|
Steve2
Posts: 750
Joined: 10/11/2012
Location: Annapolis, MD
 User Profile |
I started OV live trading late last week. I plan to collect detailed slippage data that I'll share from time to time. The attached table shows you what I'm collecting. If there's any other way you'd like to see the data please let me know and I will try and accomodate. If anyone would like to share slippage data for gxtrader, it would be greatly appreciated.
Steve
Caveat: Don't draw any conclusions from the data so far. It only represents 12 trades :)
 Attached file : IB Slippage 3-12-13.jpg (187KB - 282 downloads)
|
|
Jim Dean
Posts: 1059
Joined: 10/11/2012
Location: L'ville, GA
User Profile |
Probably too much to ask - but if you could collect the ATR(14) and the EMA(V,14) related to the symbol associated with each entry and exit, it would be very valuable for creating useful parametric models.
You could create two custom focus list columns in OT to display these values btw.
|
|
hafnium
Posts: 43
Joined: 10/11/2012
Location: Vienna, VA
User Profile |
Steve2
I don't understand the request or logic for collecting Gxtrader slippage.
If you place a MOO in GXtrader before Mkt opens - the OV price is the GXtrader MOO price. There is 0 slippage. Am I missing something?
LD
|
|
Steve2
Posts: 750
Joined: 10/11/2012
Location: Annapolis, MD
User Profile |
LD, not really sure. I thought that MOO orders were guaranteed to be filled within the opening range but if say 100 users submitted MOO orders for EBAY to GXtrader they might not get the exact same fill price. Are you saying that the entry and exit prices you get on GXtrader orders always match OV entry and exits? Also, have you had any issues with MOO orders being cancelled or only partially filled?
Steve
|
|
Steve2
Posts: 750
Joined: 10/11/2012
Location: Annapolis, MD
User Profile |
Jim, I knew I shouldn't have asked... :-) Let me take a look and see how much work it would be to do that. Will let you know.
Steve
|
|
Jim Dean
Posts: 1059
Joined: 10/11/2012
Location: L'ville, GA
User Profile |
#indicator
Return atr(14)
#indicator
Return ema(V,14)
Save those two indics, then add them as custom FL columns. Put the symbols used by the strats in the list, and sort by symbol name.
You might also want to record the overnight change:
#indicator
Return O-C[1]
Voila!
|
|
hafnium
Posts: 43
Joined: 10/11/2012
Location: Vienna, VA
User Profile |
Steve
Your point is well taken. When Omnivest trade volume ramps up - it may become an issue. However to date - with 4 months of live trading
My MOO price with GXtrader (entries and exits) have allways matched the OV site price. There is no slippage. Also to date - no orders cancelled and no partial fills.
Perhaps someone with a larger account (six figures and up) may have had a different experience.
LD
|
|
Jim Dean
Posts: 1059
Joined: 10/11/2012
Location: L'ville, GA
User Profile |
I believe the issue will become relevant when real time trading is implemented. VERY relevant.
|
|
BrianD
Posts: 302
Joined: 2/23/2013
Location: Grand Rapids, MI
User Profile |
I know from experience IB "Smart" has made me "smart" (as in hurt) over the years in slippage. Interesting hafnium has experienced no slippage on GX versus OV Open/Close values - I always expect slippage of some amount - sooner or later. But as Jim is trying to define, how much can be expected based on some basic factors, and then apply in Strategy modeling?
Steve, $11 slippage across 2,600 shares and 12 trades may be great, or really painful. Can you provide some total $$ or % perspective - what's the slippage impact against gains (loss) and total equity in these trades? That may even help lend comparison to real vs. strategy simulation output, maybe ;-)
In questioning Ed during one of the OV presentations, he said he has seen slippage 'average' out over time - sounds plausible, but is a big 'what if', especilly in aggressive, high trade volume RTM type strategies.
|
|
John W
Posts: 654
Joined: 10/11/2012
Location: Sydney, NSW, Australia
User Profile |
Question number 3 in the survey addresses the question of slippage by each broker against the published market opening price:
"3. Broker Fill Prices - Build a capability into OmniVest to record and report broker specific fill prices at market open"
John
|
|
Jim Dean
Posts: 1059
Joined: 10/11/2012
Location: L'ville, GA
User Profile |
Recording the info is only the start. The purpose of this is to create a viable, dynamic model that can be used for historical simulation. Since symbols vary widely as to volatility and liquidity, between symbols and over time, the model is IMHO useless as a "dumb average". It needs to be dynamic, a function of historically-measurable driving forces that logically relate to it.
Slippage is very important to some kinds of trading and some kinds of symbol lists, and less important to others. Real time and RTM and any other short-term trading is significantly affected by it.
Just saying "it averages out" is IMHO a way of sidestepping the issue. Not until careful studies are done using many thousands of data points could such a statement be reliably made. If such a study has been done it would be great if someone would publish it.
|
|
BrianD
Posts: 302
Joined: 2/23/2013
Location: Grand Rapids, MI
User Profile |
Jim: I could not agree more.
Based on experience, I'm very fixated on slippage when modeling Strategies.
The name 'slippage' is so appropriate - walk on ice, think all is well, until you fall and you bang your head a few times.
|
|
John W
Posts: 654
Joined: 10/11/2012
Location: Sydney, NSW, Australia
User Profile |
It appears from this thread and the other thread on Opening Gaps that we have identified two types of slippage.
The first type of slippage is simple and easy to measure. It’s the broker fill price versus the OV simulated fill price.
It can tell us how well an individual broker is able to get a ‘better’ price compared to his competing broker. It allows us to simulate our broker’s specific performance, and to complain or move our account if we are consistently not getting as good a fill price as the other brokers.
The second type of slippage appears complex. It is caused by many OV customers acting in concert at the same time on the same strategy.
This results in a higher buying price or a lower selling price. The larger the OV customer base and the larger its account size, the more this effect will be ‘felt’. It will be ‘felt’ in the hip pocket nerve because there will be bigger opening gaps from one period to the next, therefore reducing returns.
So how to ‘see’ what is going on, and what to do about it?
Nirvana is in the best possible position to ‘see’ what is going on; there is plenty of information available from OV and the TP that can be garnered into meaningful statistics.
There is pre-OV return data and post-OV return data, and as has been suggested there are gap sizes, ATR’s and EMA’s. There is also account size data and $ and position sizes placed on different strategies and symbols. A set of summary statistics could be useful for all.
What to do about it in trading? I echo the comments on limit orders or delayed entry. Perhaps an intelligent release of orders into the market when buy/sell ratios are suitable could be another approach.
I’m looking to Nirvana for guidance on this topic; there are lots of good ideas in this thread and related threads but this topic goes to the design of how OV is simulated and also how the TP places trades into the market and I believe it falls into Nirvana’s area of expertise and responsibility.
I’d like to see Nirvana take leadership on this issue, perhaps suggesting a way forward.
John
|
|
Steve2
Posts: 750
Joined: 10/11/2012
Location: Annapolis, MD
User Profile |
Brian, sorry I missed your question above. With only 12 trades, it's way too early to assess slippage impact. But to answer your question it's an $11 slippage against a profit of $1,383. Currently, slippage has about the same impact as commissions but, again, way to early to draw any conclusions. Stay tuned...
|
|
Jim Dean
Posts: 1059
Joined: 10/11/2012
Location: L'ville, GA
User Profile |
Regarding John's "second kind of slippage" - | The second type of slippage appears complex. It is caused by many OV customers acting in concert at the same time on the same strategy. |
Two points:
1. This is essentially a "macro" form of regular trading slippage - think of all the nearly-simultaneous orders as if they were "one big trade". If that "big trade" is a significant fraction of the daily volume for that stock, and if prices on that stock tend to be volatile, then the "big trade" slippage effect could well be a "scary" magnitude kind of number.
2. To help us proactively avoid being hit by that effect, it would be very helpful if OmniVest could report in an analysis column what the overall composite magnitude of participation there is amongst ALL OVest customers in a particular strategy used on a particular symbol. This report could be something simple like small/med/large, or something like a normalized percentage of max possible users participating in that combo, or (best) a percentage that specifically indicates the fraction of the average daily volume that the composite total "big trade" number of shares represents. If that percentage is high, we as users can opt NOT to participate in that trade. In fact, OVest could give us a money-management input option to either ignore or to scale back participation in a given trade as a function of what the big-trade percent of daily volume might be. This would be COOL. And powerful!
I hope I've explained this well enough.
|
|
Juan
Posts: 74
Joined: 10/11/2012
Location: Round Rock, Tx
User Profile |
This thread has zeroed in on what I think is the "Achilles' heel" of OV. Say 500 OV users go in regularly on the same symbols over and over again. Most certainly this will reduce returns especially for lower cap stocks.
Here's a solution that I think would work to reduce this. Since OV and TP are supposed to work as an automated solution, why do we have to be limited to buying and selling at the 8:30am CST OPEN? Why not recreate each strategy to run and execute trades at say 8:30am market open (current default), 9:30am, 10:30am, 12noon, 1:30pm, 2:30pm, and 3pm CST?
The idea is really quite simple. For each strategy, run historical equity curves based on buy/sell signals at "after market hours", 9am, 10am, 11:30am, 1pm, 2pm, 2:30pm CST based on real-time data and then execute these 30 minutes later at market open 8:30am (current default), 9:30am, 10:30am, 12noon, 1:30pm, 2:30pm, and 3pm (market close).
For example, pretend at exactly 2pm is when the strategies are run and trades are determined, then execute these trades at 2:30pm CST. Take all the strategies and split them across these designated time frames and build the historical equity curves based on this. Then we can select these and reduce the impact of say 500 users buying/selling the same symbol at 8:30am CST.
For example, the strategy R18-B-SP, Nirvana would build seven new strategies with the following names: R18-B-SP(keep orig name which runs at 8:30am), R18-B-SP_0930AM, R18-B-SP_1030AM, R18-B-SP_12PM, R18-B-SP_130PM, R18-B-SP_230PM, and R18-B-SP_3PM.
Even though 8:30am or 3pm will show the highest gains, I would surely select strategies that executed at other time intervals simply to reduce the risk and slippage. It may be difficult to build equity curves back to 2000 but surely we can go back to 2007. I know several data vendors have this. Also, the historical equity curves at these time different time intervals are an absolute must to buy into these strategies.
Since OV and TP are supposed to be a fully automated trading solution that runs unattended, executing trades during regular market hours should not be a stretch. And if OV/TP is really moving to real-time trading, then it's a must to have.
I think this would be easiest to implement and reduce the open gap issue that will certainly reduce profit over time if everyone is making the same trades in concert.
Lastly, I've done much analysis on NSP-33, NSP-41, and the T3 suite and noticed that returns are typically higher when you buy/sell at the close (3pm CST). The problem with executing this, of course, is if you have a regular job, you're not able to do this during business hours. Thus, you're stuck running this overnight. But the TP/OV solution can overcome this and much more with the multiple timeframe solution I just described above.
IMHO, I really believe this solution makes sense and hopefully the Nirvana Team will bring this into consideration as they progress into OV/TP real time trading.
|
|
Jim Dean
Posts: 1059
Joined: 10/11/2012
Location: L'ville, GA
User Profile |
Unfortunately there is a problem with the different-times-for-different-strategies approach. The backtesting is not set up that way - it's done on daily data and there is inadequate granularity in the EOD historical data to model specific times of day.
However - some strategies are not as sensitive to specific entry and exit times as others might be - generally speaking, strats with short hold times are more sensitive than those with long hold times.
So - if Nirvana wants to spread the processing and submission of the orders across the day, they could do so with an arbitrary rule of some sort, based on the historically typical trade duration - strat's with short hold times (like RTM) would fire in the first hour or so, and those with the longest historical hold times (like longterm trending) would fire later in the day.
The impact of doing this would be a guess - it could not be back tested.
Finally - reality check - for now and probably for some time to come, as long as good LIQUIDITY CHECKING is done (per my suggestion two posts ago), this concern can be minimized. Sure, OVest will be like a fund manager, but at first it will be a pretty small fund. Eventually the number of participants will grow large enough for this to be a bigger issue, but by that time Nirvana should have plenty of operating income from OVest to warrant more exhaustive solutions to this.
I really hope that the suggestion I made in item #2, two posts ago, will be seriously considered - that is, teach OVest to "monitor itself" regarding the overall number of shares that it is submitting (for all users that selected that strat, factored by percentages etc) for a particular symbol entry or exit on a particular day, as a percent of that symbol's average daily volume. If we could see that stat in an OVest, and have a Money Management user input to block or scale back participation when that percentage is too high, I believe that would provide a practical and relatively easy to code solution.
That is - TELL us what the big-composite-trade-excess-slippage danger is on a case by case basis, and LET US DECIDE (manually and via automated MMgmt control) what we want to do about it.
|
|
Jim Thorpe
Posts: 36
Joined: 10/11/2012
User Profile |
Wouldn't the use of limit orders help with this? The limit orders could be assigned as a percentage of the ATR perhaps (percentage allowed assigned by the user). This way we wouldn't be filled if the price creeps up too much at the open. But if the price pulled back again any time during the day, then we would be filled around our allowed price. So maybe I set up to assign limit orders at 20% of the ATR over the closing EOD price for example. If within that range any time during the day, I am filled.
|
|
Jim Dean
Posts: 1059
Joined: 10/11/2012
Location: L'ville, GA
User Profile |
Limit orders can help - and we need them. That's an "easy to code" solution.
However, lest we see this as a panacea for the problem -
They won't singlehandedly solve the "composite big trade slippage" problem - what they will do is essentially prevent many participants from fully entering the trade in some cases, if the price takes off and never pulls back.
Also, for strategies that use limit orders, the historical simulation will be less legitimate - slippage will be reduced but the number of shares (and the number of trades) will not match what would have happened in real life.
And, they will create partial fills (unless FOK is used) which are harder for the TP+OVest to properly track and manage (or so it seems).
All that being said - the option to use Limit orders on a strategy by strategy basis is a very important need IMHO.
|
|
Jim Thorpe
Posts: 36
Joined: 10/11/2012
User Profile |
Thanks, that all makes sense. The limit orders would protect me (albeit, with the mentioned effects). But, isn't the end all solution to the problem at hand.
|
|
hafnium
Posts: 43
Joined: 10/11/2012
Location: Vienna, VA
User Profile |
In the name of full disclosure.
I did go back and verified that my GX trader fills (entries and exits) on MOO orders exactly matched the OV numbers for the past 4 months.
However - I just experienced two slippage events this morning. Out of 7 entries today, I did experience 2 events (vs OV)
DVA long OV - 117.63 GXT - 117.71
EXC long OV - 31.93 GXT - 31.96
So, with some egg on my face - I recall my previous statement. Perhaps I was just experiencing Early Adopter Credits.
LD
|
|
Jim Dean
Posts: 1059
Joined: 10/11/2012
Location: L'ville, GA
User Profile |
Thanks LD
You might want to check the numbers again tomorrow.
Maybe OV / TP is correcting their numbers, after the fact.
No, Jim. That is not happening. - Ed See my reply to this thread below.
|
|
Fred Gordon
Posts: 481
Joined: 10/11/2012
Location: Fayetteville, Ga
User Profile |
Jim,
An interesting concept for OV; to morph into a "Nirvana managed" investment portfolio. No conflicts, no system or software failures, no cloud concerns, no requirement for broker stop protections, no fear of being away from a monitor for extended periods, no need to worry about slippage. The only drawback for some would be the hands-off element.
|
|
Jim Dean
Posts: 1059
Joined: 10/11/2012
Location: L'ville, GA
User Profile |
Hi FG
I suppose that is a possible future branch-alternative for nirvana but of course they would need to be licensed to do so.
I'm not one of those people who would be comfortable with "hands off", but I would prefer a more integrated cloud environment, to reduce the number of connectivity-links and points of potential failure.
OmniVest offers a lot of potential for the future, no matter how you look at it.
I do hope, however, that Nirvana will soon be able to refocus on OT to get some bugs fixed. OT is where my main interest lies since it allows me to create my own strategies. It will be great when OT is stable and Elite Trader is active!
|
|
Steve2
Posts: 750
Joined: 10/11/2012
Location: Annapolis, MD
User Profile |
Jim,
Need some OT help. I created the custom indicators for the stats you want and added them to an OT focus list. That all works fine. However, I won't be available every day to record them so I need a way to occasionally display the indicator values for previous bars. I suppose I could create a chart template that plots them over time but I don't know of a way to display the actual values for a given bar. Is there a way to do this?
Steve
|
|
Jim Dean
Posts: 1059
Joined: 10/11/2012
Location: L'ville, GA
User Profile |
Hi Steve
Do it one of two ways:
1. Create a second set of the simple indic's that use [1], and add them as additional columns, such as
Return atr(14)[1]
… or …
2. Add a parameter to the indic's that allow you to change the lookback using the right-click edit > parameters option, ie:
#indicator
#param "DaysAgo", 0
Return atr(14)[DaysAgo]
That's better than plotting them, since you won't have to flip from chart to chart and mess with the cursor to get the value.
You can btw use SnagIt to capture the focus list including the custom column outputs as a text file - but that's another story entirely ;-)
|
|
Steve2
Posts: 750
Joined: 10/11/2012
Location: Annapolis, MD
User Profile |
Thanks Jim,
Attached is an example of the output. Let me know if it doesn't meet your needs. I'll post an updated spreadsheet about once a month. The portfolio I'm using should generate about 117 TPM. How long I do this will probably depend on how the portfolio does :)
--Steve
Attached file : Slip Stats 3-13-13.jpg (203KB - 223 downloads)
|
|
Ed Downs
Posts: 645
Joined: 2/7/2007
Location: Austin, Texas
User Profile |
To all in this thread....
I think this is an important issue, i.e., if a large number of users trade the same symbol at the Open, will that affect results?
A few thoughts that I'd like you to consider, and an answer on this...
1) The Machine by Larry Connors has the same approach as ours, multiple Strategies trading specific symbols. The difference is the # of Strategies, which are mostly based on small variations of parameters. With the addition of new Strategies (like the ones last week) the effect of the same users trading the same symbols will become less and less significant. (Note: Users have said our approach is better than Larry's - see the Forum on competitors.)
2) Our trades are a TINY percentage of the market. Even with about 400 users in OmniVest, unless some people are trading multi-million dollar accounts (which I doubt) we are very unlikely to affect the market at this point.
3) Market on Open is Market on Open. If you are using a broker that supports MOO orders, market makers resolve all the orders before market open, and a fair price is posted. It is THE Open, not the OmniVest Open.
4) We have the most liquid lists in OmniVest - no penny stocks. In my personal OmniVest account, I further limit selections using Account Settings, to 0.5% of daily volume. Doing this, my results have improved a great deal. I suggest users look carefully at their settings and limit trades to the most liquid stocks if they are concerned about "slippage".
Now, to answer your question...
I find it a little odd that this item did not make the Enhancements Survey I just published. The closest was John W.'s request that we REPORT on fill differences among the various brokers - but no items (that I can find) related to managing trades differently from OmniVest. On this topic, here are my ideas...
A) We can establish a Trade Processor condition that establishes Limit orders based on the Open. That is, wait until the Open passes and then establish a Limit price at the actual Open.
B) As Mr. Jim Dean suggests, we could provide a quantity of shares that WILL be traded in OmniVest, so users can establish a filter that says, "Only take this trade if OmniVest volume is less than X% of the prior market's volume." But that requires multiple passes across all users. It's a circular process - not easy at all to implement.
I see a lot of concern in this thread without a whole lot of empirical evidence, though I can see folks are trying to collect that, which is good. Everyone should know that IB is notorious for trading against customer accounts i.e., looking at demand and trading against the opening orders. So I'd be suspicious of IB if you are using them. We could expend the effort to generate GXTrader vs. IB Fills (per John W's request) and I think that would be instructive. Take a look at Brian D's post earlier in this thread, "hafnium has experienced no slippage on GX versus OV Open/Close values." All the "success interviews" were with GXTrader customers. Makes you think, doesn't it?
The question is what we focus on RIGHT NOW.
Look, you guys know me. I want this product to be as successful as possible for our members. What helps the most is constructive suggestions on specific things we actually can do in a reasonable amount of time. Extremely complex enhancements that require weeks to program don't help the mission. I mean it takes a user a few minutes to post a complex analytical idea that takes weeks or months to implement. I have asked before, "Do you want us to spend 3 weeks implementing something complex, or do you want us to add more ARM4 Strategies?"
I'm putting out another survey next week for the 2nd tier issues. I suggest you guys formulate a specific enhancement regarding Broker Fills that actually addresses the issue, and submit it in the Enhancements Forum so we can get the user population to vote on it. I have to move OmniVest forward in this manner - i.e., the things that benefit the greatest number of users. I'm sure everyone is aware we could NEVER implement all the ideas that some users have posted. Would love to do it, but resource constraints are what they are.
|
|
Jim Dean
Posts: 1059
Joined: 10/11/2012
Location: L'ville, GA
User Profile |
Hi Steve:
That format looks great!
|
|
Juan
Posts: 74
Joined: 10/11/2012
Location: Round Rock, Tx
User Profile |
Ed, I thought you're supposed to be on spring break. Thanks for taking the time off your vacation to converse on this very interesting topic. I regret that I said 'This thread has zeroed in on what I think is the "Achilles' heel" of OV'. Truth is the equity curves are so incredible that I find them hard to believe...and I feel that this issue will dampen the curve somewhat. But I feel that even so, this system will still generate very generous returns.
The idea I proposed about multiple strategies based on real-time set points ("execute trades at say 8:30am market open (current default), 9:30am, 10:30am, 12noon, 1:30pm, 2:30pm, and 3pm CST") is something that I'm hoping the Nirvana team can consider sometime down the road. Certainly not right away since it would be quite a development effort. I do believe it has merit and will be beneficial for real-time trading. I also believe it's very doable but only in the long run.
For slippage and risk controls, I'm all for this but I have to be able back test this and confirm that any new risk controls won't dramatically lower the equity curve. I hope that as your team decides to add risk controls that we're able to run historical simulations.
I've noticed in my trading, over the years, that when I put in too many risk controls my returns go way down...like the old saying no risk no reward. I'm all four risk controls but need to confirm that they don't damage the equity curve historically.
Once again thanks for the great product and I hope you're having a great vacation.
|
|
Ed Downs
Posts: 645
Joined: 2/7/2007
Location: Austin, Texas
User Profile |
Juan,
Yes, I'm on Spring Break with my family, but this is such an important time I feel staying plugged-in is important. I thought your post was fine. You must be the person in the Survey who wrote in about adding too many risk controls. Well said. If you remove all the risk you remove all the returns. Absolutely!
I have been pondering the potential effects of large opening orders, and think I have a solution which is similar to what you are saying. The fact that we have many Trade Processor running at the same time gives us a unique advantage.
What if the Trade Processors coordinated their submissions so they don't all go in at the Open? Like 8:31a, 8:33a, 8:35a, 8;37a Central Time (and so on). They could be randomly assigned a time slot and then go with that. We can't do it tomorrow, but we could create a new order type ("Market in Sequence") that would do this. The effect would be a slower metering of orders into the market, which is what Buy-Side traders do.
Anyway, just a thought I wanted to throw out there.. Thanks.
|
|
gbarber
Posts: 282
Joined: 12/30/2012
Location: Pearland, TX
User Profile |
I don't claim much experience with this but it seems to me that your earlier post Ed, that doubted Omnivest traders were causing the gaps was right (unless you have very many users of Omnivest, I have no idea). I have my account settings set to filter out any trade with a volume less than 500,000 and trade size must be less that .1% of trading volume. But I still suffer pretty badly from gaps up followed by a retrace. I am thinking that the systems in use have some commonality with many systems used by many other people. It seems to me that some automation to avoid trading into a gap would be the best solution. A limit order seems to be the simplest way to do that. But you guys know this stuff much better than I do. I know something isn't working right. I have been using a set of strategies with excellent stats and outstanding recent performance (from 1 Jan to now) but I have been losing since I started trading with it about two weeks ago.
|
|
BrianD
Posts: 302
Joined: 2/23/2013
Location: Grand Rapids, MI
User Profile |
Gary: What broker are you using?
I concur with Ed's above comments about IB from experience.
|
|
gbarber
Posts: 282
Joined: 12/30/2012
Location: Pearland, TX
User Profile |
I am using GxTrader.
|
|
John W
Posts: 654
Joined: 10/11/2012
Location: Sydney, NSW, Australia
User Profile |
Originally written by edowns on 3/14/2013 12:55 PM
2) Our trades are a TINY percentage of the market. Even with about 400 users in OmniVest, unless some people are trading multi-million dollar accounts (which I doubt) we are very unlikely to affect the market at this point....
I see a lot of concern in this thread without a whole lot of empirical evidence...
Ed, I think there is empirical evidence that if OV trades its position all at once that it is likely to affect the price of smaller traded symbols and as OV grows this is likely to become more pronounced. I had a stab at this on Feb 8 Post #23570 http://www.omnitrader.com/currentclients/omnivestforum/thread-view.asp?threadid=4277) and also on Jan 24 Post 23195 http://www.omnitrader.com/currentclients/omnivestforum/thread-view.asp?threadid=4243).
If my numbers below are wrong re the average OV account size then please adjust my attempt at quantifying this so that the empirical data is better.
On 8 February I wrote:
"In another thread on January 24 I mentioned that in the Russell 1000 I ran an OmniScan to find out how many stocks on average traded less than 1000K, 2000K and 3000K per day [e.g. Avg(C,14)*Avg(V,14) < 1000000]. I’ve run that again today and there were 44 in the less than 1000K category, 156 in the less than 2000K category, and 258 in the less than 3000K category.
Using my belief that 3% is the maximum to trade before the price starts to move then 3% of these 3 categories is 30K, 60K and 90K respectively.
So let’s take the higher example, the 3000K/day traded stock.
At the moment there are 400+ members in this forum. Let’s imagine a really popular strategy is followed by 50% or 200 OV users. Let’s imagine the average account size is 50K, and that 10% or 5K is the normal trade size. So at a particular time 200 OV users place 5K on stock ABC at market open (or the next bar open on a real time strategy), and sell it another few days (or bars) later.
This is more than 3% of the daily traded volume to be placed by a group of OV users acting as one at one point in time. My belief is that this will cause a gap in the price for the 258 stocks in the less than 3000K category, bigger gaps in the 156 stocks in the subset trading less than 2000K and even bigger gaps in the smallest 45.
My other belief is there will be the fall off in performance of the particular OV strategy in play in the smaller symbols lists because now the group of OV users trading this strategy as one will experience a gap up when they buy and a gap down when they sell, lessening returns over time.
It’s not hard to imagine that some people trading OV may put on larger trades than the $5K example, that the number of OV users could exceed 1000 or even 10,000 (one competitor claims more than 65,000 members!), and even hedge funds and other big players may be attracted to OV.
The 100K traded on single stock could grow massively which will cause many of us to re-evaluate our returns on smaller symbol lists.
If the number of users using OV jumped to 1000 then the cut off in my example is 6500K and only the top 500 stocks have the possibility they will avoid some form of slippage caused by OV users acting together. At 10000 users only the top 62 stocks escape the possibility of OV induced slippage if we all act together as one!
In summary I'm saying that it may be a waste of time building a slippage model for a single account. There won’t be observable slippage caused by that single account, just gaps and future performance issues in smaller lists in particular.
The way that I’m hoping Nirvana plans to combat this is to consider adding many more strategies and introduce other data sets such as futures and options, even FX and different time frames, even other country lists, limit orders and other Elite strategies offered by others to keep growing the universe of available alternatives so that as the number of members in OV grows the risk of opening gaps is kept manageable so as not to damage the future performance of OV.
Nirvana could also introduce random small purchases throughout the day in the same manner as the big players in the market, and also could limit membership or the number of strategies to be traded.
Ed has already mentioned some of these alternatives so I’m firmly in the camp that the guys in the control room at Nirvana have a path mapped out, and over time the building blocks will come together."
John
|
|
Jim Dean
Posts: 1059
Joined: 10/11/2012
Location: L'ville, GA
User Profile |
I agree with John's rationale re the eventual significance of impact, but I also believe that Nirvana will find a way to help deal with that, as the numbers of shares traded en-masse increases. I think that the volume filtering capability that is already present is a good start.
I disagree however regarding the value of slippage modelling. I think that slippage, for whatever cause, is a relevant and potentially important aspect of trading - regardless of whether it's individual or via OmniVest. There is nothing that Nirvana can do to eliminate this inherent aspect of trading. Bid and Ask will always be different. If not there would be no market!
The REASON that I believe it's important to model slippage in the historical simulations is to make those simulations more true to life. I strongly believe that SOME kinds of strategies are much more susceptible to slippage, while others are less sensitive (easy to prove if anyone disagrees). The core concept of OmniVest is to mix and match strategies to get consistent gains with low drawdowns and wise usage of capital. We can only do that intelligently if the equity curves and statistics that are the building blocks of our decisions are as true to life as possible.
Since slippage is not an "even Steven across the board per share" kind of impact (like commissions or margin costs) - since it does impact some strats more and some less - and some symbols more and some less - I think it's critical to include an telligent parametric model of it for the simulations.
The important thing is not for the slippage amount to be absolutely correct - but rather that it realistically represents how different strats and symbols would react differently to it.
And once again, I'm offering to provide working code to Nirvana, at no charge, to implement a healthy slippage model, so that it won't steal too much time from other development endeavors. The model is such that it can be turned off by users who don't want it, and adjusted/scaled by users with differing "beliefs" about its magnitude.
Thanks for listening.
|
|
John W
Posts: 654
Joined: 10/11/2012
Location: Sydney, NSW, Australia
User Profile |
Unfortunately my quote from 8 Feb included the line "In summary I'm saying that it may be a waste of time building a slippage model for a single account. There won’t be observable slippage caused by that single account, just gaps and future performance issues in smaller lists in particular."
I agree with Jim's rationale about slippage too. It's just that I think it’s going to be felt at the OV level by all users; it’s not just an individual account issue. Even if there was one single large account using OV and inducing slippage, that slippage would be felt by all. It’s a moot point one way or the other, let’s just accept slippage will occur and Jim has proposed a solution.
Jim has a number of times offered his own work to Nirvana, not only on the issue of slippage but also risk management and position sizing and I'm hoping that Nirvana will take him up on his offer(s), but I've seen nothing yet to indicate that may happen. This is overdue for response.
I've proposed to Ed a simple empirical formula above that shows if OV trades its position all at once then it is likely to affect the price of smaller traded symbols and as OV grows this is likely to become more pronounced.
I’m genuinely hoping that Nirvana will take this input (and Jim’s) on board in their system design.
John
|
|
Steve2
Posts: 750
Joined: 10/11/2012
Location: Annapolis, MD
User Profile |
Regarding gaps, OV users are but one set of users that might have a significant impact on a stock at market open. Presumably there are many other sets of players that for one reason or another might generate enough opening volume to influence price. While we can determine if OV generated opening volume is significant, we can't do that for other players. Perhaps we need to think about doing historical analyses of gaps on a stock by stock basis. If the frequency and size of gaps are increasing above historical norms then that might be cause for filtering out a stock whether or not OV is generating significant opening volume.
I'm also in favor of slippage modelling.
|
|
Jim Dean
Posts: 1059
Joined: 10/11/2012
Location: L'ville, GA
User Profile |
Gap filtering has long been a part of a sophisticated "smooth movers" culling process that I've developed over the years. It's one of ten generic filter considerations. If/when Nirvana decides to implement that, I'd be willing to help with code for several alternative approaches. It's an interesting and complex field of study.
|
|
gbarber
Posts: 282
Joined: 12/30/2012
Location: Pearland, TX
User Profile |
It is really great to hear from all the erudite and vastly experienced people participating in this discussion. A lot of wisdom has been laid on the table. I am hoping Nirvana will soon state what will be done to OV to handle the opening gap situation discussed here. Trading into an opening gap raises risk in a manner not contemplated at the time of ordering the trade. This is similar to trading into a dividend or earnings report (I think the answer here is generally, don't order the trade). All of these need to be accounted for and handled in some way when using an automated system. Since I have been hurt by opening gaps, IMHO this should be high on the priority list and we will hear soon how Nirvana will handle this.
|
|
Jim Dean
Posts: 1059
Joined: 10/11/2012
Location: L'ville, GA
User Profile |
I think the idea is this:
TP users want the TP to be generally as smart as they are, in terms of obvious general rules that the "average guy" would agree with.
... or to put in another way ...
TP users are unhappy if the TP acts in an obviously dumb manner that costs them money which they definitely would not have done manually.
Big opening gaps are an obvious example of this ... but there are probably other such "aww, come ON, Mr.TP, you shuda known better than t'do THAT!" kinds of things that will be noted as time passes.
So, I hope Nirvana will create a "functional basket" in the TP for rules like this that the User can activate &/or tune in the TP settings.
|
|
Mark Holstius
Posts: 744
Joined: 10/11/2012
Location: Sleepy Hollow, IL
User Profile |
Good idea Jim...
Something like the current "Account Settings" page with rules / parameters that could be checked and adjusted as desired;
"Allowable Gap +/- %"
Different order types and limits
Etc
Not a programmer, but it seems that if the "basket" is included in this early beta stage it would simplify the process of adding improvements as they become apparent later.
Mark
|
|
kmcintyre
Posts: 890
Joined: 10/11/2012
Location: Portland, OR
User Profile |
I was ignoring this thread because I thought it was IB centric...
Ed wants simple solutions that can be implemented in a reasonable amount of time.
I'm with Jim on wanting the TP to be "intelligent" enough to not do "stupid" things, and to trade as I would. (only better...)
Overnight gaps, slippage, unscrupulous brokers, sundry market conditions beyond my control. To me, it all comes down to this - can I buy the stock at the entry target my strategy made it's recommendation based on, or not?
If I can't get the stock at the target entry, I want TP to forget the trade and move on to a more favorable trade.
OV can generate MANY more trade opportunities than I can afford to take. OV doesn't know what market conditions will be tomorrow, and hence can't really make intelligent decisions about what trades should be taken.
Trade selection needs to be a "late binding" affair (borrowing from computer science vernacular). TP should have a long list of potential trades, look at the market (bid prices), and take the most favorable entries. IMO that means limit fill-or-kill orders taken after the market opens.
What does late-binding buy? Too many people wanting the same stock? TP will look for a better deal. (So the demand is spread out by TP real-time.) Overnight gaps? TP will pass on the trade since the stock can't be bought at the target entry price. Internet outages? Broker not available? TP will look at the market once service is restored and decide if any favorable entries exist.
Add MOC exits and TP would be a KILLER product for EOD traders.
(And process closing orders prior to new entries to free up buying power.)
I love to call it Intelligent Order Processing - (IOP). And if Nirvana doesn't do it, I might write it myself. :-)
Simple, elegant, implementable, applicable to real-time trading...
Cheers
Keith
|
|
Mark Holstius
Posts: 744
Joined: 10/11/2012
Location: Sleepy Hollow, IL
User Profile |
Excellent synopsis Keith
Mark
|
|
Steve2
Posts: 750
Joined: 10/11/2012
Location: Annapolis, MD
User Profile |
Attached is IB slippage data from my first month of live trading (actually less than a month since I started on Mar 6th). The Summary worksheet shows average slippage by order size and order type. The Trade Data worksheet shows the details for each trade including volatility, volume, opening gap, and daily range data, as well as the exchange that was used. I'm currently submitting market orders before market open (although there is one MOO order in there). At some point I will switch over to MOO orders for awhile and then begin to explore limit orders. Will post updates at the end of each month.
I've made only 60 trades so far so way too early to draw conclusions. Currently, average slippage is $.005 per share or just about equal to commission cost.
Attached file : Slippage History - IB Live 1 Thru 13-03.xlsx (21KB - 235 downloads)
|
|
Fred Gordon
Posts: 481
Joined: 10/11/2012
Location: Fayetteville, Ga
User Profile |
Great format for data history. Since, for now, your trades are mostly @Market, have you thought about checking slippage without rounding the # of shares for each position? Not sure it would make a difference but the trades may be more quickly placed by the brokerage thereby transacting closer to the managing strat's price.
|
|
Steve2
Posts: 750
Joined: 10/11/2012
Location: Annapolis, MD
User Profile |
FG, I am planning to do that at some point since using round lots reduces P/L in the simulations. I actually started with round lots because I thought that might help reduce slippage. I have the execution times for the orders. They were just not included in the spreadsheet. The vast majority of the orders completed execution within 1 sec of market open so I'm not sure things can get much faster than that.
Steve
|
|
BrianD
Posts: 302
Joined: 2/23/2013
Location: Grand Rapids, MI
User Profile |
Impressive work. Thanks for sharing.
|
|
Fred Gordon
Posts: 481
Joined: 10/11/2012
Location: Fayetteville, Ga
User Profile |
Probably right. That is probably as quick as we can expect, round lots or not, MOO or not.
|
|
Steve2
Posts: 750
Joined: 10/11/2012
Location: Annapolis, MD
User Profile |
Attached is IB slippage data from my live trading account (Mar 6th through the end of Apr). The Summary worksheet shows average slippage by order size and order type. The Trade Data worksheet shows the details for each trade including volatility, volume, opening gap, and daily range data, as well as the exchange that was used. I'm currently submitting market orders before market open (although there is one MOO order in there).
Slippage for the month of April was positive and overall slippage is now slightly positive (plus $5 since inception). 153 trades have been executed so it's still early but so far there appears to be little correlation with order size.
Starting in May, I am changing opening orders from market orders to day limit orders at the previous trading day's closing price. My analysis of the first two months of trading show this would significantly increase portfolio returns. So we shall see. Will report back in a month.
Steve
Attached file : Slippage History - IB Live 1 Thru 13-04.xlsx (30KB - 215 downloads)
|
|
Steve2
Posts: 750
Joined: 10/11/2012
Location: Annapolis, MD
User Profile |
Attached is IB slippage data from my live trading account (Mar 6th through the end of May). The Summary worksheet shows average slippage by order size and order type. The Trade Data worksheet shows the details for each trade including volatility, volume, opening gap, and daily range data, as well as the exchange that was used. For the first two months I submitted market orders before market open for both opening and closing trades. In May, I submitted day limit orders at the previous trading day's close for all opening orders and continued to submit market orders for closing trades.
Slippage for market orders submitted before market open remained slightly positive at $0.001034 per share. Slippage for the limit opening orders was more positive at $0.042941 per share. This resulted in a positive gain of $322 during the month, however, to assess the true impact one has to factor in the P/L of trades that did not fill. During the month 10 opening trades did not fill and 9 of those have been closed out by OV. The impact of those 9 trades was a profit of $496 that was not realized. So, the net impact of using opening limit orders was $322 - $496 = ($174). Since, I believe the use of opening limit orders still has promise, I'll continue to use them for another month and report back at the end of June.
Steve
Attached file : Slippage History - IB Live 1 Thru 13-05.xlsx (39KB - 264 downloads)
|
|
Fred Gordon
Posts: 481
Joined: 10/11/2012
Location: Fayetteville, Ga
User Profile |
Steve, your slippage data is very informative. Thank you,
Fred Gordon
|
|
Mark Holstius
Posts: 744
Joined: 10/11/2012
Location: Sleepy Hollow, IL
User Profile |
Thanks for all your work Steve - much appreciated!
Mark
|
|
BrianD
Posts: 302
Joined: 2/23/2013
Location: Grand Rapids, MI
User Profile |
Once again, very valuable and informative info. It is great to have accumulated 90 days worth of live experience to review.
Thanks again for the detailed info and time spent.
|
|
Steve2
Posts: 750
Joined: 10/11/2012
Location: Annapolis, MD
User Profile |
Attached is IB slippage data from my live trading account (Mar 6th through the end of June). The Summary worksheet shows average slippage by order size and order type. The Trade Data worksheet shows the details for each trade including volatility, volume, opening gap, and daily range data, as well as the exchange that was used. For the first two months I submitted market orders before market open for both opening and closing trades. In May and June, I submitted day limit orders at the previous trading day's close for all opening orders and continued to submit market orders for closing trades.
Slippage for market orders submitted before market open turned slightly negative at $(0.000352) per share. Slippage for the limit opening orders remained positive at $0.069690 per share. This resulted in a positive gain of $1,248 during the past two months, however, to assess the true impact one has to factor in the P/L of trades that did not fill. During the past two months 18 opening trades did not fill and have been closed out by OV. The impact of those 18 trades was a profit of $1,334 that was not realized. So, the net impact of using opening limit orders was $1,248 - $1,334 = $(86). Since, this is an improvement over just the May results, I'll continue to use them for another month and report back at the end of July. Also, I'm adding funds to the account (normally the kiss of death), so I'll be collecting data on larger order sizes.
Steve
p.s.
If anyone from Nirvana happens to read this, I was wondering if you all simulated the use of limit orders at the previous close when designing OV. If you have done this and the simulation showed it does not improve results then please let me know and I can abandon this approach... thanks
Attached file : Slippage History - IB Live 1 Thru 13-06.xlsx (51KB - 243 downloads)
|
|
Fred Gordon
Posts: 481
Joined: 10/11/2012
Location: Fayetteville, Ga
User Profile |
Steve, super info; thank you.
|
|
Jim Dean
Posts: 1059
Joined: 10/11/2012
Location: L'ville, GA
User Profile |
Hi Steve
Thanks very much for doing this. I wish Nirvana would do this - drawing info anonymously from all TP transactions and filling in the table. We really need a huge amount of data to draw useful conclusions.
You mentioned that you are about to "add funds to the account". I thot that your prior trades that you were reporting on were from a funded account. Was I incorrect? Was it a Paper account?
I would NOT expect paper accounts to show any kind of meaningful slippage since the orders aren't actually being fed to the exchanges.
Or, if the orders are small ones (such as you see on the far left of the equity curve), I would not expect the slippage to be meaningful unless VERY illiquid stocks were being traded.
The gotcha with slippage is when trade sizes become an appreciable fraction (say, 0.1% or more) of the avg daily vol for that stock. That's what the middle and right of the 13 year equity curves are generally dealing with. That's what IMHO needs automatically estimated slippage modelling in OVest and PortSim - to help make those wild and wonderful equity curves a bit more realistic.
So - if you truly want to be a martyr for the cause and throw some real money at a portfolio that has significantly less-liquid stocks BUT does show a pretty nice OVest curve and decent strats, let me know and I'll pass on the portfolio and allocation info. It's using the strats that are all-long, based on MG groups of symbols for market-rotation type trading. I found in my simulation R&D that the curves were extremely sensitive to ANY filtering of the volume - which tells me that the trade sizes got to be appreciable fractions of the daily vol for those stocks.
I'd try it myself but I just don't have adequate funds to do meaningful testing of it at this time.
|
|
Steve2
Posts: 750
Joined: 10/11/2012
Location: Annapolis, MD
User Profile |
Jim,
The data posted here is from a live account. Since the strategies are performing as expected, I'm just adding more funds. This will push the average order size up to about 500 shares. I agree completely that slippage is an issue when trade size becomes an appreciable fraction of daily volume. That's why I'm only trading stocks with average daily volumes greater than 1M shares ( and not interested in trading illiquid stocks, sorry). I'm really trying to see if there are correlations with other things e.g., order type, exchange used, order submission time, etc. I was also hoping others would post slippage data for the other brokers so we could compare but so far no takers.
I too wish Nirvana would publish daily aggregate trading volumes for each symbol so we could track how much volume OV drives or might theoretically drive since I don't know how much live trading is actually being done.
An interesting observation about paper accounts. I've been running TP against a paper account for quite awhile submitting market orders before market open (not MOO). While I haven't been formally tracking slippage, it appears that the fills are worse in this account than in my live account. So, one of the things on my list is to figure out whether this is due to the fact that it's a paper account or an artifact of the order parameters TP is using.
Steve
|
|
Mark Holstius
Posts: 744
Joined: 10/11/2012
Location: Sleepy Hollow, IL
User Profile |
Thanks for all your work Steve... great info.
Mark
|
|
BrianD
Posts: 302
Joined: 2/23/2013
Location: Grand Rapids, MI
User Profile |
Steve: I find IB demo accounts produce slippage quite often. Fortunately IB appears interested in good real time trades. That's why it has been comforting to see your actual results over the last 4 months.
|
|
Steve2
Posts: 750
Joined: 10/11/2012
Location: Annapolis, MD
User Profile |
Attached is IB slippage data from my live trading account (Mar 6th through the end of July). The Summary worksheet shows average slippage by order size and order type. The Trade Data worksheet shows the details for each trade including volatility, volume, opening gap, and daily range data, as well as the exchange that was used. For the first two months I submitted market orders before market open for both opening and closing trades. In May, June, and early July, I submitted day limit orders at the previous trading day's close for all opening orders and continued to submit market orders for closing trades. On July 8th, I switched back to submitting opening orders as market orders. Nirvana has indicated that they will add opening limit orders to OV, so I'll just wait for that to happen so their effectiveness can be properly back tested.
Slippage for market orders submitted before market open is once again slightly positive at $0.0043 per share. Slippage for the limit opening orders remained positive at $0.062678 per share. This resulted in a positive gain of $1,213 during the past two+ months, however, to assess the true impact one has to factor in the P/L of trades that did not fill. During the past two+ months 21 opening trades did not fill and have been closed out by OV. The impact of those 21 trades was a profit of $2,115 that was not realized. So the actual impact of submitting opening limit orders was $1,213 - $2,115 = $(902).
Note: I'm not sure how much longer I will continue to publish this data. The data collection is a pain and I've accomplished my primary goal which was to assess the impact of IB slippage on P/L. My assessment is that IB slippage is a non-issue for order sizes that I expect to be submitting over the next several years.
Steve
Attached file : Slippage History - IB Live 1 Thru 13-07.xlsx (60KB - 208 downloads)
|
|
gbarber
Posts: 282
Joined: 12/30/2012
Location: Pearland, TX
User Profile |
Steve
Thank you very much for sharing this very valuable data with us.
|
|
Jim Dean
Posts: 1059
Joined: 10/11/2012
Location: L'ville, GA
User Profile |
Thanks Steve
If anyone has data re less liquid stocks (Steve is using > 1mill/day) please let us know. Nirvana if you are testing with iB then hopefully you can tell us something.
|
|
Steve Mayo
Posts: 414
Joined: 10/11/2012
Location: Austin, TX
User Profile |
Great work Steve. Can you publish one more metric from your work, namely how did your actual return compare against the simulated return in OV over that time period?
I find, running fully automated, that a number of trades never get filled and my results are less than what OV shows. I've switched portfolios lately so I don't yet have a good metric I can report...so interested to know what you are finding.
Thanks!
Steve Mayo
|
|
Bob123
Posts: 121
Joined: 10/11/2012
Location: North Andover, MA
User Profile |
Steve2 - tremendous work and much appreciated considering the time you put into this for the benefit of all OV subscribers.
|
|
BrianD
Posts: 302
Joined: 2/23/2013
Location: Grand Rapids, MI
User Profile |
I have been running straight Market orders for a few weeks. I have not been as diligent as Steve in recording so I do not have a good spreadsheet to share, but slippage has been in my favor so far.
My TPM is over 100 on this account.
|
|
Steve2
Posts: 750
Joined: 10/11/2012
Location: Annapolis, MD
User Profile |
Steve,
Unfortunately, the new interface changes have broken the ability to get all of the OV Historical Positions. I'll get you the metric once that's working again.
Steve
|
|
Steve2
Posts: 750
Joined: 10/11/2012
Location: Annapolis, MD
User Profile |
Steve Mayo,
I can answer your question now.
My actual return from Mar 6th through Aug 7th was 19.36%. The OV Trade History for the same period showed a return of 16.80%. The difference was primarily caused by two things. On the positive side, I stopped out 4 trades for significantly higher gains than OV achieved. I have my account settings setup in a strange way. I round up to the nearest 100 shares and I do not have a limit set on the percentage of my account value that can be allocated to a single trade. This means that when OV selects a high-priced stock (e.g., GOOG, AAPL) it can chew up a high percentage of my account value. I was very nervous about doing this but the simulation showed it resulted in significantly higher gains. So, for the first few months of trading I used stops to try and preserve gains and protect against large loses for any trade that was greater than 40% of my account value. I've since relaxed a bit and no longer bother to do that but it did prove to be a successful approach early on. On the negative side, I experimented with placing opening day limit orders at the previous trading day's closing price. While this resulted in positive slippage against OV's entry points, it resulted in some no fills that actually reduced my returns over what OV was achieving so I've discontinued that.
I am not using TP on my live account and have not experienced any issues with trades not being filled (other than when using limit orders). Out of the 218 opening trades I've submitted (excluding the limit orders that didn't fill), I've had one partial fill and zero no fills. I always submit market orders before market open rather than MOO orders.
I have been running TP continuously for the past 80 days on a paper account and have not experienced any no fills but I'm not sure that really means anything.
Which broker are you using?
Steve
|
|
Steve Mayo
Posts: 414
Joined: 10/11/2012
Location: Austin, TX
User Profile |
Thanks Steve. Very helpful info.
I'm running live on GX, virtually unmanaged, on an AWS cloud server. I started with an account running about 30 strats back in March. It had good returns, admittedly much better before the commission increase, but even after margin costs and the platform fee, making dozens of trades a day. But it missed a lot of trades for various reasons...and I don't even want to think about doing my tax return now. :-(
In the end, however, the return, although significantly beating the S&P, was only about half of what OV showed over the same period. So, I switched to an account with fewer strats. It still misses the ocassional trade or two (not shortable, couldn't get a price, exceeding overnight limit, data hickups, etc.) but has tracked OV much better over the last month. Unfortunately, the return hasn't been as good (albeit, its hard to compare short term returns in different market conditions). This one seems to bear much more market risk, meaning it more closely tracks the S&P500. I think that is probably because it uses more of the trend-following strats than my prior portfolio which was mostly just quick RTM trades.
|
|
kmcintyre
Posts: 890
Joined: 10/11/2012
Location: Portland, OR
User Profile |
Steve2,
I think your work supports the Intelligent Order Processing (IOP) approach I've evangelized.
If TP used real time pricing data to only enter trades at the target price (or better), that would map your experience placing limit orders,
If TP had a plethora of trade opportunities supplied by OV and performed "late binding" of opportunities to actual trades based on real time market pricing, those 21 missed trades would have probably been replaced by 21 other trades that met or bettered the target entry price. Assuming those trades performed as well as the trades you actually took, you would be even further ahead. (Closer to 100% invested anyway...)
IOP would also provide "load balancing" between trade opportunities, allowing more OV users to trade the same strategies without impacting entry prices due to too many orders chasing the same symbol.
Nice work!
Cheers
Keith
|
|
Steve2
Posts: 750
Joined: 10/11/2012
Location: Annapolis, MD
User Profile |
I agree Keith!
Steve
|
|
Steve2
Posts: 750
Joined: 10/11/2012
Location: Annapolis, MD
User Profile |
Attached is IB slippage data from my live trading account (Mar 6th through the end of August). The Summary worksheet shows average slippage by order size and order type. The Trade Data worksheet shows the details for each trade.
Note that I have stopped collecting volatility, volume, opening gap, and daily range data -- sorry Jim but it was too much work. Those columns have been deleted from the spreadsheet.
For the first two months I submitted market orders before market open for both opening and closing trades. In May, June, and early July, I submitted day limit orders at the previous trading day's close for all opening orders and continued to submit market orders for closing trades. On July 8th, I switched back to submitting opening orders as market orders. Nirvana has indicated that they will add opening limit orders to OV, so I'll just wait for that to happen so their effectiveness can be properly back tested.
Note that on Aug 11th I began allowing multiple strategies to trade the same symbol. While the Summary tables and Trade Data show each trade generated by each strategy, trades for the same symbol on the same day were actually combined into a single trade that was submitted to my broker. This means that the slippage by order size numbers in the Summary tables are no longer correct as a single large order is sometimes shown as multiple smaller orders in the tables. The bottom line slippage numbers in the table columns are correct and the slippage numbers by order size are not too far off as the number of times this happens is small compared to the overall number of trades. I did this because it's more important for me to be able to compare my trade record to OV historical trades and to assess individual strategy performance. Also, given my trade sizes and the liquidity of the stocks I'm trading there is no correlation between trade size and slippage.
Slippage for market orders submitted before market open is once again slightly negative at $(0.000792) per share. This was caused by a single anomalous closing trade (LLTC on 8/13). The stock gapped up 5% at the open and retraced down to the previous close within one second of market open. OV, of course, recorded the opening price as its exit price but my market order filled after the retracement. This resulted in $798 of negative slippage. Based on pass results, I would expect this to be recovered over the next 3 or 4 months. Well see...
No opening limit orders or MOO orders were submitted this month so those stats remain the same.
Steve
Attached file : Slippage History - IB Live 1 Thru 13-08.xlsx (57KB - 216 downloads)
|
|
Jim Dean
Posts: 1059
Joined: 10/11/2012
Location: L'ville, GA
User Profile |
Thanks, Steve.
|
|
kmcintyre
Posts: 890
Joined: 10/11/2012
Location: Portland, OR
User Profile |
Steve,
My hats off to you for keeping such meticulous records!
Nice returns too.
Mind sharing the ballpark account size? I see similar returns with my 100K simulation account, but real world returns on a smaller account (started at 5K, now 20K) have gone nowhere. (I still contend size matters with RTM trading...)
So now I'm starting to keep more detailed records too. (Which obviates some of the benefits of set-and-forget black box trading!)
Keith
|
|
Steve2
Posts: 750
Joined: 10/11/2012
Location: Annapolis, MD
User Profile |
Keith,
I started with $50K in March and added another $50K at the beginning of July. My commissions and margin interest are running about 8% of profits. With a $1 minimum commission, there's definitely a minimum account size that makes sense for RTM strategies but I haven't investigated what that is. When I started, I wanted an account size that would enable the average trade to be above 200 shares to minimize the impact of the minimum commission. $50K gave me an average trade size of 250 shares given my strategies and account settings.
Steve
|
|
kmcintyre
Posts: 890
Joined: 10/11/2012
Location: Portland, OR
User Profile |
Steve,
Makes sense. Time to rachet up the account size.
Keith
|
|
Fred Gordon
Posts: 481
Joined: 10/11/2012
Location: Fayetteville, Ga
User Profile |
Steve2, have you compared results between using pre-open Market orders (I am assuming "Day orders) and MOO orders? Is IB your online broker?
I use MOO for most of my transactions and differences between closing and entry prices can be signicant. Pre-market "Day" Market orders may answer. I have tried LOO and "Day Limit set to the previous evening's closing price but experienced too many dropped orders for both. I use a live IB account.
Thank you,
Fred
|
|
Steve2
Posts: 750
Joined: 10/11/2012
Location: Annapolis, MD
User Profile |
Hi Fred,
IB is my broker but I haven't tried submitting pre-market or post-market orders. As the spreadsheets posted in this thread show, I've only used market orders submitted before market open (but executed at market open) and day limit orders at the previous trading day's close. Since I'm able to on-average match OV's entry/exit prices using market orders submitted before market open, I'm sticking with them. Submitting day limit orders at the previous trading day's close does handily beat OV's entry points but, as you point out, there can be a number of unfilled orders. In the 6 weeks that I tried this, the lost profit from the unfilled orders outweighed the gain from beating OV entry points but 6 weeks is not really much of a test. Ed, indicated at one point that they will add limit orders to OV so we will eventually be able to simulate the impact of this.
If you do try submitting pre-market orders, I'd be very interested in knowing how it works.
Steve
|
|
Fred Gordon
Posts: 481
Joined: 10/11/2012
Location: Fayetteville, Ga
User Profile |
Steve2, I think we are talking about the same kind of order type. I called it a pre-market order but meant a Market order that transacts at market open or very near to it but the price will not be at opening print like MOO and LOO orders. I should think the main advantage would be, not being caught in an opening price gap with an attendant snap back. When I try them, I will advise.
Fred
|
|
Steve2
Posts: 750
Joined: 10/11/2012
Location: Annapolis, MD
User Profile |
Fred,
Ok, I misunderstood. If you look at the latest spreadsheet posted in this thread, the first table in the Summary worksheet (Market Before Open) shows the slippage I've gotten using this method for 450 opening trades. It's virtually the same as submitting MOO orders. The one caveat is that I think the slippage you get will vary based on the liquidity of the stocks you are trading. I'm trading pretty liquid stocks (10-bar average volume greater than 1M shares).
Steve
|
|
Steve2
Posts: 750
Joined: 10/11/2012
Location: Annapolis, MD
User Profile |
Attached is IB slippage data from my live trading account (Mar 6th through the end of September). The Summary worksheet shows average slippage by order size and order type. The Trade Data worksheet shows the details for each trade.
Note that I have stopped collecting volatility, volume, opening gap, and daily range data -- the July spreadsheet is the last one that contains these columns.
For the first two months I submitted market orders before market open for both opening and closing trades. In May, June, and early July, I submitted day limit orders at the previous trading day's close for all opening orders and continued to submit market orders for closing trades. On July 8th, I switched back to submitting opening orders as market orders. Nirvana has indicated that they will add opening limit orders to OV, so I'll just wait for that to happen so their effectiveness can be properly back tested.
Note that on Aug 11th I began allowing multiple strategies to trade the same symbol. While the Summary tables and Trade Data show each trade generated by each strategy, trades for the same symbol on the same day were actually combined into a single trade that was submitted to my broker. This means that the slippage by order size numbers in the Summary tables are no longer correct as a single large order is sometimes shown as multiple smaller orders in the tables. The bottom line slippage numbers in the table columns are correct and the slippage numbers by order size are not too far off as the number of times this happens is small compared to the overall number of trades. I did this because it's more important for me to be able to compare my trade record to OV historical trades and to assess individual strategy performance. Also, given my trade sizes and the liquidity of the stocks I'm trading there is no correlation between trade size and slippage.
Slippage for market orders submitted before market open is once again slightly positive at $0.001986 per share.
No opening limit orders or MOO orders were submitted this month so those stats remain the same.
Steve
Attached file : Slippage History - IB Live 1 Thru 13-09.xlsx (66KB - 221 downloads)
|
|
Mark Holstius
Posts: 744
Joined: 10/11/2012
Location: Sleepy Hollow, IL
User Profile |
Thanks for all your work on this Steve - nicely done & a lot of work on your part!
Appreciate your sharing...
Mark
|
|
kmcintyre
Posts: 890
Joined: 10/11/2012
Location: Portland, OR
User Profile |
Yes, Steve!
Thanks so much for keeping this data and sharing the results.
Keith
|
|
BrianD
Posts: 302
Joined: 2/23/2013
Location: Grand Rapids, MI
User Profile |
Steve: I have been tracking my real time slippage closely for the last few months. If memory serves me correctly, you use IB and Smart routing. Do you have "Seek Price Improvement" in the routing settings enabled too?
In a very unscientific manner, I enabled this routing feature and made all trades 'Min Trade Size = 100', at the same time. Once done, these two settings provided definable positive improvement to my actual versus OV entry/exit values - both long and short.
I'm curious to know if you have Seek Price Improvement setting enabled. I can build a case that shows Min Trade Size was not as important as the Seek Price Improvement setting.
I do miss allowing OV to select some smaller size (<100 share) picks, and I've been having a hardest time entering Priceline.com in 100 share lots ;-)
Attached file : NBB0.png (22KB - 197 downloads)
|
|
Steve2
Posts: 750
Joined: 10/11/2012
Location: Annapolis, MD
User Profile |
Brian,
Interesting. My TWS doesn't show those configuration options (see attached screenshot). What version of TWS are you using. I'm using Build 939.3.
Steve
Attached file : TWS Smart Routing Options 10-4-13.png (63KB - 194 downloads)
|
|
BrianD
Posts: 302
Joined: 2/23/2013
Location: Grand Rapids, MI
User Profile |
Well, I guess that answers the question ;-)
My image is from the TWS API config settings drop down. You are showing your TWS settings in the .png image (I assume). I should know that as you are executing manually...
Not sure where it is in TWS client. Seems logical it would be with Smart routing. Definitely a different setting box between API and TWS client...
My evidence is way too little to draw conclusions, but across around 120 trades (120 before and 120 after) long & short, the results are nice. I think >100 smoothed out some big 'pothole' slippage hits, and the "price improvement" added a nice overall touch.
But ask me after 1,000 trades :-0
I'm at 939.3 too. I attached image of the price improvement 'info' bubble for the check box. Found the setting in one of IB's demos or manuals.
Attached file : TWS-API routing.jpg (75KB - 182 downloads)
|
|
Steve2
Posts: 750
Joined: 10/11/2012
Location: Annapolis, MD
User Profile |
I see. Just checked the IB Gateway and those Smart Routing options are available there but not through the TWS client which I use for my live trading. Will check with IB customer service and see why.
Thanks,
Steve
|
|
BrianD
Posts: 302
Joined: 2/23/2013
Location: Grand Rapids, MI
User Profile |
Gateway only? That makes no sense. Please let me know IB response.
|
|
Steve2
Posts: 750
Joined: 10/11/2012
Location: Annapolis, MD
User Profile |
Heard back from IB customer service. That option is available in both interfaces. Only some order options are enabled by default and the defaults are different in each interface. To enable "Seek Price Improvement" in TWS, go to Configure > Features > Order Management > Order Attributes and check "Seek Price Improvement".
I'll give this a try for awhile with my live account and see if it improves entry/exit points. Will let you know at the end of the month.
Steve
|
|
BrianD
Posts: 302
Joined: 2/23/2013
Location: Grand Rapids, MI
User Profile |
Makes me wonder wbat client and Gateway settings operate independently. Certainly a lot of settings in client not included in gateway.
|
|
Steve2
Posts: 750
Joined: 10/11/2012
Location: Annapolis, MD
User Profile |
Attached is IB slippage data from my live trading account (Mar 6th through the end of October). The Summary worksheet shows average slippage by order size and order type. The Trade Data worksheet shows the details for each trade.
Note that I have stopped collecting volatility, volume, opening gap, and daily range data -- the July spreadsheet is the last one that contains these columns.
For the first two months I submitted market orders before market open for both opening and closing trades. In May, June, and early July, I submitted day limit orders at the previous trading day's close for all opening orders and continued to submit market orders for closing trades. On July 8th, I switched back to submitting opening orders as market orders. Nirvana has indicated that they will add opening limit orders to OV, so I'll just wait for that to happen so their effectiveness can be properly back tested.
Note that on Aug 11th I began allowing multiple strategies to trade the same symbol. While the Summary tables and Trade Data show each trade generated by each strategy, trades for the same symbol on the same day were actually combined into a single trade that was submitted to my broker. This means that the slippage by order size numbers in the Summary tables are no longer correct as a single large order is sometimes shown as multiple smaller orders in the tables. The bottom line slippage numbers in the table columns are correct and the slippage numbers by order size are not too far off as the number of times this happens is small compared to the overall number of trades. I did this because it's more important for me to be able to compare my trade record to OV historical trades and to assess individual strategy performance. Also, given my trade sizes and the liquidity of the stocks I'm trading there is no correlation between trade size and slippage.
IB provides a Smart routing option to seek price improvement. When this option is used, Smart routed orders sweep certain exchanges for hidden liquidity at more favorable prices, then checks the dark pools as usual before routing to an exchange. Beginning on 10/9 I began submitting all market orders with this option set. This continued through the end of October. An additional slippage table has been added to the summary worksheet to evaluate slippage using this option (order type is MBI). 69 orders were submitted using this option. The average slippage for these orders was $(0.006039) per share or about $(155) for the 25,700 shares traded this way. As a result, I am discontinuing use of this option and will continue to submit market orders before market open with no options set.
Slippage for market orders submitted before market open (with no options set) continued slightly positive at $0.001813 per share.
No opening limit orders or MOO orders were submitted this month so those stats remain the same.
Steve
Attached file : Slippage History - IB Live 1 Thru 13-10.xlsx (75KB - 274 downloads)
|
|
Fred Gordon
Posts: 481
Joined: 10/11/2012
Location: Fayetteville, Ga
User Profile |
"I am discontinuing use of this option and will continue to submit market orders before market open with no options set."
Hi Steve,
Are you including min. vol. settings, round lots only, "Smart routing" and "Seek better price", when you say: (no options set)?
Also, are you using this scheme for exits only?
Thanks,
Fred
|
|
Steve2
Posts: 750
Joined: 10/11/2012
Location: Annapolis, MD
User Profile |
Hi Fred,
Sorry for the confusion. I'm still trading stocks with 10-day average volume > 1M shares. I'm trading round lots and using SMART routing. Orders are submitted as market orders the night before market open. I'm doing this for both entry and exit orders. The only difference during the test was that in addition I checked the "Seek Price Improvement" routing option. This was done for both entry and exit orders.
Steve
|
|
Fred Gordon
Posts: 481
Joined: 10/11/2012
Location: Fayetteville, Ga
User Profile |
Thank you Steve, are you submitting the night before (assuming any time after OV posts next day's trades?) for convenience or do you think you get a better price for some reason? I understand you have eliminated "Seek Price Improvement" to compare the effect or you feel it was hurting rather than helping.
fg
|