macrocephalopod

1/2 Random risk thoughts. - Loss cutting is key, but at what level? Recently, seems wider stops are needed; ideal thresh may b changing over time, needMoreData - Hard2know if a stock drop is abnormal; need models. Rebalancing can distort. - Risk exposure not obvious; I own

I'll give you the classic quant perspective. Quant traders don't tend to think in terms of profits and losses. You have an alpha signal which, let's say, is a prediction of expected returns. You have a risk model which models volatility, correlation and sets exposure limits.

Ignoring trading frictions, the job is to maximize the ratio of expected returns (given the alphas) to some measurement of risk (from the risk model) - subject to the other limits imposed by the risk model.

We wouldn't really think about "cutting losses" in a trade other than... If a position has gone down a lot, it would appear to be "riskier" in the risk model - which would cause size to be cut If +ve auto-corrs are modelled in the alpha, this might cut size on a down leg too

Usually, recent P&L in a position isn't correlated with future P&L - which is why stop losses and the like are usually not the best way to be managing position risk.

Very insightful, along w what @saah1lk mentioned. In my scenario, these trades aren't really quant tho, they're discretionary - trying to get a feel for what themes the market likes (recently, cyclical/value) and finding names w value, growth, & momentum, & trade price action

So for this type of thing, the alpha can't really be modeled very well, bc sometimes awesome stocks breakout, fake out, then do nothing. As far as stop losses, though they are crude, for discretionary trades they are useful. Tudor Jones and other legends found value in them

Yeah but everything trended pretty strongly back when PTJ, Druck etc. Nowadays that's not true. A stop loss is a bet on positive autocorrelation in returns. It is useful to the extent returns are positively auto-correlated. They usually aren't.