? Surely with their large number of trades (not to mention their servers, pipes, data and PhDs) they must have a pretty good idea on at least some of their trading systems. I wonder how close they get to Kelly or (insert unspecified ‘better approach’ here) on the funds they allocate to these systems. My guess would be that they get close. I have nothing to back that up, but if I assume on a proven long-standing system, that their win/loss is not too far from even, and their risk/reward also not too far from near even (I know, big assumptions, I have no idea) then to explain the massive results/compounding the firm has achieved, they would probably need to be quite aggressive in bet sizing.
Size came directly from calculating your win probability and win expectancy. If you don't change your methodology, then over time you should be able to get your average probability of win, expectancy with some consistency. But anytime you change/adjust your method, your Kelly will change. I trade short and long options differently and I use different Kelly for them. Best to you.
Serendipity I just posted an interview with James Simons in another thread. https://www.elitetrader.com/et/thre...teria-beats-random-entry.313759/#post-4525450 From the horse's mouth, he states that volatility is a major issue and they use complex formulae to keep it under control. From my trading of today, an example. It was NFP Friday so had to work around how that news hits the market, and the hurricane impact was an unknown. Started on spot USDJPY after the London open. Scalp it on the way down into the news. Profit. News hits, jobs are worse than forecast, USDJPY should drop, but it spikes. No argument from me, trade the price as it is, not what I think it should be. Profit. Now I know this price will fade. I know that on a Friday many people do not want to hold a JPY cross over the weekend in case "Little Rocket Man" wants to let one off over Japan. Decent chance that he might, which would cause a gap down on Sunday night (Monday morning in Sydney). So I load up short on USDJPY for the fade, I know people want out. I go far above my normal position limit because I am certain. I make a good profit. BUT BUT BUT I left a lot of money on the table. Since I had loaded up the position I was being far more cautious on the exit then I normally would have been. Had I traded my normal size I would have been willing to follow the price down much further and would have walked away with more money. Good trade, completely understood the situation, had the price action read, but by loading up with too much at risk, I made less money then I could have. I hate leaving money on the table from my being stupid. EDIT: And for those north of the 49th - Happy Thanksgiving
%% Actually that could work [ insane size]; if one only did it twice, only .LOL But ''bet a million Gates'' won a big bet on a UK horse race + considered that all life was gamble. Come to find out '' bet a million Gates'' did not bet a million anyway. I rechecked it =about twice as many millionaires shop[ WMT] Walmart..... as Sears, USa . NOT saying WMT is twice as good; but many stores can help;not a stock tip