Hacker News Reader

Comment by mac01021 | Hacker News Reader

I can't speak for Quantopian but traditionally the goal of a "hedge fund" is to preserve your vast wealth through a diverse range of market conditions. This means it should substantially overperform the index funds when they're all tanking, but could very well underperform them when they're doing well.

I think probably this only makes sense if you want to be able to liquidate your assets at any time without having to worry about taking a tremendous loss if the market is down.

View on HN

Wealth preservation isn't really the point of most hedge funds. Bond index funds that invest in only the highest rated sovereign and corporate debt are cheaper ways to preserve wealth which outperform stock index funds when they're tanking and underperform them when they're doing well. Hedge funds serve other purposes.

But if you're heavily invested in bonds then you can't "liquidate your assets at any time" like the commenter above you said. It sounds like you're talking about different investing goals, not really refuting their claim.

What are you talking about? Bond index fund holdings are extremely liquid. And even if you directly hold individual bonds in your own account, investment grade bonds from large issuers are also very liquid. The vast majority of wealthy hedge fund investors still don't have investments large enough that liquidity becomes a serious concern for bond trading.

Well given that quantopian hasn't existed long enough for their machine traders to be trained during adverse market conditions, I wouldn't hold my breath.

Trading algos are trained on historical market data, this can (and should) include crashes etc.

Obviously, but 1) the data are not the same then so it's possible the algos just learn that featureset and not the actual patterns, and 2) I'm more talking about the meta-pruning and algorithmic balancing quantopian uses to determine which bot wins, which fundamentally can't be trained on the past.

Back testing based on OHLC on a crash day isn't very useful for many systems.

The bid/ask of instruments become illogical. Bad trade data in the live feed won't be in your historical data.

Intraday during a liquidity drop is mostly based on your relationship with your broker. How they handle margin calls will decide if your trades survive.

For example, traders have had margin calls on option spreads with a known fixed risk at the time of opening. But the individual legs of the trades were making very incorrect markets, so the broker algorithms issued margin calls and liquidated at the worst time possible.

Reply on news.ycombinator.com