Hi,
since I mainly trade rather illiquid microcaps, I naturally encounter rebalance dates where the stocks I want to sell/buy just have no proper bid/ask, volume or both. I account for this in my simulations with high slippage respectively and take precautions (liquidity filter, limiting turnover, etc.). In realtime, for my manual trading, I try to apply some "tricks" to soften the blow, e.g. placing a limit at mid price first to check if I get a better execution as expected (which often works at US exchanges). If not possible, I often take partial execution with a limit order at bid (or ask) with as much volume I can get and check later if I should adjust the limit (especially in thin exUS trading).
Just wanted to ask if you have some additional tips and tricks from your personal trading or experience regarding order types, best trading hours, best trading days etc.
Maybe you can also comment on liquidity and placing orders in phases like 2009 or 2020 (I was more of a SMIDcap trader in 2020, so liquidity was not a giant issue)
I use VWAP orders and place them over several days, often skipping days, all in an attempt to avoid market impact. A lot depends on your broker. IB has a percentage-of-volume limitation on VWAP orders and I'm not sure what it is. With StoneX (a prime broker) I use a 50% of volume limit but with Fidelity I don't. I've been known to trade 200% to 300% of volume in one day, but it's expensive. For extremely illiquid stocks it might be worthwhile paying your broker to execute a not-held or desk order. I do that for quite a few stocks.
Thanks for your reply. I assume you are moving much more $ than me. Can you roughly tell what's your personal border definition of "extremely illiquid"?
I personally prefer to trade stocks with sub5% spread and sufficient offer to swallow my order but that's not always possible of course. Eg. on WSE microcaps, I sometimes have maybe 3% spread and up 3% market impact if I eat through the order book with a 5-figure trade. The problem I have is not knowing what is the better compromise:
a) Scaling into (out of) the illiquid position over several days risking that the price moves away from signal entry but potentially saving $ by not blowing up the orderbook if it doesn't.
or
b) Trying to execute 100% during the rabalancing date losing $ to market impact but staying "true to the signal"
I have a rather low-turnover strategy, still, my entries can be based to the big part on analyst revisions and surprise signals. In those cases I'd assume the signal has priority over market impact concerns? On the other hand, market impact is a certain loss. Short-term signals and price movements are not. Not sure what would be the best way to find a rule of thumb for decision-making in this regard.
Nice! I remember from a previous post that you mentioned randomizing which days you skip. For most VWAP orders, the VWAP algorithms also randomize the timing of trades throughout the day, so other traders can’t easily detect your pattern or exploit your orders.
I know you already get this, but I just wanted to highlight how cool and clever that approach is!!!
If I place a 50%-of-volume VWAP order and less than 20% of it actually gets filled, then for me it's an "extremely illiquid" stock. Some examples are WSTL, AKT.A:CAN, TI:CAN, and IDXG.
To solve this problem try to figure out your "delay cost." What percentage return are you giving up by delaying your trades? You can base this on your daily return, maybe doubling it because entry and exit signals may tend to yield more than you would get just sitting on a good stock. Then try to figure out the market impact cost and see whether it's higher or lower than the delay cost.
Is there a good way to test fixed signal delay? Simplest way would be going from weekly to biweekly rebalance but this seems not very clean because the timing can be anywhere between 0 and 10 trading days of delay.
Yes. Because ranking is done on Mondays, use the following rules with daily rebalance. For your base case, use WeekDay = 2 or (WeekDay = 3 and PrevBarDaysAgo >= 3) . Then for a one-day delay use WeekDay = 3 ; for a two-day delay use WeekDay = 4 ; for a three-day delay use WeekDay = 5 ; and for a four-day delay use WeekDay = 6 .