Slip and Commission
Slip and Commission
Dear c.f. ,
Sorry to bat on about slippage but Im beginning to realise its a huge huge , if not most important part of the system.
I have designed a couple of portfolios of systems now with great risk adjusted returns up to 30 % slippage or $ 40 per trade min. slippage .
Is there an opinion again , on if this sounds too optimistic
What level are other people using for max $ slippage and % slippage.
Thanks again
Chris
Sorry to bat on about slippage but Im beginning to realise its a huge huge , if not most important part of the system.
I have designed a couple of portfolios of systems now with great risk adjusted returns up to 30 % slippage or $ 40 per trade min. slippage .
Is there an opinion again , on if this sounds too optimistic
What level are other people using for max $ slippage and % slippage.
Thanks again
Chris
-
- Roundtable Knight
- Posts: 1842
- Joined: Tue Apr 15, 2003 11:02 am
- Contact:
Chris,
I generally test with 8% slippage.
If you have a good broker, this corresponds pretty closely with what I have seen in real orders. Sure, at times, I'll get much higher slippage, but there will be many times when you get no slippage.
If you have a lot of systems that only use stops at volatile pricepoints, or if you don't have a good broker, or you trade illiquid markets, you might go higher.
I don't use minimum slippage. The other slippage parameters are there so we can emulate other methods in common usage. Many traders prefer to use flat assumptions.
For commission you should use whatever you expect to get charged which you should be able to find out from your broker ahead of time.
Using $75 slippage plus commissions which correponds to a $25 commission and 0% slippage and $25 minimum slippage is a common figure, but I don't use this method.
I generally test with 8% slippage.
If you have a good broker, this corresponds pretty closely with what I have seen in real orders. Sure, at times, I'll get much higher slippage, but there will be many times when you get no slippage.
If you have a lot of systems that only use stops at volatile pricepoints, or if you don't have a good broker, or you trade illiquid markets, you might go higher.
I don't use minimum slippage. The other slippage parameters are there so we can emulate other methods in common usage. Many traders prefer to use flat assumptions.
For commission you should use whatever you expect to get charged which you should be able to find out from your broker ahead of time.
Using $75 slippage plus commissions which correponds to a $25 commission and 0% slippage and $25 minimum slippage is a common figure, but I don't use this method.
-
- Roundtable Knight
- Posts: 1842
- Joined: Tue Apr 15, 2003 11:02 am
- Contact:
Sorry,
Welcome to the forum!
We were both referring to a setting in VeriTrader.
The 8% is 8% of the distance between the desired price and the worst price of the day. In the case of a buy this would be the difference between the entry stop and the high of the day.
For example, if you have an order to go long Gold at $410.50 and the high of the day is $420.50. You would get a fill at ($410.50 + (8% of ($420.50 - $410.50)). This calculates to a simulated fill price of $411.30 for slippage of $0.80, or $80 per contract for 100 oz. Gold.
- Forum Mgmnt
Welcome to the forum!
We were both referring to a setting in VeriTrader.
The 8% is 8% of the distance between the desired price and the worst price of the day. In the case of a buy this would be the difference between the entry stop and the high of the day.
For example, if you have an order to go long Gold at $410.50 and the high of the day is $420.50. You would get a fill at ($410.50 + (8% of ($420.50 - $410.50)). This calculates to a simulated fill price of $411.30 for slippage of $0.80, or $80 per contract for 100 oz. Gold.
- Forum Mgmnt
-
- Roundtable Knight
- Posts: 113
- Joined: Wed Jun 04, 2003 9:44 am
- Location: Somewhere, Hyperspace
Dear Forum Mgmnt,
Your method for calculating slippage is interesting. I tend to use a more conservative method and- upon reading your approach- I wonder if mine is perhaps too conservative.
This might ultimately lead me to prefer one set of parameter values over another set that may actually perform better in real time.
Perhaps you might share your thoughts on striking up a comfortable balance.
Your method for calculating slippage is interesting. I tend to use a more conservative method and- upon reading your approach- I wonder if mine is perhaps too conservative.
This might ultimately lead me to prefer one set of parameter values over another set that may actually perform better in real time.
Perhaps you might share your thoughts on striking up a comfortable balance.
Last edited by shakyamuni on Sat Jan 08, 2005 6:19 pm, edited 1 time in total.
-
- Roundtable Knight
- Posts: 1842
- Joined: Tue Apr 15, 2003 11:02 am
- Contact:
shakyamuni,
You've hit upon one of the significant problems with following the conventional wisdom.
Arbitrarily deciding to add additional costs to each trade will skew your results towards favoring longer-term less frequent trades as you point out.
Your assumptions will often drive your testing results. You won't trade many short-term systems, if any, if you assume every trade skids 50% of the distance to the high.
I want to emulate the actual costs I expect to have, so I test with the actual commision I expect to pay. Likewise, I use the slippage method I do because I it is the most realistic and accurate algorithm I know for daily bar data.
The best way to test accurately is to keep track of your actual trade slippage. If you use a percentage-based slippage like I do, then you should keep track of the percentage slippage in your actual trades to see if you have under or overestimated the slippage.
Slippage is very dependent on your broker. If you have a good broker, you can use smaller slippage assumptions. If you have a bad one, you should get a better one![Smile :)](./images/smilies/icon_smile.gif)
You've hit upon one of the significant problems with following the conventional wisdom.
Arbitrarily deciding to add additional costs to each trade will skew your results towards favoring longer-term less frequent trades as you point out.
Your assumptions will often drive your testing results. You won't trade many short-term systems, if any, if you assume every trade skids 50% of the distance to the high.
I want to emulate the actual costs I expect to have, so I test with the actual commision I expect to pay. Likewise, I use the slippage method I do because I it is the most realistic and accurate algorithm I know for daily bar data.
The best way to test accurately is to keep track of your actual trade slippage. If you use a percentage-based slippage like I do, then you should keep track of the percentage slippage in your actual trades to see if you have under or overestimated the slippage.
Slippage is very dependent on your broker. If you have a good broker, you can use smaller slippage assumptions. If you have a bad one, you should get a better one
![Smile :)](./images/smilies/icon_smile.gif)
Forum Mgmnt,Forum Mgmnt wrote:Chris,
I generally test with 8% slippage.
If you have a good broker, this corresponds pretty closely with what I have seen in real orders. Sure, at times, I'll get much higher slippage, but there will be many times when you get no slippage.
If you have a lot of systems that only use stops at volatile pricepoints, or if you don't have a good broker, or you trade illiquid markets, you might go higher.
I don't use minimum slippage. The other slippage parameters are there so we can emulate other methods in common usage. Many traders prefer to use flat assumptions.
For commission you should use whatever you expect to get charged which you should be able to find out from your broker ahead of time.
Using $75 slippage plus commissions which correponds to a $25 commission and 0% slippage and $25 minimum slippage is a common figure, but I don't use this method.
I think we all would like to have this 8 % slippage in average. If I ask people who are running a hedge fund I hear
- * $ 150 round turn /contract commission included
* $ 100 round turn/contract commission included
* $ 90 round turn/contract commission included
* $ 75 round turn/contract commission included
* 50% +commissions (reasoning because experiences sometimes %100)
* 12 - 15 % with minimum slippage + commission
* 8 % no minimum slippage + commission
* 5 % ATR slippage + commission
Maybe other´s could tell their real average slippage and what they assume for their backtesting?
Best regards as ever
Helmut
PS: I edit to include new opinions in the list....
Last edited by Austrian on Wed Apr 05, 2006 6:51 pm, edited 3 times in total.
-
- Roundtable Knight
- Posts: 2071
- Joined: Fri Apr 25, 2003 3:33 pm
- Location: London
- Contact:
What I would also like to know is how people calculate their slippage. It is very clear how to calculate slippage where you have an order at a certain limit. Less so where you have a spread or an MOO order.
Let us take MOO. CSI gives you an opening price which it gets from the exchange. That is the price you use in backtesting. Therefore it seems to me that is the price from which you calculate slippage. Assuming you roll on a spread at the open, again I assume you should calculate the slippage from the opening price as given by CSI for each contract month of your roll.
The other thing however is timing. What symbols does one use in CSI and does this match with how you or the broker executes your spreads or MOO orders. Are you using pit and electronic prices for instance.....................
Let us take MOO. CSI gives you an opening price which it gets from the exchange. That is the price you use in backtesting. Therefore it seems to me that is the price from which you calculate slippage. Assuming you roll on a spread at the open, again I assume you should calculate the slippage from the opening price as given by CSI for each contract month of your roll.
The other thing however is timing. What symbols does one use in CSI and does this match with how you or the broker executes your spreads or MOO orders. Are you using pit and electronic prices for instance.....................
-
- Roundtable Knight
- Posts: 137
- Joined: Thu Sep 23, 2004 11:19 am
- Location: California
excellent questions by all, and great post Helmut. Personally, I have seen developers use $75 c/s for years as a figure for testing. I thought that those numbers were much too low back when commissions would take up $50 of that total. Now, with brokerage costs being lower, I think a $75 c/s figure is fairly accurate, and with more and more markets moving towards electronic execution, the slippage cost should, in theory, decrease as time goes on.
This whole slip and comish thing is always being discussed. I take an approach that most probably would not agree with. As c.f. points out above, over estimating slip means one will favor longer term trading systems. This seems to me to be throwing the baby out with the bath water. I do include Comissions - those are fixed costs. But slippage - well that's another thing. Yes, they are real. But really they are the difference between the price at the time you puts your order in, and the price you get filled at. These can be "managed" to some extent. If you know the nature of the market you're trading, e.g. coffee, slip is terrible, so one might price in your entries (limit orders) or, what I'll do a lot, anticipate the triggering of an system entry signal and get in (or out) early. This takes art and skill, but over time, and in general, I beat all my systems and get positive "slippage". In other words, I get better entries and exits than my system. This enables me to trade good short term systems which look terrible if you factor in things like 8% slippage costs. Now, I traded over 4,000 mostly pit contracts last year so I'm not saying this is easy or for everyone. Just another point of view on the devil known as "slippage".
When it comes to LTTF systems, it almost becomes irrelevant because $75 slip and comish is going to be miniscule compared to the millions that a good LTTF system can make.
At the end of the day, from the point of view of trying to compare one system to another, short term and long term ones, I would rather just use my known comissions and zero for slippage. Then take the average trade net profit and the percent time in the market for each system and compare those numbers. That will tell me how tight my system is to covering slippage costs. Or said another way, what my margin of error is if I don't do better than the system (and therefore get positive slippage).
Just my 2 cents on this subject.
When it comes to LTTF systems, it almost becomes irrelevant because $75 slip and comish is going to be miniscule compared to the millions that a good LTTF system can make.
At the end of the day, from the point of view of trying to compare one system to another, short term and long term ones, I would rather just use my known comissions and zero for slippage. Then take the average trade net profit and the percent time in the market for each system and compare those numbers. That will tell me how tight my system is to covering slippage costs. Or said another way, what my margin of error is if I don't do better than the system (and therefore get positive slippage).
Just my 2 cents on this subject.
Now let us look at the slippage estimates from attain (attain´s slippage averages to rd. 65 $ round turn )- I am still wondering about c.f.´s 8 % slippage assumption because - what slippage has to be calculated with trading size????
Best regards as ever Helmut
Best regards as ever Helmut
- Attachments
-
- slippage.gif (28.78 KiB) Viewed 21451 times
If I use the 8 % slippage value on a more short term futures system with 51 more or less liquid markets I get in TB 2.1b2 rd. 17 $ slippage per round turn.
And with in mechanica´s info-package used trading cost assumption ("Trading Costs: slippage and commission set to equal the dollar value of 15% of the day's range, any day that a trade occurs (entry, rollover, exit).") I get in the same system with 51 markets
$ 39 slippage and commission per round turn.
![Razz :P](./images/smilies/icon_razz.gif)
![Cool 8)](./images/smilies/icon_cool.gif)
And with in mechanica´s info-package used trading cost assumption ("Trading Costs: slippage and commission set to equal the dollar value of 15% of the day's range, any day that a trade occurs (entry, rollover, exit).") I get in the same system with 51 markets
$ 39 slippage and commission per round turn.
Today in the morning I get a message regarding my research on slippage telling me and I believe this message in system design you have to be very very carful and very very cautious - basically I believe this.
... and I think my fellow traders, system designers and hedgefund managers have a problem in revealing their data regarding slippage in public. If so PM or e-mail me
![Cool 8)](./images/smilies/icon_cool.gif)
... and I think my fellow traders, system designers and hedgefund managers have a problem in revealing their data regarding slippage in public. If so PM or e-mail me
![Razz :P](./images/smilies/icon_razz.gif)
![Cool 8)](./images/smilies/icon_cool.gif)
Copied from elitetrader-Forums ›› Technically Speaking ›› Strategy Trading ›› Original Turtle Rules - Discussion with Murray Ruggiero
As one of the most successful turtels could you please comment on the issue of trading cost and slippage, because I cannot understand the difference of your 8% testing approach and Kim´s experience. I think we all in this forum would like to learn about one of the most important factors for system design.
Best regards as ever, Helmut
PS: edited to include following:
ATR system 10 years till today - liquid.set markets
8 % range + commission = $ 46.11
50 % range + commission = $ 58.55
Forum Mgmnt,..........also you use a different instrument portfolio than the original turtles used. that's okay, because some of those markets don't exist anymore, and there are a lot of new ones available.
i have a huge issue with your trade cost assumptions: $75 skid + commission is a huge underestimate. 10 of the 21 markets in the original turtle test have a tremendous amount of skid when you start trading size. now, you are also using instruments such as lumber and orange juice where you are accounting for a large portion of the volume and open interest. when you "print" in your test, any trade above 10 contracts in those markets is going to experience increasing skid with increasing trade size. i am sure that the lumber trade you have open on 12/22/05 might take more than a day to put on, and you will be pushing the top end of the range (meaning that your real skid might be 100%+.) the more accurate way to simulate the backtest is to use slippage as a percentage of the Average True Range relative to the entry/exit. overall, i use 50% slippage for the entire period of my testing, because i sometimes will experience 100%+ skid when i put on a large transaction in some markets.
are you accounting for contract rolls in your backtesting? if you are holding that lumber contract for a year, you are rolling in out and out several times a year, and you get hit on both ends.
i mention these things because my goal of backtesting is to simulate real trading environments and scenarios. otherwise, what's the point of backtesting? making the numbers look pretty get people excited, but it will most likely hurt their real trading.
i want to share this link as an example:
http://www.iasg.com/mpage.asp
if you look at the 5-year track record of the top 100 managers on IASG, only 8 of them have returns above 30% - with some having significant drawdown levels. you will find very very few managers with a 10-year+ track record returning more than 30% per year. three big reasons for understated performance (if those fund managers are at least partly mechanical and swinging for higher returns) are 1) they understate their trading costs and over-optimized their system, 2) they have a structural error in the way they test, and 3) the date they start trading. number 3) is an eye-opener for me: i see fantastic trading systems that work independent of the start date 90% of the time, but that 10% of the time - they have mediocre to poor performance.
the best system that i have been able to make is 26%/yr over a 35-year period with a 66% drawdown (using 50% entry and roll slippage). that 66% drawdown happened in 1980-1981: at that time, only 18 of the 65 liquid futures markets in my system were trading. since then, my largest system drawdown was about 40%. if i choose my start test date in 1981, my MAR ratio performance increases by 73%, my annual Sharpe ratio more than doubles, and my drawdown levels and lengths are cut by a little more than a third. starting dates do make a difference in testing and in real trading. my system performance is 94% independent of my starting date, and that 6% is a realistic killer.
maybe others can come up with better long-term trend-following systems, but i want to understate my system performance with a NON-optimized parameter set using all 65 liquid futures instruments available to me (regardless of the instrument being profitable or not.) i choose to trade that for my hedge fund going forward because i know my performance level has a better chance of standing up in the future.
Edward Kim
2GTT, LLC
As one of the most successful turtels could you please comment on the issue of trading cost and slippage, because I cannot understand the difference of your 8% testing approach and Kim´s experience. I think we all in this forum would like to learn about one of the most important factors for system design.
Best regards as ever, Helmut
PS: edited to include following:
ATR system 10 years till today - liquid.set markets
8 % range + commission = $ 46.11
50 % range + commission = $ 58.55
-
- Roundtable Knight
- Posts: 1842
- Joined: Tue Apr 15, 2003 11:02 am
- Contact:
As I mentioned in the turtle rules document, slippage can be controlled. As turtles we never used stop orders. We had prices we used as mental stops but we almost always used limit orders.
Occasionally, if I knew the floor broker who would be handling the order personally, I would use a market order, but even then I would send the order out in pieces.
Keep in mind that I sometimes had positions of 1,000 to 1,200 sugar contracts, 500 to 1,000 coffee contracts etc. This was twenty years ago. Even with these my slippage seldom exceeded 20% of an ATR. On average it was much smaller, and based on my best recollection less than the 8% figure I use in my personal testing.
I have always suggested that you track your costs and use a number that reflects reality for your trading style. The 8% figure I use has tracked well in actual trades for the markets I test with and might even be a little pessimistic for the way I trade.
I don't recommend trading any illiquid markets like rough rice, lumber, orange juice, etc. If you trade those you should adjust your numbers accordingly. I don't know what the skid would be like in those markets since I don't trade them.
I have no idea what markets Murray used for his test but that may be some of the basis for Edward Kim's judgement. I think his test included several illiquid markets. Slippage is very market dependent.
You are free to disagree with me and use whatever numbers you wish. I offered my opinion above and still agree with that opinion. TBB lets you use whatever assumptions you wish for your personal testing.
One should make a best estimate rather than an artificially "conservative" figure so as not to overly penalize high-frequency trading systems. One should keep track of actual slippage as soon as one starts trading in a real account and then adjust your backtesting assumptions accordingly.
- Forum Mgmnt
Occasionally, if I knew the floor broker who would be handling the order personally, I would use a market order, but even then I would send the order out in pieces.
Keep in mind that I sometimes had positions of 1,000 to 1,200 sugar contracts, 500 to 1,000 coffee contracts etc. This was twenty years ago. Even with these my slippage seldom exceeded 20% of an ATR. On average it was much smaller, and based on my best recollection less than the 8% figure I use in my personal testing.
I have always suggested that you track your costs and use a number that reflects reality for your trading style. The 8% figure I use has tracked well in actual trades for the markets I test with and might even be a little pessimistic for the way I trade.
I don't recommend trading any illiquid markets like rough rice, lumber, orange juice, etc. If you trade those you should adjust your numbers accordingly. I don't know what the skid would be like in those markets since I don't trade them.
I have no idea what markets Murray used for his test but that may be some of the basis for Edward Kim's judgement. I think his test included several illiquid markets. Slippage is very market dependent.
You are free to disagree with me and use whatever numbers you wish. I offered my opinion above and still agree with that opinion. TBB lets you use whatever assumptions you wish for your personal testing.
One should make a best estimate rather than an artificially "conservative" figure so as not to overly penalize high-frequency trading systems. One should keep track of actual slippage as soon as one starts trading in a real account and then adjust your backtesting assumptions accordingly.
This is what TBB does with it's slippage assumptions. It calculates the slippage starting with the open for MOO orders or stop orders which are hit on the open. It will use a percentage of the open to the high for buy orders or open to the low for sell orders for it's estimated slippage.In an earlier post, AFJ Garner wrote:Let us take MOO. CSI gives you an opening price which it gets from the exchange. That is the price you use in backtesting. Therefore it seems to me that is the price from which you calculate slippage.
- Forum Mgmnt
-
- Roundtable Knight
- Posts: 144
- Joined: Mon Apr 19, 2004 3:04 pm
- Location: Old Europe
I have been systematically tracking the slippage on my futures transactions over the last two years. About 45% are electronic and 55% are pit traded contracts. And I am also active in some less liquid markets like DX, RR and LB.
The slippage figure I come up with is about 9% and very much in line with c.f.' figure.
Cheers,
Old European
The slippage figure I come up with is about 9% and very much in line with c.f.' figure.
Cheers,
Old European
-
- Roundtable Knight
- Posts: 144
- Joined: Mon Apr 19, 2004 3:04 pm
- Location: Old Europe
What I find quite remarkable is that illiquidity does not necessarily lead to large slippage.
Lumber is a good example. The typical daily volume in the front month is only about 500 lots. It has many lock-limit days. You can hardly call it a liquid market.
... but the slippage on my MOO lumber orders is almost always zero. I really mean 0.
I often wonder if the market opening price would have been the same without me placing an order and thus participating in the opening price discovery. Can you call this also slippage?
I'm afraid that this is a pure philosophical question!
Cheers,
Old European
Lumber is a good example. The typical daily volume in the front month is only about 500 lots. It has many lock-limit days. You can hardly call it a liquid market.
... but the slippage on my MOO lumber orders is almost always zero. I really mean 0.
I often wonder if the market opening price would have been the same without me placing an order and thus participating in the opening price discovery. Can you call this also slippage?
I'm afraid that this is a pure philosophical question!
Cheers,
Old European
Attain Capital calls it "hidden slippage" and I'm afraid it is very difficult to account for it in backtesting. What backtests show is the performance you may have achieved in the past on condition you had been too small to move the market. Otherwise, your system may generate signals that it would not produce without you participating in a particular market, or if you enter/exit markets at the opening price, you may experience hidden slippage as described below.Old European wrote:I often wonder if the market opening price would have been the same without me placing an order and thus participating in the opening price discovery. Can you call this also slippage?
http://www.attaincapital.com/alternativ ... un2005.htmIt's not only the wide bid/ask spread in illiquid market that is troublesome. Another issue in illiquid markets can be trading system orders moving the market. This is obviously not ideal, as the trading system starts reacting to price moves that it caused. If the system caused the price move, and not some underlying event or stimulus, the odds of prices continuing in that direction once the system is done pushing prices that way is remote at best.
This scenario plays out quite frequently in in less liquid markets like Lumber, Palladium, and Propane. Even a single trading system coming into any of these markets with an order of 10 to 20 contracts can send prices on wild moves as an order imbalance causes prices to shift and open at the much higher or lower prices. The result of this can be an increase in the actual risk per trade while the system tells you the risk is much less (as its basing the risk off the prior day's closing price). While an investor won't technically see slippage on these "gap opens", as the system will show it entered at the next day's opening price and that will be the price investors get, there is a "hidden slippage" in these scenarios due to the "shifted" prices and often dramatically increased risk.
-
- Roundtable Knight
- Posts: 144
- Joined: Mon Apr 19, 2004 3:04 pm
- Location: Old Europe