Hello, I'm looking for literature with ideas about risk management in statistical arbitrage (or pairs trading) approach. How to identify possible "red flags" or fundamental shifts, suggesting significantly reduce or liquidate positions... How to to set optimal risk limits, define strategy stop losses. May be some books (papers) describing historical examples from commodity and equity markets, when events suggesting breakout of arbitrage strategies. Also should note, that - A ways of finding ideas not interesting right now. - Taking care of liquidity (or how not to become an "elephant in the room") probably not important, I can consider my positions is too small for market.
First of all: Idea generation comes from watching flow. You code after that. Watching flow also alerts you with regards to risk. Build your risk management rules and cut offs around that. e.g. if you traded STIRs in 2018 from just looking at stats alone you would not have realized that the blow outs are due to a shift in sentiment and FED regime. For stat arb to work you NEED to know the markets you trade in because your return distribution is usually skewed to the right. If a spread moves 3sigmas is this an opportunity to add or is this a potential blow out? Any mechanical risk management approach is doomed to fail here. Risk management rule number one for shorting convexity: ALWAYS KNOW YOUR MARKET! Every MM or stat arb/delta neutral guy I know usually sits down and studies at least 1y worth of data by hand and researches the blow outs. Why they happened, the magnitude, infrastructure related implications (e.g. was one leg limit up/down, has there been a trading halt, changes in borrow rates) and especially if they find any indications in other assets. Let's say you were spreading ENB vs Gazprom your risk definitely cannot be quantified by looking at 1000 days of 5min data. For everything else just read "The mathematics of money management" by ralph vince
Haha, I'm not an insider can watch the flow. Ideas come from simple fundamental understandings, like profits of gold miner depends on price of gold, or WTI crude oil correlates with Brent crude oil and etc. Yes, you've got my idea - I'm looking for the way of "fundamental" learning, what factors can blow out. Yes, my mean-reversion holding takes some days or even weeks, so only data-driven tests not enough, I think (for hourly trades with stop-losses it might be enough) Related to Gazprom ... it seems a guy should have a very versatile imagination what can happen on emerging markets. A lot of study needed on cases what governments can do in extreme situations (capital control measures)...