Rats trained to play the forex market | Hacker News Reader

Rats trained to play the forex market

artmarcovici.com

56 pointsmjtokelly21 comments

Much like the monkey throwing darts at the stock market chart who clocks professional money managers on average, this result proves nothing about rats or markets. It provides further experimental proof for the worst kept secret in history: professional money manager systematically destroy value.

While it is certainly true that professional money managers don't beat the market on average, this does not imply that they destroy value.

Professional money managers and speculators in general, at least in principle, more efficiently allocate capital as a result of their speculative activities. This causes overall returns to increase. In a world without money managers/other speculators, market returns would be lower overall.

The simple model: in a world without money managers/speculators, overall returns might be 2%. In a world with 1 speculator, returns might be 2.1% and that speculator might achieve returns of 4%. In a world with many speculators, they might all achieve returns of 4%, and none of them would be beating the market. They are creating value, however.

(Another reason they don't beat the market is that many of them are not trying to. The manager of my Vanguard Target Retirement 2050 fund is currently trying to beat the market. In 2040 his goal will be to minimize risk.)

The stock market has little to do with allocating capital. The bond market does, but the stock market is mostly about valuing short term and long term risk with associated rewards.

Huh? The stock market absolutely has to do with allocating capital. It, together with the bond market, is the place where large companies go when they need to raise funds.

Bond traders and stock traders alike both balance the risk and reward of a company's offering in order to determine a fair value for that offering. Valuing risk and reward is valuing the company and valuing the company is the most critical part of a stock offering. Any post on here about raising money from a VC will talk about the valuation of the startup.

A small percentage of stock market transactions are Company to Investor. Most transactions are Investor A to Investor B which only indirectly impacts a company’s value. Clearly they are related, but it's generally much better for investors in healthy companies for that company to get money from the bond market than the stock market.

PS: Consider a company with consistent revenue/dividends but zero transitions on the stock market. Its value is going to be heavily influenced by stock market conditions, but it's ability to raise money is going to be dominated by the state of the bond market.

Edit: My point is what it means when a transaction between Invester A and Invester B happens on the stock market at a given price.

It is often better for a company to raise money from the bond market, but only up to a point and only because of the tax implications.

http://en.wikipedia.org/wiki/Trade-off_theory_of_capital_str...

If you take 80 people and tell them to randomly play the stock market, buying or selling - you will always have some that succeed and some that fail over the short term.

The question is, if the same people who 'succeeded' are put back to the start again and play randomly with a new batch of 80 people - would they still rise to the top, or perhaps this time have worse luck.

Comparing this to the rats, the long term testing / new rounds with a new set of 80 rats (including the elite) would help determine whether these rats are the best, or if it was chance.

If I understand this correctly, the decision the rat makes is based solely on hearing a certain frequency (pattern of trading emerge), unfortunately even if this is the case - trading by patterns could be extremely risky without the common sense / experience of a real trader to see what is really happening to the market.

What the rats are doing (if this is not a well done prank) is exactly equivalent to what 'technical' traders and chartists do. As to why these techniques work; they are reacting to the intentions of other traders signaled in data. This may be a case where knowledge of the "underlying reality" that the market is reacting too (supposedly) is a disadvantage because it prevents noticing and reacting to the actual data.

That said, this whole thing reads like an elaborate practical joke.

You mean, like they're noticing a 'bearish gartley' pattern (in audio) and acting appropriately?

My point was, that although these can be 'indicators', you still require a human mind to determine whether the trade should be made. If it was as simple as matching patterns and trading, anyone would be able to do it.

Not necessarily, if (again taking this at face value) only a few rats are able to pull this trick off, that's not too surprising. Of humans who trade forex, only a certain percentage are relatively successful at it.

These rats clearly deserve $1000,000 bonuses. They are the indispensable financial innovators which drive our economy. These rats are too big to fail.

They are providing valuable liquidity to our financial markets. Is it too much to ask that they be rewarded for their contribution to society.

Seriously, this is the most ingenious thing I have seen in a while.

Marcovici is a conceptual artist and this is satire. Wasn't sure if some comments here thought this was for real.

Take a look at his satirical illustrations about BP here:

http://sites.google.com/site/artmarcovici/

This is the first thing I thought of when I read this:

http://dilbert.com/strips/comic/2010-07-03/?CmtOrder=Rating&...

I wish he had compared the rats' performance to the current publicly available "best of breed" programmatic approach that uses technical indicators. Other than that, the interesting part of the experiment seems to be how it may relate to training human traders more efficiently. Starting a fund based on trading with rats seems a bit satirical indeed..

Perhaps there is an elaborate rat society running around mazes and pressing switches somewhere in the bowels of Goldman Sachs?

Dupe: http://news.ycombinator.com/item?id=514798

Survivorship bias explains his results.

He claims after selectively allowing the best performing rats to breed the second generation of rats performed better than the first generation. If it were survivorship bias the second generation should be no better than the first.

How is this different than training a neural net on historical data?

That is ratsome.

Reply on news.ycombinator.com