Does anyone have any data on the tracking error of IB’s ‘best efforts’ VWAP algo vs. true VWAP, and how it varies by liquidity band? I’ve been meaning to email them, but haven’t gotten to it.
Thanks.
They’re pretty secretive about everything relating to their algos. I’ve scoured the internet and talked with support and never come across any hard data from them.
But, they added a transaction cost analysis feature a while back so I have a good amount of data from my own trades. (You may be able to find data from other traders with a google search as well.) I’ve placed over >1,000 orders using the IB vwap algo with some slight settings changes. About 80% all of those orders were in the 5-10% ADV range. And over all those trades I’ve trailed the true daily VWAP by 6 bps, and lost to the daily range (H/L midpoint) by 3 bps.
The rest of my evidence is just anecdotal, but in my opinion the IB algos are as good as any. Tracking error is minimal at <5-10% adv. Personally I would use simpler orders or set a vwap to complete in a short time period for <5% adv, and spread trades >10% adv over multiple days.
Thanks for sharing. Are these mostly lower liquidity stocks? Have you been using VWAP with an upper limit in these comparisons or just VWAP? I haven’t done a formal study, but have seen some really bad fills in a small initial VWAP test on lower liquidity stocks (ADT100 at $500k). I think the duration I’d want VWAP to fill in would be related to system type. For fast turn systems with rapid factor decay, I’d want really quick fills - but for most of my systems (3 to 6 month turnovers), ADT is fine with fills over a full day or 2?
Those trades have been almost entirely in small cap stuff with mktcap 100m-1B area. I don’t use a limit.
My preference is to try to never take liquidity (resulted in $.0018 per share over a very large sample for me), set vwap with an end time that should make my order comparable to about 5% volume participation assuming normal daily volume, allow trading past the end time if necessary, and set a max participation threshold at 10%. Ideally you can set the end time to be at least an hour before close, otherwise the algo often fails to fill the entire order.
But adjusting the end time on a vwap order is one of the most flexible all-around order types and I really recommend it in most situations. Personally I always prefer to let my orders fill over a few hours even with high turnover systems. I just get better fills and lower commissions if I allow it a little time. But, when volume really disappears this order type isn’t good.
I’m curious what order type you found success with “when volume really disappears?” I currently have a position that I’ve been holding for a year that has had liquidity dry up since when I bought, and now I need to sell around 100% of the average daily volume. All of the action seems to happen near the close, and I’m only able to sell a few lots per day. Since this is a slow system, I should be able to sell over a couple of weeks or so.
Armando,
If your primary concern is just closing the position with minimal market impact then an algo like vwap of volume participation are probably still best. Just make sure to set a max volume participation level that you don’t think will cause you to move the price too much. Personally I’m comfortable with 10%, and I’ve seen other experienced people recommend higher than that. It may take a couple weeks to move 100% adv, but if the stock has any volume at all you’ll be able to trade with its flow.
If it would be beneficial to close more quickly though then it doesn’t hurt to check for liquidity in dark pools. Just use a limit order or something similar and route it to the dark pools only. More often than you would think there is someone checking for bulk lots in dark pools and you’ll just fill your whole order there in an hour.
There are more aggressive algos for crossing the spread and executing orders as quickly as possible, but I have no experience with them and can’t think of many circumstances in which they would be worthwhile. If you start crossing the spread and trying to fill abnormal volume the price can move against you very quickly.
As long as you’re never in a position where you NEED to close the position then I advocate sitting on the favorable side of the spread with small lots and trying to trade slowly with the flow of volume.
That makes sense. Thank you. I’ll experiment with using crossfinder in addition to the usual vwap. I’ve also been using it with similar params, and will also experiment with adjusting end time as well. I’ve been using vwap with a wide limit price, but sometimes get a bad fill if the volume comes in at once at the worst price. Do you not use a price limit with your vwap orders?