Frosty's trading bot goes live part 2

[nonlinear5]

There is a big hole in the IB's algorithm which drives the simulated account. Take a look at the pic below. I stacked two windows so that the comparisson can be made. The top window is my real account, and the bottom window is my simulated account. Notice a huge difference in the ER2 quote between the two accounts. You can easily and reliably reproduce this problem yourself:

1. In your simulated account, place a bid well below the current bid and place an ask well above the current ask.
2. Place two market orders: one buy to hit the current ask size, and one sell to hit the current bid size.
3. After your two market orders execute, your ridiculously low bid and high ask that you made in step 1 become the current market!
4. This will presist until another simulated trader will outbid you or a trade occurs in the real market, at which point the simulated bid/ask will be updated.

In summary, in a relatively thin market, the bid/ask on a security in a simulated account is whatever one (or a couple) simulated traders decide it to be!

This is much less likely to happen is a more liquid market, such as ES. Not sure how this may have affected you, Frost, but this abnormality in the IB's simulated trading is definitely something to keep in mind.

 

[maxpi]

Quote from austinp:

<i>"Maybe, however I don't really believe "they" are smarter than me."</i>

If the Native American Indians had equal weaponry as white settlers = invaders did, this country would never have been taken from them. Many of the indian leaders were absolute military geniuses, and far more wise & intelligent (in every way) than their European adversaries.

There was no comparison of armament, hence the indians' demise.

==

Not really, the Amerind Warrior could distinguish himself by acts of valor, the Caucasians would distinguish themselves by owning land. The White guys, with better, longer term goals, simply showed how goals make the difference. The better weapons were available to an Indian that could get one, and many did, it did not make up for the goal aspect.

Personally, I don't think the pros have much advantage over a determined individual with goals and willingness to put in the work. Some recent psychological findings show that experts are made, not born, and it takes 10 years of concerted effort to master something.

 

[TSGannGalt]

It seems like people believe access to equipment is a major difference...

Though who can actual go in there and be able to use it?

Who can actually program in Java/C++? Work their way around UNIX/Solaris? Write a SQL script for Sybase? Who has experience writing their own execution platform using FIX?

I've been in this business for quite sometime... And I'm still far from being able to apply the access I have to the full potential. Actually, I'm currently teaching myself to write a better UNIX Shell script.

So before someone envies and feels jealous about large hedge funds and banks... They need to realize what it takes and how much is required on the systematic traders and developer. I don't believe institutional guys are smarter... we are just expected to do more...

 

[TSGannGalt]

Quote from nonlinear5:

There is a big hole in the IB's algorithm which drives the simulated account. Take a look at the pic below. I stacked two windows so that the comparisson can be made. The top window is my real account, and the bottom window is my simulated account. Notice a huge difference in the ER2 quote between the two accounts. You can easily and reliably reproduce this problem yourself:

1. In your simulated account, place a bid well below the current bid and place an ask well above the current ask.
2. Place two market orders: one buy to hit the current ask size, and one sell to hit the current bid size.
3. After your two market orders execute, your ridiculously low bid and high ask that you made in step 1 become the current market!
4. This will presist until another simulated trader will outbid you or a trade occurs in the real market, at which point the simulated bid/ask will be updated.

In summary, in a relatively thin market, the bid/ask on a security in a simulated account is whatever one (or a couple) simulated traders decide it to be!

This is much less likely to happen is a more liquid market, such as ES. Not sure how this may have affected you, Frost, but this abnormality in the IB's simulated trading is definitely something to keep in mind.

Conclusion is, he should have tested his system considering all the potential slippage.

Correct?

 

[nonlinear5]

Quote from TSGannGalt:

Conclusion is, he should have tested his system considering all the potential slippage.

Correct?

I think it even goes beyond slippage. Look at the attachment. I made the ER2 market to be $100 bid, $200 ask. If Frost's bot is running right now against a sim account, I don't know what would happen.

I looked at my sim account value, and predictably, it dropped from $80,000 to $7,000. If anybody's bot is making reqAccountUpdate() and reqPortfolio() calls at this time, she (the bot, that is) would be panicking.

 

[bigmoose]

Frost, that is why I asked before if you traded intra-bar in a previous question of mine. I am not sure IB sim account and real account get the exact same data feed...

 

[maxpi]

NonLinear, thanks for the post. I was just on the verge of starting some tests with the sim account.. now I have to make some comparisons of the YM in the two accounts and see about getting feedback on fills and whether they would happen with the real market.... IB has time stamps on orders to the second, it will be easy to check all simulated fills with real data after the fact.

Earlier today I was comparing quotes from Tradestation and IB sim account on the YM in Multicharts.. I actually did not see a difference... that was very anecdotal of course.

I just did a market sell of YM, 1 contract, in the sim account, it reduced the size on the bid from 11 to 1!! Assuming mine was the only order at that time, and it is really slow currently, something is messed up bigtime... I'm not sure what that was about, I tried another order and it had the effect it should have.. probably, as long as you sim trade one contract and during RTH things should be basically ok.

 

[BigBubba]

my best ats i have running on IB's simulated acct. so, i applied it to the same data period on real acct. was still decent.. but an obvious change noted on the equity curve. trades today were different.. worse actually. looks like i have some work to do..

 

[maxpi]

Quote from BigBubba:

my best ats i have running on IB's simulated acct. so, i applied it to the same data period on real acct. was still decent.. but an obvious change noted on the equity curve. trades today were different.. worse actually. looks like i have some work to do..

Probably if we are using the sim account we have to go back at the end of the day and edit the results for bad fills. I hope that is doable.

 

[BigBubba]

i'm gonna rebuild mine on the real data.

i won't try to patch it up. it's obvious the system behaves a bit different. i don't know precisely how it's different than the sim acct.
..so i best not take a chance.