No.
I know more than one person who've gone all in on worse bets and won. Made their personal fortunes.
I've also known a few who went all in on similar bets and lost.
Maximum sensible bet size for great setups
- Thread starter Ghost of Cutten
- Start date
I know more than one person who've gone all in on worse bets and won. Made their personal fortunes.
I've also known a few who went all in on similar bets and lost.
I've considered that approach, and the problem I see is that if you have a nasty drawdown (which betting at Kelly sizes practically guarantees), then you have to scale back radically, and this inhibits account growth in future.
For example, let's say you allocate 20% of equity to your Kelly sizing account. You see a slam dunk trade and Kelly says bet 70% of your capital. You lose. Now you have 7% equity left to risk. Another slam dunk comes along later, and you lose again. Now you have 2.1% left. You will thus be sizing at 1/10th of your original amount, because your Kelly account is only 2.1%, despite having 82.1% of your original net worth left - this seems suboptimal, it could take you many normal trades to dig out of that hole, and your other 80% of capital is sitting their being under-employed.
Am I missing something here? No matter how much you 'pretend' that you only have 2% of equity left, the fact is that you have 82% left. Your drawdown is a sunk cost and it seems like you're just inhibiting account growth by trading super-small there. I suspect that betting a fixed fraction of the total account value, and making it small enough to keep the risk of major drawdowns low, would prove superior. Would be interesting if someone good at maths/stats could run the numbers on this.
Yes, I appreciate that there's always a small chance of something worse happening. By 20% drawdown tolerance, I'd say that a 95%+ confidence of that not happening would be sufficient. After all, once you get to very long odds like <5%, chances of things like dying, becoming disabled, being sued into bankruptcy, being sent to jail, your country turning socialist, or losing a major war are starting to become just as likely. It makes little sense to reduce your trading risk to a number significantly lower than your risk simply from being alive.
12 losses in a row, with 75% win rate, is a 1 in 16 million chance, a bit conservative there Lucias
More realistic is system or market feel degradation leading to a lower win rate than anticipated - this is a possibility that must be considered, and serves as a realistic cap on bet size. For example, if you think you have a 75% 5:1 trade setup, and in reality it's 40% and 2:1, you are going to be in serious trouble if you size based on the former. This is the logic behind scaling back size during equity drawdowns. However, note that even with such a catastrophic misjudgement, Kelly still recommends betting 10% of equity on the latter odds.
My instinct is that something like 5% (still only half Kelly for the inferior trade setup) should be the maximum on any given bet, for exactly this reason. To avoid the danger of setup degradation, you should probably reduce it quite quickly as you have losers. E.g. starting with capital of 100, bet 5 at first, then 4, then 3, then 2, then 1. So at a 15% drawdown, you are only betting 1.25% of capital even on your slam drunk trades. Only 10 losers in a row will put you over the 20% drawdown threshold, and even at 1% bet size from then on, you only need 4 winners more than losers to claw back to your old account equity high. If they don't come, then you know your system or approach is probably kaput. Losing about 20% of your account in that scenario is acceptable risk IMO.
With 5% bet size, scaling back to 4%, 3% etc as you have a losing streak, you will still be making 10-25% on your winning slam dunk trades, as long as you don't have 4 or more losers in a row. At 75% win rate, the odds of 4+ losers is 0.39%. So you still have good trading firepower in all but the most unlucky runs.
If you want to stick to fixed fraction, rather than scaling back size during losing runs, then obviously you have to trade smaller. I'd say about 2.5% is probably the most you should bet on a fixed fraction with those odds. But I think this would reduce profitability compared to the dynamic position sizing approach, your winning streaks will be only half as good (even less so once you factor in compounding).
Yes. A good example is the outbreak of WWI, the markets were closed for months and then gapped down 25%+. Imagine if you were leveraged 3:1 in that scenario, and thought you had hedged yourself by buying some 1 month out-the-money put options. Ouch.
Kelly suggests betting 70%+ of account equity on high probability setups with good payoffs. A 25% chance of having a 70% drawdown is something only a complete fool or reckless gambler would contemplate. The vast majority of market operators are going to be out of the game if they lose that much.
75%+ win rates and 5:1 payoffs happen from time to time in the markets, don't try to denigrate others just because you haven't yet worked hard enough to find them.
Sorry but you're wrong on the latter part. If you were right, I would not have asked this question in the first place.
The problem with betting 20% is that a single loser hits the maximum drawdown tolerance, and 2 losers (6.25% chance) make it a >35% drawdown. For something like a hedge fund, this means it is quite likely to go out of business. For an individual trader, it's not pleasant, and can be psychologically destabilising.
It comes down to how much are you willing to tolerate in a drawdown. 20% is too low for me, if I were running OPM I would make sure I never lost 20% but for my own money, heck, I have had positions where I risked that much(And I regret I didn't risk more). Had I lost the bet I would not have kept those risk limits of course, I would have cut down materially
Yes, its possible the trader will be debilitated mentally after a large loss but then he belongs to a psychologist chair not a trading desk
To me, not betting big when a great trade comes along is like not doubling down on blackjack when you got a 11 against the dealer's 6. You need to close your eyes and do it. Yes, trading is different because the probabilities are unknown but in some situations you can be fairly confident the odds favor you massively(Buying Feb 2010 Fed futures in June 2009 comes to mind)
Yes, its possible the trader will be debilitated mentally after a large loss but then he belongs to a psychologist chair not a trading desk
To me, not betting big when a great trade comes along is like not doubling down on blackjack when you got a 11 against the dealer's 6. You need to close your eyes and do it. Yes, trading is different because the probabilities are unknown but in some situations you can be fairly confident the odds favor you massively(Buying Feb 2010 Fed futures in June 2009 comes to mind)
It is a property of Kelly betting that it's only virtually guaranteed to be optimal at "terminal wealth", but may be suboptimal at any specific step along the way. Over 500 trades, Kelly is the way to go. Over 5 trades, who knows? There is a good paper on Thorp's website which shows this in simulation of 700 Kelly bets:
http://www.edwardothorp.com/sitebuildercontent/sitebuilderfiles/KellySimulationsNew.pdf
Note the important point he makes on page 3, right under the table of results of the various Kelly and fractional Kelly bets: "Most of the gain is in the last 100 of the 700 decision points". In other words, you will only get the benefit of Kelly as you get further and further along the path of using it versus other position-sizing methods. At the 20th trade, you could be in a bad drawdown using Kelly and a not-so-bad drawdown using a fixed fraction or fractional Kelly, with fixed fraction or fractional Kelly being the better performer. By the 200th trade, Kelly is more likely to be the better performer. By the 2,000th, even more likely. I recall having read that there is a formula for figuring out the probability at bet "N" of Kelly or non-Kelly being the better performer, and as "N" gets higher, the probability of Kelly dominating non-Kelly gets higher, too.
To go back to your example, where you only have 2.1% of your Kelly portfolio remaining, the optimal thing would be to continue Kelly betting with that 2.1%. Assuming your winning percentage and payoff rate remain static (yes, this is the most dangerous assumption of all and I'm not trying to gloss over that fact and they aren't exactly static because you would incorporate the new trade information into the Kelly calculation for your next trade), and bets are divisible down to the tiniest percentage of your portfolio size, i.e. you can buy at least 1 unit of your trading vehicle even with less than 2.1% of your net portfolio size, the optimal thing to do is place another Kelly bet, using whatever your set-up's Kelly sizing is now.
Of course, Thorp's example is one with static odds and, as everyone knows, market odds are not static. Yet, we also don't know how variable they are. Does a 75%/5-to-1 payout turn into a 50%/1-to-1 payout zero expectancy set-up overnight? How would you know? If all edges eventually degrade to zero or negative (after commission and slippage) expectancy, one would assume the 75%/5-to-1 edge will eventually get there, too. So, you have two choices: Bet the heck out of it until it stops working at whatever level you are comfortable making that judgment or keep betting fractional Kelly, maybe for hundreds of trades, experiencing that same winning percent/payout ratio for all of them, and leave a bunch of money on the table.
Yes, this is why I think that the only constructive way to conceptualize risk is "risk of ruin". Thinking of risk in terms of "drawdown" will stop you from risking the optimal amount based on your strategy's historical performance. This is also why I think you should only trade with money you are willing to lose 100% of. If that's $10K, trade with $10K. If it's $100K, trade with $100K.
At some point, I would bet that everyone who made a fortune in the markets was a Kelly bettor at some point early in their early accumulation phase. Evidence suggests that those people then tend to level off their betting size due to various constraints, including liquidity, age, institutional, etc. Unless you face those, I don't see the issue with betting as close to the maximum you can justify historically on each bet. Yes, there are things that happen, like WW I breaking out or some other unknown unknown, but who's to say that you wouldn't be positioned short at just such a juncture?
i traded and earned massive profits risking 16.67% on setups like this for 3 years in a row but recently lowered my tolerance to 5%. i just wanted to make sure i as more sustainable through my badluck. it has worked very well for me so far.