Post New

Interactive Brokers has a detailed transaction cost analysis (TCA) which provides a lot of information about trades.
Has anyone looked at it and have recommendations for interpreting?

Friction is important to minimize especially if one is actively trading screens.
Back in the day we spent a lot of time, for example, trying to evaluate FolioFN transaction costs.
See for example http://www.datahelper.com/mi/search.phtml?nofool=y...

Here is a top level example IB TCA for some of my recent trades.
These trades were all done using basket trader with the MidPrice algo.
https://1drv.ms/i/s!AmCzJdeikH-ciop2bheJHAmsW9R5Ng...

They report price improvement in basis points versus various standards - order arrival time, later times, close, and several VWAP references.
Seems to me the most useful one to estimate actual friction is the "vs VWAP" column.
This means, as I understand, the execution price vs the VWAP (volume weighted average price) from the time the order was submitted to when it was executed.

In the case of the above trades, the figure is (0.108) basis points or a loss of 0.108% average per trade.
Commissions using IB Pro fixed tier seem to be around 0.08% on average.
Would it be reasonable to use 0.108%+0.08% = 0.19% as an estimate of one-way friction?
Rounding to 0.2%, this would correspond to using 0.4% friction in GTR1, as GTR1 applies all the friction to sales IIRC.

Just wondering if this is a reasonable way to estimate real-world friction or if there's a better way.

Mark

Isn't the Friction that's relevant the 'difference between the final transaction cost & the price that the backtester (for example GTR1) uses'? That is the difference between the transaction price including fees, taxes etc. & the day's closing price in the case of the GTR1.

That's one of the reasons why I've always favored the MOC (market on close) transaction type at IB for MI work. In that case the friction is essentially IB's fees only x2 which is lower than the traditional 0.4% friction. And of course if one could reliably transact at 'better than the close' pricing then friction could even be negative.

Isn't the Friction that's relevant the 'difference between the final transaction cost & the price that the backtester (for example GTR1) uses'? That is the difference between the transaction price including fees, taxes etc. & the day's closing price in the case of the GTR1.

That's one of the reasons why I've always favored the MOC (market on close) transaction type at IB for MI work. In that case the friction is essentially IB's fees only x2 which is lower than the traditional 0.4% friction. And of course if one could reliably transact at 'better than the close' pricing then friction could even be negative.

Thx. Have you tried that enough to confirm that it works? Meaning that your trades always come in at the published close price?

I'm pretty sure I'll try that some time, perhaps soon. There are a couple points of concern. One is that many SiPro screen picks do not trade a lot of volume.
Even with a filter of $750K daily over last 10 days, you will have picks that go through stretches during the day where there are no trades.
Even if you have a modest order, how is the close price determined if yours happens to be the only order at the close?
Maybe you get the close price, but your order has affected the close price.

Another concern is that IB routes trades to different exchanges. Most of mine have been ISLAND, some IEX, and some DRCTEDGE, whatever that is.
It would be interesting to see if trade prices always match the close price.

One last point is that from my statistics in the past, it was better to buy earlier in the day rather than later. (Selling did not matter as much).
So maybe trading buys with MOO orders and sells with MOC orders would be a strategy. For a fully invested account that might mean using margin.

I guess I see three possible trading strategies for actively trading MI screens at IB: using MIDPRICE, using a VWAP algo, or using the MOO/MOC strategy.
It would be interesting to figure out which seems to be the most reliable.


Mark

I'm pretty sure I'll try that some time, perhaps soon. There are a couple points of concern. One is that many SiPro screen picks do not trade a lot of volume.

For my smaller volume SIPro picks, I have been using Vanguard and limit orders. For buys, I place the limit at the ask. If the spread between bid and ask is more than a couple of cents, my execution price is usually better than my limit price. Similar experience with selling, execution price generally better than the limit price. I have not tried this at IB as the screen I use at IB is for very liquid stocks and I trade market orders.

Craig

One last point is that from my statistics in the past, it was better to buy earlier in the day rather than later.

My experience, anecdotal since I kept no records on this item, was that I got perfectly fine prices when I bought and sold shortly after lunch time, Central Time (a time frame driven by my daily schedule more than anything). "Perfectly fine" for me was usually a price better than the extant market price by a few bucks across the whole order; although a few times I got a worse prices of roughly the same magnitude. This was with TDAmeritrade; now that I've been forced to Schwab by that acquisition, I'll find out whether that's still the case.

Eric Hines

I've had a retail trading & IRA account with IB for decades. I'm very happy with them.

Last year I also opened up a separate paper trading account with them to test out different portfolio ideas & trading strategies. No charge, but as best I can tell I have to log in/out to switch accounts. Except for that I've been happy with that, too.

Thanks for all your contributions, everyone.
Happy New Year.

:-)Shawn

Thx. Have you tried that enough to confirm that it works? Meaning that your trades always come in at the published close price?
I've done MOC orders for my MI work and it's always worked as intended.

You can also enter the order more or less at any time as long as it's earlier than 30 minutes before closing (& allowing for a little price movement when calculating the amount of the stock to buy) and you'll always pay the official closing price + IB's fee. It results in minimal friction (as I've defined it above which in this case is just IB's fee) and so I use a far lower friction number for myself than the standard 0.4% used normally.

Post New

Mechanical Investing FAQ

---

Contact Shrewd'm
Contact the developer of these message boards.