Hacker News Reader

Comment by blfr | Hacker News Reader

Nothing works forever but there are funds like Renaissance Technologies' Medallion which may just be getting heads for a long time but rather seem to have some secret sauce that works for a long time. Usually though they're not very large and don't seek out investments.

View on HN

I think that's an important point. The problem good funds have is not finding more capital but being able to manage it without cannibalizing their own returns. Anybody trying to sell you a hedge fund is probably not selling a very good one.

Yes, Renaissance is an exception. I would argue that their nature of their strategies almost puts them into a pseudo market maker category (I've personally never worked with/for them, but I hear much of their strategy is statistical arbitrage, made possible by favourable transaction fee set ups and fast, colocated execution infrastructure).

Plus apparently some dodgy "creative" tax code interpretations around short-term/long-term gains.

Is it dodgy? You buy a share in a fund, that fund buys and sells, should you pay capital gains on every sale that fund makes? I don't think that has any precedent. You pay tax when you sell your share.

The type of strategy they employed isn't something small time investors could. It was a complex option scheme essentially buying options on their own short term trading.

While what you're saying is true, in fairness there are two caveats:

1. Virtually all profitable strategies run by hedge funds are inaccessible to small time investors nearly by definition, because they have more capital and lower capacity constraints,

2. The partnership you're talking about with Deutsche Bank doesn't account for most of their returns, and was only in effect in the last decade, not decades before.

Any info on the colo'd infrastructure?

Unfortunately not, I'm afraid. I would guess they have much of their execution infrastructure in the same data centre as the exchanges they execute on.

There's a considerable early period (1980s, 1990s) when Renaissance was turning up amazing profits when electronic trading largely didn't exist. That would suggest other methods.

Perhaps they were just market making back then as well? See this link: https://www.wallstreetoasis.com/forums/is-renaissance-techno...

There are several options that have noting to do with competence that allow for such returns. The most obvious is feeding the fund using another fund. Luck is another as the best returns from the largest outlier taking high risks looks great especially if you ignore large initial losses.

Look at their past performance, luck is just too unlikely for their string of successes.

Supose they had a 10% chance the fund is worth nothing in a given year. Over 20 years there is only ~12% chance a given fund makes it but taking such risks significantly boosts returns. Further, assume many such funds and you are only looking at the lucky ones.

This is why you need to make predictions of performance ahead of time instead of analyzing past performance.

PS: A standard trick is to start 20 funds and the 'best' one has high returns. Create another 20 funds to have a new 'best' when the old one reverts to the mean. Thus you need to analyze total returns weighted by funds size of a company not just individual funds.

This doesn't reflect how the industry actually works among funds that have alpha.

Clearly funds try and avoid going broke. However, many funds also have losses much greater than the market including the one we are talking about.

The truth is risk is hard to measure accurately and most Alpha is simply risk hidden from their investors.

Which is why a statistically significant Alpha takes more than a single funds past performance. And how someone just lost a 1 Million dollar bet on this crap.

Not so unlikely if you consider survivorship bias.

Actually yes, it's still quite unlikely. If you formalize your thesis here and actually do the calculation, you'll see that the chance of a fund achieving a consistently annual average 70%+ return over 20+ years is overwhelmingly unlikely to emerge with the number of funds that existed in the same time period.

"Survivorship bias" is a meme that is commonly thrown out, but to date no one I've challenged on it has empirically demonstrated that this accounts for the emergence of ultra-successful funds. Model this out a bit - what is your single unit of trading to judge and what is your time interval? How many other participants are there in the same interval, and how is each unit judged? You can't just judge on an annual basis - no firm has an actual 50% chance of beating the market each year. Funds like Renaissance make hundreds to thousands of trades each day. Moreover, different firms have different chances of beating the market each year.

Basically, I want you to rigorously formalize how a firm like Renaissance maps to monkeys throwing darts at the wall, because as much as people like to use these analogies (coin flipping, etc), they're never empirical. How do you account for a firm that beats the market by an overwhelming margin for 2 - 3 decades and never having a return poorer than the market (and in fact only rarely being down per quarter or month).

EDIT: Elsewhere: https://news.ycombinator.com/item?id=13797635

They don't consistently average 70% returns.

Feel free to list the years they had over 70% returns vs less than 70% returns.

That was a misnomer, you're right. But that doesn't meaningfully impact my point. An average annual 70% return, some years greater (notably, 2008) and some years lesser, but only a very small number of down quarters or months in the same time frame.

Don't forget to include the preceding 30% drop which reduces both the total returns directly and rate of return as the fund operated over a longer period.

Further, they stopped publishing returns suggesting an even lower long term average.

So, you are looking at a biased subset of a funds returns not total returns which greatly shifts the probability's.

The trick is that any non-illegal investing strategy has finite potential, because it is exploiting some sort of mispricing by the market. Eventually either your buying and selling will correct the mispricing, or other investors will catch on and start doing the same thing, spreading the profits thinner among a larger crowd.

This means that if you want to continually beat the market, using any strategy beyond an impressive gut, you need to be continually developing new tricks and strategies, finding new market mispricings to exploit as your old ones dry up. It's like juggling: someone who is skilled or lucky can keep the balls in the air for quite a bit of time, but eventually you're going to run out of tricks, and the best you can hope for when that happens is returning to average returns.

Renaissance is an exception at levels that makes me wonder if it isn't a cover for other methods or operations.

Howw many sigmas divergent is it?

It's a highly profitable market maker more than a long term investment fund.

I'd really like to see deets. I've looked, somewhat, and what I've found (including Jim Simon's very few talks) have been ... pretty uninformative.

You're not going to see details. No one who works there (or at similar firms) has any incentive to give away that information. If you come across people who work at one of these firms, you'll find they are extremely cautious about saying much of anything about their work.

The comments you see on forums like this one are just best guesses - educated guesses, but generally underinformed and relatively out of date. I personally consider it a fluke that RenTec is as famous as it is - other comparable firms like TGS have far less notoriety (and are probably better off for it).

Reply on news.ycombinator.com