hft, What inventory management algos do you use in your mm strats? (ex: If you get a position building up and the market goes more and more against you, how do you hedge? When will you liquidate the whole inventory? will you still make markets with higher spreads? edgerequired = f(inventory,...)) thank you
Yeah, a combination of the above. Complete liquidation is usually the last resort, but it's certainly there as an option.
Hft, Can you be more specific (a link, an emoticon... ) ) about the inventory management algos used by hft firms? thank you
EdgeRequired can scale linearly or in a number of ways as you suggested. As far as hedging, ideally you want to hedge FOR edge, but as your inventory builds the edge you require to hedge would lessen and go negative (so if you can lose a half tick hedging vs 3/4 tick puking in the market that your inventory is on from, do that). In all honesty it doesn't happen all that often. A vast majority of trades you get into are scratches that end up negative due to fees, with a few winners thrown in to make money, followed with a very small portion of gross-negative trades, at least in my style of trading. In other words, I'm very aggressive at not putting on inventory and at scratching at the first sign of trouble, at the price of losing opportunity cost. That's just me though, other strategies I've seen (that are very successful) have a much higher risk/reward tolerance.
Which type of hft strategies are more popular on cme - market taking or market making? Does this depend on product - ES, CL, GC ? Many thanks!
In general, rule of thumb is that bigger tick sizes/values lead towards more market-making, and vice versa. So there's more HFT MM strats on ES than on CL and GC, but in most of the popular CME products there is so much liquidity and volume that there is plenty of both HFT taking and making going on and you wouldn't characterize the product, let alone the entire exchange, as dominated by one hft practice or another.
Hi hft, thanks for keeping this thread alive, it's very interesting. i have a question, Does one has to work on a predictiv model and then figure out if backtests are good, Or, Do you work on the equity curve, i mean you run a "naive" strategy and then you try optimise this curve trying to filter out loosing trades/add profitable trades ? Cheers
Definitely not PNL-based in market-making. Market-takers will use it sometimes though. Predictive model is more like it, though it varies from person to person. Real post-trade analysis generally prevails over back-testing, since it's hard to accurately simulate fills in HFT MM'ing.