IB adaptive algo - how effective is it?

[Abundance Magnet]

I'd argue anything is better than a simple market order, unless you're dealing with something like SPY and relatively small orders.
Have you tried comparing it to something like CSFB's Guerrilla?

I'm almost not that surprised. I had the sense that the algos were more marketing than actually effective. I have even found their limit orders slower than other brokerages. Ie. Two lmt orders but IB was slower to fill. Its happened a few times now so I don't think k it is a coincidence. Makes me wonder about their performance edge studies that claim they are provide better fills than the competition. An IB funded study (or studies) that eventually proved they were superior perhaps?

D08, do you still use IB but have your own algos? Or are you institutional, and on institutional platforms?

Thanks

 

[GRULSTMRNN]

Out of curiosity did you compare vs guaranteed vwap (for which most banks charge several dozen basis points depending on name, liquidity, and corporate action) or attempted vwap?

My opinion might be outdated but a few years ago it was poor. I didn't do quantitative analysis but it didn't seem to catch enough liquidity and it was poor compared to VWAP.
The IBAlgos I've tried have all been subpar compared to competition.

 

[GRULSTMRNN]

I believe your statement needs to be better qualified. If someone wants to get into a position no matter the cost (which is a reasonable thought for highly liquid stocks) then its very hard to beat a market order.

I'd argue anything is better than a simple market order, unless you're dealing with something like SPY and relatively small orders.
Have you tried comparing it to something like CSFB's Guerrilla?

 

[d08]

I'm almost not that surprised. I had the sense that the algos were more marketing than actually effective. I have even found their limit orders slower than other brokerages. Ie. Two lmt orders but IB was slower to fill. Its happened a few times now so I don't think k it is a coincidence. Makes me wonder about their performance edge studies that claim they are provide better fills than the competition. An IB funded study (or studies) that eventually proved they were superior perhaps?

D08, do you still use IB but have your own algos? Or are you institutional, and on institutional platforms?

Thanks

I use CSFB algos now alongside IEX routing. Both are available at IB. I have use my own algos but they're very basic, just TWAP (randomize size, time but keep a certain interval).

 

[d08]

Out of curiosity did you compare vs guaranteed vwap (for which most banks charge several dozen basis points depending on name, liquidity, and corporate action) or attempted vwap?

I compared to the historical VWAP on IB. Should be a good enough metric or do you disagree?

I believe your statement needs to be better qualified. If someone wants to get into a position no matter the cost (which is a reasonable thought for highly liquid stocks) then its very hard to beat a market order.

It depends on the liquidity at the time. Even with liquid names, during the quiet hours when the price is not moving, putting on some reasonable size will alert the HFT and the price moves away. What I see on the screen offered and what I realistically can get are two very different things. Now if you're trading 2000 of SPY near open or close, obviously it will not make a dent and market orders are the best (I use them myself).

You mentioned trading fundamentals or news, so it makes sense to get a position whatever the cost. Another scenario is a major trendline break. You want as much as possible as fast as possible.
For other, technical styles, when you're not trading something obvious - slippage matters and it's much better to attempt to hide size (not literally).

 

[GRULSTMRNN]

Agree with everything you said. I think historical vwap is a better measure because it does reflect cost of execution other than commission in the most realistic way via price and volume. Guaranteed vwap would skew the picture because it is a metric each liquidity provider prices differently as function of their own inventory, block and program trades they see and other variables that may differ from house to house.

For my algorithmic orders I never send market orders, so yes, only for discretionary trades when I am fairly certain of liquidity, available.

I compared to the historical VWAP on IB. Should be a good enough metric or do you disagree?



It depends on the liquidity at the time. Even with liquid names, during the quiet hours when the price is not moving, putting on some reasonable size will alert the HFT and the price moves away. What I see on the screen offered and what I realistically can get are two very different things. Now if you're trading 2000 of SPY near open or close, obviously it will not make a dent and market orders are the best (I use them myself).

You mentioned trading fundamentals or news, so it makes sense to get a position whatever the cost. Another scenario is a major trendline break. You want as much as possible as fast as possible.
For other, technical styles, when you're not trading something obvious - slippage matters and it's much better to attempt to hide size (not literally).

 

[dstyers]

I'd argue anything is better than a simple market order, unless you're dealing with something like SPY and relatively small orders.
Have you tried comparing it to something like CSFB's Guerrilla?

No, I have not. Thanks for the suggestion.

 

[dstyers]

I believe your statement needs to be better qualified. If someone wants to get into a position no matter the cost (which is a reasonable thought for highly liquid stocks) then its very hard to beat a market order.

This is my situation exactly. I cannot attest to the efficiency of adaptive algo for limit orders ect. I have only used the adaptive market (patient) on a few of my algorithms which require speed of fill, rather than efficiency of fill as in entry.

 

[GRULSTMRNN]

Then I would consider a limit order that is by a couple points already within the current price level. It grabs best bids/offers right away but protects against predatory algos to a limited degree

This is my situation exactly. I cannot attest to the efficiency of adaptive algo for limit orders ect. I have only used the adaptive market (patient) on a few of my algorithms which require speed of fill, rather than efficiency of fill as in entry.

 

[Lee-]

I use adaptive on illiquid option orders occasionally. When I've set it to patient, sometimes it would take maybe 30+ minutes to sort of walk the price up to the limit. However, for example, on Friday I gave it another try on "normal" setting and it appeared to just about immediately set my order to the limit price and get a fill. This is in contrast to the last time I tried it (a couple months ago) when I used normal and it still took a few minutes.

Without details as to how the algo works, we can only speculate. I suspect they have some statistical metrics about fills that they use to determine how to work and that's why we see different behaviors. Possibly depends on underlying, time of day, and so forth, but I really don't know. I know when trading options with large spreads, it appears to be inconsistent in how quickly it bumps the limit price. I've have not looked to see if it favors certain exchanges or moves my order around exchanges.