Please note that these are just my thoughts and they may be wrong.
Trades whose return driver is that your money is being lent to some profitable enterprise and put to use.
Trades whose return driver is that you are helping the market to better predict the future (the market pays you for this service).
Returns which are highly correlated to the returns of some underlying productive asset class, such as equity indices. Since productive assets are expected to always go up in the long term, this is a positive return driver. The extent of the correlation is called "beta". Its disadvantage is that if multiple strategies have a high beta with respect to the same underlying asset class, then they will have a high correlation with each other, and hence they do not do much for diversification. Many productive asset classes are highly correlated with each other, hence the notion of a "global beta".
The global beta return driver is often quite strong, because investing in equities has a relatively high expected return, so high that it is relatively hard to beat.
A subcategory of predictive-based. In these trades, predictions are made about large-scale events, such as politics and macroeconomics. They generally involve the use of human intuition in addition to quantitative analysis. These sorts of predictions are relatively poor and hence you shouldn't bet too large a proportion of your capital on any one prediction being right.
In addition, you should try to realize the same predictions different ways via different trades -- this allows you to partially 'hedge out' extraneous/idiosyncratic exposures in each trade (particularly exposures that you are unaware of), leaving more chance that the overall performance is highly correlated to the success of your prediction.
Technical means using some formula based on market-endogenous variables such as price and volume to determine when to trade.
They have a predictive-based return driver, although they may have others too.
Fundamental means using some formula based on market-exogenous variables such as earnings and book value to determine when to trade.
They have a predictive-based return driver, although they may have others too.
A strategy is systematic to the extent that there is an algorithm which completely determines the strategy.
The systematic part of a strategy can be backtested.
A strategy can be either absolutely systematic (e.g. there is an algorithm which completely determines the strategy) or it can involve human intuition. If it is absolutely systematic, it can be entirely backtested, which is a huge advantage.
You can commit a larger portion of your capital to a well-backtested strategy.
However, it is always possible that a strategy was only coincidentally profitable in the past, so you shouldn't bet too much on a backtested strategy if it is not supported by theory.
To the extent that you have a well-supported theory that explains why a strategy is profitable, you can bet more on it. However, the underlying conditions for a strategy to be profitable can change, so unless it is statistical arbitrage or arbitrage, you must still limit your confidence in it.
A systematic by default strategy is one in which there is an algorithm which could determine the trades, but a human has the option to intervene. Such a strategy is not absolutely systematic.
An abortable systematic strategy is one in which there is an algorithm which determines the trades, except that a human has the option to pause the strategy when they feel that the conditions for the strategy to do well may not be present. Such a strategy is not absolutely systematic (although the underlying strategy is).
With such a strategy the main danger is that the strategy may only be profitable in some conditions, and the human may selectively abort it in those conditions. Abortable systematic strategies are safer when the underlying absolutely systematic strategy is stochastic arbitrage, because in this case the strategy should still be profitable no matter when the human chooses to abort.
A subcategory of technical. A momentum-based strategy looks at recent moves in price (for some value of 'recent') and makes the assumption that the price is likely to go up if it has gone up recently, or vice versa.
A subcategory of technical. A contrarian strategy looks at recent moves in price (for some value of 'recent') and makes the assumption that the price is likely to go down if it has gone up recently, or vice versa.
Contrarian is the opposite of momentum, however a composite strategy may be momentum on some time scales or conditions and momentum-based in another.
Trades whose return driver is that you are willing to lock up your money for a long period of time in exchange for a higher expected return in the long-term.
Trades whose return driver is that you are willing to assume more volatility in exchange for a higher expected return in the long-term.
These are similar to illiquidity-premimum based, in that, if you only have a little bit of time to invest before you will withdraw your money, they are not a good choice (because the value of the investment may go down and take a long time to go back up).
A hedged trade is one that starts with a simpler trade and then cancels out some of its return drivers. The goal is either to make the trade's profitability more predictable (by subtracting unpredictable return drivers) or to make the trade less correlated with others in your portfolio (by subtracting return drivers already in other trades in your portfolio).
These are nice because you can get a lot of data on how well the formula works, due to the short term.
Short-term technical trading strategies are a good choice for automation.
A strategy with trades which lose a reasonable amount when they lose money, but which sometimes make a lot. Such a strategy can be profitable even when most trades are losing, but of course also if most trades are winning.
Trades that usually cost you money (e.g. most trades are losing) but then every now and then pay off big (e.g. asymmetric payoffs). If the time scale is long, then these are dangerous because the sample size of the big payoffs is relatively small, so you may be wrong about how frequent/how big they actually are -- but i suspect they may be extra profitable, because most people probably don't like to lose money most of the time.
Synonym for "profitable strategy". A strategy whose expected log return (geometric return) is positive.
Strategies which have a negative (geometric) expected return.
For example, a strategy in which most trades pay off but every now and then lose big, such that the payoffs are not enough to make up for the occasional losses.
Sometimes losing strategies can be inverted to create winning strategies. But often they cannot, due to execution costs and bid-ask spreads.
An absolutely systematic strategy that is 100% guaranteed to make money, provided there is no execution errors or counterparty failure. These are hard for retail traders to find because big institutions with low execution costs and fast computers can generally fill the capacity for these strategies, unless you discover one that is relatively unknown.
These tend to be super-sensitive to execution costs (because otherwise it's likely that other retail traders would have discovered it).
If you find one of these (and if you are SURE you understand it), you can commit a large proportion of your money to it, and perhaps even leverage it.
An absolutely systematic trading strategy that is mathematically guaranteed to make money, under some highly probable assumptions.
You can commit a relatively large part of your money to these, if you are relatively certain the analysis is correct.
A trading strategy that takes huge losses every now and then but that should make money over all if you can hold on long enough.
These trades are sometimes disguised forms of selling insurance.
Backtesting means seeing how an absolutely systematic strategy would have done in the past.
One has to worry about "overfitting", that is, about a strategy with so many free parameters such that the free parameters can be chosen to make the strategy coincidentally work well in the past.
A well-backtested strategy should be consistently profitable over many time periods.
Backtesting can be forgone if there is a solid theory that guarantees or nearly guarantees profitablity (arbitrage and statistical arbitrage) (but if possible, even those should be backtested).
Non-absolutely-systematic strategies cannot be truly backtested, making them much more dangerous.
A strategy which is unknown.
In some cases, it is even unknown whether or not the strategy is absolutely systematic.
These strategies are dangerous because there is no theory to explain the return drivers, and especially if the degree of systematization and number of free parameters is unknown, because in this case there is little basis which which to determine how much confidence to place in the hypothesis that future returns will be similar to past returns.
If historical data is available, then backtesting can be done, and extra emphasis should be placed on consistency of returns over many time periods. However even with well-backtesting, such strategies are dangeous due to the possibility of coincidental profitability.
Note that you are more likely to hear about a strategy which has been profitable in the past, so there is survivorship bias which makes coincidental profitability more likely in strategies that you hear about.
My personal guess is that black box strategies could be useful in a portfolio provided that emphasis is placed on consistent historical returns over many time periods, and on low correlation with other strategies in the portfolio.
"don't catch falling knives" -- that is, don't try to guess when a falling asset will reach its bottom
"buy on the rumor, sell on the news" -- for positive rumors
If your strategies are all absolutely systematic you don't have to worry much about psychology, you can do everything with math.
Otherwise, you have to worry about psychology. Even if you were rational, you will have to learn over time how to convert a subjective level of certainty in your head into a concrete number.
But you aren't rational. Not only are you not rational, but your specific form of irrationality is shared with many others, and has been studied by other traders. Mr. Market, who is trying to predict the future based on inputs from people like you, must attempt to filter our your irrational biases. As a side effect of this filtering, it will appear to you as if the market is being manipulated by an adversary who is able to anticipate your biases and who attempts to move against you and take your money whenever you make an irrational mistake.
Unfortunately, knowing our own biases in certain cases doesn't help us to counter them, and more intelligent people are slightly more prone to certain kinds of irrationality ( http://www.newyorker.com/online/blogs/frontal-cortex/2012/06/daniel-kahneman-bias-studies.html )
more to come here.
If your strategies are all absolutely systematic then you can size positions with math, otherwise they must also reflect your subjective levels of certainty in your various trading theses, in which case your position sizing becomes subject to psychological factors also.
Your position size may be too small if:
Your position size may be too large if:
Types of strategies in order from smallest to largest position sizes:
If you find that your position size is too large, you should strongly consider reducing your position size immediately. The decision tree is:
Note that if you hold a call or put option with a market value of $X, that is usually comparable to a market value of the underlying which is much larger than $X. The reason is that, if you are near ATM, and the market goes against you, the value of the option will go down more, proportionally, than the value of the underlying. If you are OTM, then your problem is that the option loses a lot of money (proportionally) each day when nothing happens. If you are deep ITM then the option is mostly functioning like the underlying anyway (so why'd you pay the large option spread?).
If you are using stop-loss, you should compute your stop-loss price before you size your position, not after. First choose your stop-loss price, then compute how much money you want to risk, then compute your position size from these. If you did it the other way around (computing the stop-loss price from the combination of the money you want to risk and the position size), then you might come out with a stop-loss that is too close to the current price, causing you to be stopped-out on a short-term fluctuation even if your trading thesis is ultimately correct.
if you find yourself frequently shuffling assets between accounts to provide cash or margin, then you don't have enough cash and/or your position sizes are too large!
Your goal is the maximize the expected log (logarithmic) return of the expected proportional change in your portfolio (proportional change = return + 1.0, e.g. a 10% return = a 1.1 proportional change).
Here's a paradox: if, once every day, someone offers you a 50% chance of losing $1, and a 50% chance of winning $3, that's a winning strategy; take the bet every day. But, if they offer you a 50% chance of losing all your money, and a 50% chance of winning 3 x all your money, never take that bet; you'll most likely be broke within a week. Yet these are the same odds on offer. The difference is only in their relation to your net worth. The log expected value captures this; when only $1 is at stake, the expected return (as a proportion of your net worth) is tiny, so the proportional change is close to 1, and the log function is almost linear there, so the expected log return is similar to the expected return. But when your entire net worth is at stake, the proportional change is close to 0 if you lose and relatively far from 1.0 if you win, and the log function is not very linear there at all, so the winning side gets a big haircut by the log function.
In all cases, to break even, the proportional change of losing must be the reciprocal of the proportional change in winning; so, e.g., if you lose 2/3s of your money half the time (proportional change 1/3), then you'd better make at least 3x your money the other times.
A lot of people like to distinguish between "trading" and "investing". While i'm not too happy with that choice of words, i like to distinguish between "trying to get paid for predicting future prices better than Mr. Market", and "trying for putting your capital to use in the economy".
Two subcategories of the "trying to get paid for predicting future prices better than Mr. Market" side are:
My current hypothesis is that there are two times when speculation may be a good idea:
When it is NOT a good idea is when you do not have rare information, and the market prices seem wrong, but not so wrong that they are crazy. The problem is that Mr. Market is much smarter than you so you shouldn't disagree with hir without conviction in your hypothesis.
Three gotchas to remember about hedging:
And note that hedge rebalancings are often themselves bets:
a few things happens randomly. someone makes a strategy out of them. they randomly happen a few more times. more people notice and start following the same strategy. this increases the returns of the strategy for the first people (consider a simple strategy like 'buy gold'; the intial random events are gold going up)). more and more people notice and pile in and increase the capital they allocate to this strategy. at some point, people have plowed in as much capital as they are comfortable with for one strategy and/or the well of new recruits begins to run out. this tends to happen at the same time for many people, because exponentially more people have joined as time goes on. then the value of the strategy (asset) starts going down. people reduce their allocation to this strategy; this causes momentum; the same process happens in reverse; and the value plummets, until its price hits a floor of a combination of 'believers' who believe in a theoretical justification for the strategy in addition to the evidence provided by its performance; and people who are too lazy to notice that their strategy is underperforming; and people who are buying the strategy/asset as part of following other strategies.
from the perspective of an individual, what happens is they discover (or worse, are told about) a strategy, and they put some money in, and they make money, and they put more money in, and make more money, and then they put more and more money in, and then right at the moment that they max out with the money they are putting in, the market runs away with that money. the reason it happens right when the market has gotten you to commit as much money as it can is that there are 10,000 other people like you doing the same thing at the same time for the same reasons.