Are random trading strategies more successful than technical ones? | Hacker News Reader

Are random trading strategies more successful than technical ones?

journals.plos.org

55 pointshkai70 comments

Prices are pretty well modeled using Brownian motion. Most economists should know this while almost no one in the normal population will be aware of it. Sometimes people are just lucky, but overall the more trades you make the more you'll converge on the average return rate.

I would also like to note, that predicting price is different from predicting an overall increase in the value of the underlying security.

https://en.wikipedia.org/wiki/Brownian_model_of_financial_ma...

On average, but prices come from the value of the company behind them long term which is not always brownian and so someone who knows the company can get in/out ahead of someone who trusts only brownian motion.

Further, every time you trade, the overwhelming likelihood is that the counterparty to that trade is a financial professional with dramatically more access to company-specific research and information than you. This imbalance is minimized when you trade infrequently and maximized when you trade frequently.

For retail traders, isn't it overwhelmingly likely that you're just going to trade with the inventory of a market-neutral internalizer?

It's not either-or. You're right you're mostly trading with a market internalizer, which is pocketing the spread (ie. bid: 10.00, ask: 10.05) but isn't really making money on price movements. However, you're still against hedge funds/banks for medium/long term price movements (eg. you buying a stock after they pumped it, and selling after they dumped it).

Yeh. It would be more accurate to say something like this:

A retail trader will usually be trading with some sort of market maker. That market maker will also trade with informed traders, and sets their bid-offer spread accordingly to offset this adverse selection. So the retail trader is paying a bid-offer spread that is the result of other informed traders in the market.

Of course, as a retail investor you may have access to a broker (e.g. Robinhood) that tries to exclude informed traders so it can set a lower spread.

There are market-maker intermediaries in between but structurally you're playing the game primarily against professional finance teams with massive amounts of research, automation, and up-to-the-microsecond information you don't have access to who are the ones actually setting the price.

I don’t think that’s quite right. Otherwise, you could just follow the opposite of your unprofessional trade strategy as a cheap proxy for a professional trading strategy.

I think the market is dominated by front-running trades and randomness.

This reasoning doesn't work because the market isn't a sequence of discrete binary choices. If the "opposite" of a bad strategy was a good one anyone could have great returns by designing some obviously terrible money losing strategy then doing the "opposite".

Right, and even if the market was a sequence of discrete binary choices, bad traders will make many bad binary choices, but many of their choices will be good ones. (They just won't know which are bad and which are good.)

In other words, reversing every choice in a mediocre sequence of binary choices yields another mediocre sequence of choices.

Taking into account trading costs, a meh strategy and its inverse can both lose money.

Front-runners still have to run in front of something, though. And it's probably not going to be the purchases of some little retail investor.

Meanwhile in the Theranos thread next door… “What just amazed me is how gullible all the investors were, and how they didn't do due diligence, hire outside experts, or anything. Weird.” Financial professionals do dumb things, sometimes en masse, and I think there are still some opportunities to make money if you have a good nose for BS/mass delusions.

However, the problem with this strategy is, as Keynes put it: “The market can stay irrational longer than you can stay solvent.”

It's also very easy to convince yourself that you have particular insight that those other people lack. You might even occasionally be right. But generally speaking the odds are that they understand something you don't, not the other way around.

Specifically to your Theranos point, Theranos was never public. Conning individual investors out of private investment money is somewhat different since the people who saw through the bullshit don't have an easy way of profiting off of that sense. The options were to buy in or opt out, and plenty of investors wisely decided to opt out but there was no play for them to profit off of the collapse.

The assumptions underlying Brownian motion of prices have been disputed for quite a while now: the normality hypothesis can be rejected on most if not all historical financial returns series, as it turns out that most returns are actually fat tailed processes with very significant (and variable over time) correlations between distinct assets, which makes research around portfolio theory even harder to conduct.

Which assets are that?

Most if not all of them. Especially during a downturn, all assets of all classes tend to correlate and produce negative returns. During big crises, stocks that before seemed uncorrelated/anti-correlated have a tendency to increase their correlation and go down together; at the extreme, even bonds cease to act as a diversifier against stocks plummeting.

Depends on your timeframe right? Shorter time periods are a bit more random. But there's also clearly some autocorrelation which makes sense given the inflationary / deflationary expectations at play

Yes, definitely. The big issue there is that intraday intervals vs daily closes vs monthly prices all require very different kinds of analyses and features as they target different scales of behavior. For example, you could exploit order book models for intraday which make little sense for longer time frames. In the same way, portfolio theory on intraday intervals tends to not hold as well as it does on longer timeframes.

This is absolutely correct

If the markets were efficient and equivalent to random brownian motion, Jim Simons wouldn't be producing 50%+ returns for decades non-stop in HFT.

His net worth is 25 billion dollars, and there are other countless billionaires, made from the "efficient" markets.

That's how much this fallacy is worth.

As someone who has beat the market over a 17 year period, I attribute it to having made very few trades AND being lucky.

There are a lot of ways that a company can go under. All it takes is one mistake or one black swan event and what would have been an amazing investment 99.9999% of the time goes to zero.

> Brownian motion is the random, uncontrolled movement of particles in a fluid as they constantly collide with other molecules

sounds like the economy to me :)

Stonks always go up. /s

This reminds me a bit of a classic paper called "1/N". It compared a portfolio of putting equal money into each security, vs a bunch of fancier approaches. The 1/N almost always won.

https://papers.ssrn.com/sol3/papers.cfm?abstract_id=911512

This is widely known among practitioners, but there is a caveat -- a 1/N portfolio bears a much higher risk than, say, a cap-weighted portfolio or a risk-parity asset allocation. A 1/N portfolio receives an equal contribution in terms of volatility from each asset, meaning that very risky assets significantly increase the portfolio's volatility, while not necessarily contributing proportionally better returns, due to the nonlinearity and asymmetry of volatility's effects on prices. This way, 1/N ends up performing very poorly on a risk-adjusted basis while undoubtedly at the same time outperforming any other kind of allocation on the basis of return alone. This is rather unacceptable in a real world portfolio where the tail risks and emotions can lead an investor to ruin.

Addendum: the paper actually mentions the Sharpe ratio, which is a general, popular measure of risk adjusted returns, but which fails to take into account the non-normality of the distribution of returns; so, while my previous comment may be incorrect in a Gaussian world, I would be curious to see the results when the performances are evaluated under the assumption of fat tailed processes, which I presume would paint a very different picture.

> This way, 1/N ends up performing very poorly on a risk-adjusted basis while undoubtedly at the same time outperforming any other kind of allocation on the basis of return alone.

I fear I'm misunderstanding you. Are you saying despite having higher returns, the higher risk makes this strategy worse? That really feels like handwaving to me, since the only thing I care about is ROI. I understand nonlinearity and how it could tank your investment, but if it doesn't and you make more money then you're criticizing something that never happened. The higher risk is already baked into the ROI, because it includes the times that failed. The point is, in aggregate, you make more money - and most of the time that is the only thing I care about when investing.

Or am I misunderstanding you?

I believe OP is saying that due to the increase in risk and volatility there is a greater chance of the trader to emotionally manage their money as it peaks and falls.

And obviously there’s mathematical measurements where one tries to get the highest return per unit of risk. It’s possible that these returns might be higher but fall under the curve.

You are not misunderstanding. While there is no doubt that there is a premium on taking more risk, ROI is also a very debatable metric to consider in a vacuum, though. If you only care about ROI then you either can afford to risk everything because you have cash/safer investments in place (so you are really taking less risk), or you are YOLOing.

> it could tank your investment, but if it doesn't and you make more money then you're criticizing something that never happened

This way of reasoning is basically survivorship bias in a nutshell, and per my direct experience as a financial professional has brought down many investors who were too confident about their "strategy".

To bring this argument to the extreme: if you cared only about ROI, you could just go long some penny stock with exaggerate leverage and make big money "unless proven otherwise". In practice, what happens is you get euphoric for a couple days while you see the money shoot up to the sky, and then lose all of it to a margin call at the opening the very next day. I've seen it happen with my own eyes.

>Are you saying despite having higher returns, the higher risk makes this strategy worse? That really feels like handwaving to me, since the only thing I care about is ROI.

There’s lots of metrics that try to balance the risk and reward. Often, the risk is based on the volatility of the asset. The common alpha metric does this by incorporating the assets volatility compared to the overall market volatility. There’s others like Sharpe ratio etc.

Factoring that volatility is particularly important in long-term investing so your choices don’t, as you say, tank your investment. So maybe you interested in cyclicals over the last nine months and your investments went gangbusters. Does that mean that same strategy will work in perpetuity? Probably not, because cyclicals tend to have high volatility. Risk -adjusted metrics attempt to quantify that risk.

I think another way to characterize 'riskiness' is the volatility of a portfolio, IOW how much does it swing up and down over time. So everything else being equal, two funds with the same ROI may still have different levels of volatility, with the fund having the lower volatility being more desirable.

The reason that risk is important to quantify is because leverage could be used (in theory), to achieve risk parity between different strategies. So ideally you would pick the one that has the best risk adjusted returns (with enough diversification) and then leverage it to the amount of risk you would be comfortable with.

As a highly simplified, unrealistic example - let's say strategy A has an average return of 5%/year with max drawdown of 20%, and strategy B has an average return of 10%/year with a max drawdown of 50% (here max drawdown being a highly simplified proxy for risk). Theoretically, you could use leverage to go 2.5X long strategy A to achieve a return of 12.5%/year with a max drawdown of 50% (minus cost of the leverage - depending on how you do this, cost could be fairly small). This might do better, risk-adjusted, than just doing strategy B by itself.

Interestingly, low volatility investing has some studies that show it can provide better risk adjusted returns than naive 1/N allocation or traditional Markowitz optimization

Yes! There is a number of widely known inefficiencies (low vol, mid cap stocks, the trend following anomaly, long volatility oriented strategies...) that consistently produce better risk-adjusted returns. Some of them aren't really popular enough to stand out and reach their full capacity, others (trend following especially) are so counterintuitive that in practice almost no manager nor investor ends up being able to sustain the emotional pressure, eventually everybody cries uncle and there is no way the anomaly gets crowded enough to go away.

> It compared a portfolio of putting equal money into each security, vs a bunch of fancier approaches

They compared equal weighting, but did they also check market-cap weighting?

Sadly, the abstract doesn’t include the result, so here it is so you can decide if you want to read more:

> Our main result, which is independent of the market considered, is that standard trading strategies and their algorithms, based on the past history of the time series, although have occasionally the chance to be successful inside small temporal windows, on a large temporal scale perform on average not better than the purely random strategy, which, on the other hand, is also much less volatile.

What a ridiculously formulated sentence.

TBF the four authors have names that appear to be Italian and German. They acknowledge the contribution of someone who provided DAX data (German stock index, like the CAC or Dow). And lots of subordinate clauses are common in written German, or at least a lot more common than in English.

So while I agree the sentence is rather contorted, there is a sympathetic explanation. Especially as the authors claim no institutional affiliations.

I don’t think such a sentence would be justified coming from an institution in an English-speaking country.

Win % is a really useless metric in this business, try computing win % for something like long vol strategies (for example things like what Taleb did back in the day), it might come out to 5% or lower and still make money. And because every trade has a counterparty there's plenty of strategies that win 95% or more of the time but eventually lead to ruin. Returns pretty much never have a symmetric distribution.

Computing win % is akin to measuring software quality in terms of number of lines of code - only someone who has no first-hand experience would ever attempt to do that.

Yeah - Martingale strategy is a good example of high win % strategy that doesn't work (unless you have infinite money)

If any well known strategy was profitable presumably it will be used by people until it isn't, because knowledge of the strategy is already priced in to the relevant assets. That makes this result fundamentally unsurprising.

Doesn't the conclusion indirectly also indicate that day trading is a zero sum game?

If the answer is yes, then the only way you can make money from day trading is from commissions you earn performing day trade on behalf of other parties with money.

Poker is a "zero sum game", yet experts routinely win money from the suckers. Unlike Craps, Poker as a huge skill component.

The conclusion does not imply anything about day trading being a zero sum game.

This is a misunderstanding of zero-sums games.

Zero-sums game are actually proven to have a winning strategy.

Chess is a zero sum game.

Chess is not believed to be a forced win for either player though.

OPs claim is poorly stated. He's referring to Zermelo's theorem which states that a finite game with two players that's deterministic and zero sum with perfect information and no possibility of a draw must have a winning strategy. It's not difficult to prove that this must be true and you likely can intuit why it's true (imagine building a decision tree for such a game).

But all of those qualifiers I mentioned are needed, and that's a lot of qualifiers. If any of them are no longer true then there is not guaranteed to be a winning strategy.

In chess, it's possible to end the game in a draw, so Zermelo's theorem does not apply to it and OPs claim is wrong about chess.

I'm fairly certain one can trivially disqualify one of those criteria when it comes to financial markets as well.

> In chess, it's possible to end the game in a draw, so Zermelo's theorem does not apply to it and OPs claim is wrong about chess.

Isn't it even assumed that a perfect game of chess is a draw. Once chess is solved, it'll be all draws.

Yes it's an open question. The prevailing opinion is that a perfect game ends in a draw as you said, with a minority opinion that white can force a win. I am not aware of any credible opinion that black can force a win.

No.

First they ran in simulation, not the real market. It may be that that act of being in the market changes the market enough to make your strategy work. (though typically it is the opposite - things work in simulation but applying them to the market makes them not work). As such this paper doesn't really tell us anything useful.

Even if we ignore the above, they only tested a few different strategies. That says nothing about any other trading strategy that someone might apply: any of them might work.

I still think day trading is a bad way to invest, but this paper doesn't prove anything even though it speaks to my bias.

>things work in simulation but applying them to the market makes them not work

Can you elaborate? Is this because large flows of money eventually become the market? Insinuating that some strategies only work at low trade volume?

Pretty much. When you trade any amount of money your trade is changing the market.

In some less honest markets there are even cases where what the numbers say you can trade isn't possible because those offering the deal won't follow through, or will let a friend in ahead of you

If trading is a zero-sum game, which it is on a small scale, then random strategies are bound to be in the middle of the pack.

It is like rock-paper-scissors. A random player will win 50% of their games regardless of the other player strategy. When two non-random players play, one will successfully predict the other player moves and win more than 50% of the time, the other will fail and win less than 50% of the time.

So the ranking will always be 1. winning strategies 2. random 3. losing strategies, with as many winners as there are losers, and any number of randoms. So, random is more successful than half of the technical strategies.

Can you elaborate on what you mean by trading is a zero-sum game on a small scale? Without clarification that statement could be used to justify any conclusion.

Are you saying that someone who trades a small amount of capital is always winning an amount of money that is roughly equal (+/-) what the counterparty lost or vice-versa? That can be demonstrated to be untrue.

Are you saying that trades spanning a short period of time always win or lose an amount that sums up to zero for all participants? That also seems highly unlikely unless all participants are engaged in short term trading which is not true in practice.

At any rate, while I've heard this claim repeated often, I've never heard anyone substantiate it and as far as I can tell it doesn't really make sense.

There are financial instruments that are zero-sum by their nature with respect to dollars, for example derivatives and currencies are by nature zero sum with respect to dollars, although they are not zero sum if you factor in risk. But that has nothing to do with short vs. long term though. Equities are not zero-sum, long term or short term.

Technical analysis sort of "works" in the same way that e.g. astrology "works", in that for any given plot of stock data, you can typically draw a of a number of technical patterns which seem to fit. I've never seen any convincing evidence to the contrary.

But one thing is for sure, if technical analysis works then a neural net will trivially pick up on existing strategies and although the cutting edge is always kept secret in the financial world, we probably would have heard of ML techniques rediscovering technical analysis by now if it were truly successful, since even an amateur could build and train a neural net from free data to learn technical analysis.

P.S. if simple technical analysis techniques ever worked, I also predict that they would quickly stop working as such arbitrages eventually disappear. You're not trading against news or patterns, ultimately, whether traders realize it or not, they are trading against mass financial psychology and HFT algos. Once neural net based training becomes the predominant tool, it will be interesting to see the collective patterns that emerge, likely totally disconnected from actual fundamentals. It may be chaotic, or it may be close to steady state, but it will definitely be in a state of flux as neural nets come online and constantly train on the latest patterns. It's a battle against the arrow of time.

The paper studies trades made on financial market indexes, so over the periods of time measured I wonder if the random strategy they used is about the same as investing in index tracker funds and spreading your buys / sells out in order not to time the market.

https://markets.businessinsider.com/news/currencies/hamster-...

I think so

The technical strategies they compared it too are not strategies commonly used. It looks like they were chosen because they were simplistic and convenient to back test.

I love stuff like this. Pure comedy gold. It reminds me that someone can have all the knowledge, all the statistical tools in the world and still make huge mistakes (no, not explaining it, making too much money atm...maybe in a few decades). To the man with a hammer.

>Recently Taleb has brilliantly discussed in his successful books [15], [16] how chance and black swans rule our life, but also economy and financial market behavior beyond our personal and rational expectations or control. Actually, randomness enters in our everyday life although we hardly recognize it. Therefore, even without being skeptic as much as Taleb, one could easily claim that we often misunderstand phenomena around us and are fooled by apparent connections which are only due to fortuity. Economic systems are unavoidably affected by expectations, both present and past, since agents’ beliefs strongly influence their future dynamics. If today a very good expectation emerged about the performance of any security, everyone would try to buy it and this occurrence would imply an increase in its price. Then, tomorrow, this security would be priced higher than today, and this fact would just be the consequence of the market expectation itself. This deep dependence on expectations made financial economists try to build mechanisms to predict future assets prices. The aim of this study is precisely to check whether these mechanisms, which will be described in detail in the next sections, are more effective in predicting the market dynamics compared to a completely random strategy.

I think pundits, academics, experts etc. overestimate the randomness or unpredictability of markets and crowds. Consider this obvious thought experiment: given a choice between having to choose between a $10 bill or a $20 bill on the sidewalk, all else being equal, everyone will choose the $20.That is sorta how investing is. Quality beats crud. There is nothing mystical or unpredictable about it. Determining quality is subjective, but the FAANG index in which each company is worth at least $100 billion has pretty much beaten everything else since 2009.

Also a distinction should be made between fundamental analysis, quantitative analysis, and technical analysis (volume and chart patterns and readings). I think the the first is useful, as the out-performance of FAANG stocks shows. Quant strategies can also be very profitable. The alleged predictive power of technical analysis has long been debunked.

I don't have the impression that Taleb's thesis is anything like choosing between two known valued bills on the ground. Maybe it would be more like:

"if you were going to hunt for $20 bills on the sidewalk, which park would you go to? Central Park always does pretty well but if you were to play 'double or nothing' for tomorrow's find, you couldn't guarantee that you'd find a $20 bill there just because you found one there yesterday."

>but the FAANG index in which each company is worth at least $100 billion has pretty much beaten everything else since 2009.

The companies that have seen the largest growth, amid the longest bull market in history, have beaten everything else?

Isn't that pretty much a tautology?

In this context it's usually called "survivorship bias".

>the FAANG index in which each company is worth at least $100 billion has pretty much beaten everything else since 2009.

In some sense I think this speaks more to the way that the US regulatory framework allows dominant players in a given market segment to retain and reinforce their dominance.

You can argue that these type of investments are "quality" or "safe", but the reasoning behind that label isn't going to be based on any kind financial analysis. There's no path to dethroning these giants or constraining them in any significant way, and as a result they're insulated from market fluctuations that might crash the price of a smaller player.

That's all without going into the feedback loop of safe investments -> more investors -> higher price (or price stability) -> upgraded safety rating -> algorithmic rebalancing of index funds -> higher price -> etc.

They're also information technology companies. Maybe it shows just how of a part of daily life they are? The only ones whose I don't interact directly with (intentionally) on a daily basis are Apple and Facebook. For most people Facebook would be included in their daily use.

How many other companies are there that aren't IT related that you interact with on a daily basis? You might use your chair and toothbrush every day, but that doesn't require anything from the company that sold you the chair. Using Google does require Google's servers to respond though.

Tesco. I can't eat likes. Oil companies also, though I hate to admit it.

> Consider this obvious thought experiment: given a choice between having to choose between a $10 bill or a $20 bill on the sidewalk, all else being equal, everyone will choose the $20.

Where's the experiment part?

> Consider this obvious thought experiment: given a choice between having to choose between a $10 bill or a $20 bill on the sidewalk, all else being equal, everyone will choose the $20.

All else is never equal. If you change this experiment slightly, you'll get a more interesting result. If there is a $10 bill and a $100 bill on the sidewalk, and you have to choose one to take and one that will return to its owner, most people will choose the $10. The $100 seems suspicious and dangerous (in a "mystical and unpredictable" way.)

Quality is determined by experience and instinct. The personal valuation of a $100 bill might drop below $10 with no added information, other than that all treasure looks less like treasure than a lot of trash does.

Suffice it to say that if there were a market that accurately labeled the values of everything it sold, it wouldn't be a very interesting market.

> the FAANG index in which each company is worth at least $100 billion has pretty much beaten everything else since 2009.

That's because the politics are affected by size. To big to fail is real.

When investing in something like the FAANG index or Google you're not betting on how the companies will do. Predictions like "Google is going to do well in the future and continue to grow" are not useful for making investments.

You're making a bet that Google will do better than everyone else thinks it will. And even more than that. You're betting that it will do so by a wider margin and/or with a higher likelihood than the available alternative investments you could make with that same money.

And even more than that, you're betting that Google will do better than everyone thinks it will and that the market will acknowledge this the way that you expect and the price will adjust accordingly in a time frame that is relevant for your investment goals and solvency.

You are betting more money will come in from new investors or more money from existing shareholders. You are betting people won't sell.

You are not betting on what will think in the future because market forces are bigger than will.

Reply on news.ycombinator.com