Hedge fund rats

Why should trading be the province of humans only?:

One project is Michael Marcovici’s Rat Trader. The book describes the training of laboratory rats to trade in foreign exchange and commodity futures markets. Marcovici says the rats “outperformed some of the world’s leading human fund managers.” The rats were trained to press a red or green button to give buy or sell signals, after listening to ticker tape movements represented as sounds. If they called the market right they were fed, if they called it wrong they got a small electric shock. Male and female rats performed equally well. The second generation of rattraders, cross-bred from the best performers in the first generation, appeared to have even better performance, although this is a preliminary result, according to the text. Marcovici’s plan, he writes, is to breed enough of them to set up a hedge fund.

I don’t myself like the electric shock idea, but there you go.  That is from Diane Coyle, and for the pointer I thank Michael Gibson.


The problem with this is that human traders have always been conditioned to "call the market right" by virtue of reward (making money) and punishment (losing money) utilizing the informational content of some enormously larger stream of whatever market signals they rely on, and no such consistent trading success has emerged in the human case. Ticker tape movements converted to sounds processed by the rat brain couldn't contain enough information to allow the rats to outperform human traders except by chance. I wouldn't invest in Marcovici's rat-traders anytime soon.

If you click through, this is an art project.

Dammit Tyrone!

'and no such consistent trading success has emerged in the human case'

I'm sure that Prof. Cowen is current enough in American pop culture references that this could be at least one possible answer -

'I just want to say one word to you. Just one word. Eugenics'

I admire your determination. You want to talk about eugenics, and nothing is going to stop you!

A compulsive personality with a eugenics obsession? This won't end well.

Based in Germany!

Hear it from Behavioral economist Colin Camerer,
Human has systematic bias whereas animal like chimps (or may be rats) actually performed near the game theoretic Nash equilibrium.

"The problem with this is that human traders have always been conditioned to “call the market right” by virtue of reward (making money) and punishment (losing money) ..."

You don't understand hedge funds.

To do the rat experiment correctly to model hedge funds, the experimenter would be required to feed the rats every time they push the button, give then the food bought with 40% of the winnings when correct, and feed them normally when wrong and shock the experimenter because the investor is the loser, never the manager and trader.

Hedge funds are: heads you lose, I win; tails I win, you win just a bit more than I.

Aren't you disregarding the number of irrelevant stimuli and variables that influence human behavior? Rats probably aren't prone to selfishness, fraud, the same types of distraction, etc.; it's not outside the realm of possibility that when it comes to performing one simple task based on a reward system, rats would perform better than humans.

Makes sense ; its a rat race out there.

I think the first commenter is say that the data the rats receive a subset of the data humans have been getting, so how could a subset of data better predict than the full set which it is a pat of?

Easy - the rats don't suffer from cognitive biases.

Plus, not sure it needs to "better predict" - most blackjack card counting systems use a subset of information to make the task manageable. More info in theory is better, but not necessarily in practice (because of cognitive biases, or inability to sort out irrelevant data, or whatever). Wasn't there a post (saw it on this site or maybe another recently) that chimpanzees were better than humans at playing a game that required predicting what the other player would do (so trying to identify patterns in the other player's moves). More info and in theory being smarter may not actually lead to better result. In a way, what the rat is doing is really technical analysis - which I used to think was complete BS, but now I think there's at least something to it, either because it models human behavior (cognitive biases, or avoiding losses, or whatever) or it models humans following technical analysis.

The "full set" of data contains a lot of noise. Lower the noise floor and the signal will be more pronounced.

Electric shocks just being a little bump along the road in providing more unbiased confirmation of the benefits of eugenic breeding in a deterministic environment.

First-rate rats outperforming second-rate managers. Makes sense.

A rat-powered hedge fund is unlikely to beat human forex traders. There are some things rats won't do for money.

And they would be hell to manage as well. You couldn't fire them. I mean, people get attached to rats.

It's true, sometimes anyway.

I agree.


A direct application of neural nets.

You know, I'm going to go out on a limb and guess that this is one of that largish fraction of published results that turns out to be horse (or maybe rat) sh-t.

What does this tell us about the EMH?

Unimpressive. Doesn't white noise outperform "some of the world’s leading human fund managers"?

That's right. I think Random Walk Down Wall Street by Malkiel proved that.

However, in a contest between Vampire Squid and Trader Rats, I will go with squid since they seem to have inside information.

Aside from this being an art project, the potentially interesting question is whether the rats follow stable trading rules, and are thus at the whim of reversals, or are they dynamically adapting?

Presumably all they can do with a simple price signal is either trade momentum, mean reversion or some combination of the two. So, what is it they do, or do they randomly assign to one or the other, and then stuck with it?

Or, taking this to the (il)logical extreme (since a simple electric sock and a simple pellet food payoff doesn't reflect any sort of reality!) the "portfolio manager" should double down on the rats that do well, retire them at a particularly high level of wealth creation, and administer death blow shocks when they are massively leveraged and lose...this will impose the evolutionary mechanism more effectively!

Seriously, has no one in the entire world heard of the efficient market hypothesis?!? OF COURSE RATS OUTPERFORM HEDGE FUND MANAGERS. *Every* random number generator outperforms active trading!

after costs
and over long periods of time

Reminds me of Jesse Livermore. As it was described in one book I read, he was able to pick up patterns in the price movements that others had a hard time with or didn't have the capacity for... for rats motivated by food, I could see that facility developing. It would be pretty funny if this little art project was able to outperform the market.

This supports my decision to let my dog trade. Her investment in Nestle after it acquired the Ralston Purina company is up a fair bit, but PetSmart has been flat the past year.

Yes the rats outperform by their salary differential. The real genius of the (scaled up) hedge fund business is to have levered up the stock market and gotten massively wealthy through fees on highly correlated investment products. Calpers has figured that out. I think Tyrone still harbors admiration for the rats and his buddies at Jane Street who guarantee his benign and beneficient gaze on the matter for a pittance.

Hedge funds are already full of rats!.

Read closely, folks: outperformed SOME of the world's leading hedge fund managers.

The project hasn't been updated since 2011, so it's all a moot point now.

On the contrary, the mod date of the page was 2014-9-25. I've no connection with the project.
we were now able to make out the really gifted traders and eventually ended the test series with 4 reliable rats:
Mss. Kleinworth
Mss. Coutts
Mr. Morgan
Mr. Lehmann (we were quite surprised to find he made it)

There are video of the rat traders in action. He already has the third gen rat traders :) He claimed that certain trading house has order a thousand trained rat traders from him.

The first generation automated computer stock traders were based on the so called 'zero knowledge algorithm',i.e. pure randomness without resorting to front running or network speed and the researchers were surprised they were consistently making profits. Now that there are significant number of 'intelligent' automated traders the market dynamics must have changed.

A recent paper that showed human stuck to the suboptimal counterprediction whereas the rats adapted to the changed situations.
when faced with virtual competitors, primates resort to strategic counterprediction rather than to stochastic choice. Here, we show that rats also use history- and model-based strategies when faced with similar competitors but can switch to a “stochastic” mode when challenged with a competitor that they cannot defeat by counterprediction.

Comments for this post are closed