Many, many years ago, when first looking into investing, the books I read were mostly about buying stocks for the long haul based upon a review of the underlying fundamentals.
Sunday, February 11. 2007
Non Linear Trading
However, I'd classify myself as a trader as I don't really look at fundamentals, but look at what happens on a daily basis. My preference is to day trade, that is, get in at the beginning of the day, and get out at the end of the day. I like to get out at the end of the day because I've noticed the instruments that I trade tend to gap up or down on opening. That can be nerve-wracking. I havn't studied Single Stock Futures enough to use them to minimize the opening gap risk.
I've analyzed markets through various technical analysis methods such as moving averages, stochastics, trend-lines, linear regressions, and other interesting analytical mechanisms. The randomness of the markets are such that you can't rely on these only. Indicators and visual charts work together to set an appropriate entry and exit points. Some of these are hard for a computer automated solution to perform. Well, I've realized I need to finesses some of these algorithms to only get in to position a few times a day and ignore a lot of the noise that exists.
But I've wanted to trade the noise as well, also known as the ranging randomness. A bunch of tools come into play on this. And this is where Market Makers make their bread and butter.
I've been looking at ways to play the markets without really caring if the markets are ranging or trending. Two references have come to light over the last couple days of my research that help to refine my thoughts.
The first is an article called "Applying Pit-Trading Techniques to Electronic Trading" in the November 2005 issue of Technical Analysis of Stocks & Commodities. The article describes the art of playing the spread. The author, Clem Chambers, claims that "the market pays for liquidity". Liquidity is supplied through limit orders. By supplying short and long limit orders, one can play the spread. He also says that "if the market is wilder than you like, invent your own spread and trade outside the market's spread". This is something I'm going to try.
To reinforce and flesh out these ideas, I saw a discussion on Elite Trader where they discussed a form of this strategy. The remainder of this article assembles excerpts from the thread. If some one else can provide good references on the web for formalizing these strategies, please let me know. There are many good raw ideas in there--I have to work thorugh them and pull out the good ones.
My trading is kind of high frequency in the futures with either a profit fill order or an new entry order every 1 point level in the es sp mini. No mater what price is doing in the es at each 1 point level I have short and long orders pending---------buy to covers, sell to covers, new long orders, or new short orders. As price moves up and down on a daily basis my ratio between my short and long positions constantly changes.
If we have a day like yesterday then my long side positions are depleting and my short side positions are accumulating---------I started yesterday at about 72% to 28% longs to shorts and at the close I was sitting about 81% to 19% shorts to longs.
As priced moved about during the day I made profits mainly from the long side, but I also had numerous short trade 1 point profit fills anytime price moved lower. Every 1 point of emini movement which ever direction is giving me a profit and also at the same time an accumulation to one side of my position or the other. Thursday was an extremely high frequency trading day and I had more r/t's that day then any other for this system that I use-------several hundred r/t's in one day is high frequency enough for me.
That was suppose to be 72%/28% shorts to longs prior to the Thursday news selloff-----then at the end of Thursday I was at 81%/19% shorts to longs. At the lows of Thursday I was at one point sitting around 58%/42% shorts to longs with the big spike down. How this all played out Thursday would take a long time to explain as I had short side position buy-to-covers down every point to 1180 and new long buy orders down to 1190 sitting for the overnight session. Once the price started tearing down I was clicking off orders like a madman-----fortunately I use xtrader which was rock solid all through the price movement.
I trade this way for the ability to be neutral day after day and not having a "need" for the market to go one way or the other to be profitable. You do need multiple accounts that are cross-margined and starting this type of system is the crucial part. If you start this system correctly you will have your cost basis of the longs and shorts spreading farther and farther apart, to a point where the two sides cost basis are outside of the boundaries of the daily trading range. This creates the perfect situation as big trend days or multiple trend days do not hurt you in any way.
The repetitive and continuous 1 point profits on both the long and short side all through out the day are the profit accumulation strength of this way of trading. Yes range or choppy days are really good for this as both sides of the trade end the day with nice gains and a further spread of the sides cost basis.
I will try to take advantage of extreme moves like Thursday to let the shorts run a bit as they did below the 1180 level and my longs only had accumulation orders down to 1190. When price was moving back up after the 1170 lows, I put the needed longs that did not get added in at 1176 to 1180. 1171 to 1189 was missed for the longs since I did not have standing orders to accumulate below 1190. I never have standing long orders on for more then about 8 to 10 points below the current price levels incase some news event hits. Thursday I had to add 19 price levels of long orders and this did not happen until the 1176 to 1180 levels as I wanted to know what exactly had caused the big sell off. The short side was covered for the ones that did not have standing buy-to-cover orders between 1172 and 1175, so that came out about the same as if I had just had standing buy orders all the way from 1179 down to 1171.
There was a few good whipsaws in there during the sell off and recovery that had some good profit hits so Thursday turned out very profitable. I think I wore my mouse out that day with all the frantic order clicking.
If price levels are trading down at the lows of my plotted fib levels, like around 1130 then I would be at 35% to 65% shorts to longs. In this case I would have stops in place for my long positions at all time for any news related events. The short side in this case would have buy back orders down to the long stop level and then the rest would be able to free fall if the news was very bad. I also do not leave all these longs on for the overnight session when at the lower fib levels {I go to 40/60 or better at the end of every day session}. I want to be properly hedged day after day so these type of rules must be followed to protect all of the accounts capital.
I do have one other system that I trade during the day which is much easier to trade then my ratio trading and this system can run with the intra-day trends. This is a linear regression line based system that also trades in both directions at times until strong trends are established. If we have big trend days like Thursday and Friday up moves then I can take advantage of this in a separate system.
I have separate short and long accounts for my ratio trading that are cross-margined and separate short and long accounts for my linear regression line intraday trading that are cross-margined. At times I may move funds between the ratio trading accounts at the end of a week to keep the account balances in line with a formula I use depending on what the current ratio is on that day. This is almost like having two commingled hyperactive swing trades on in both directions at the same time----------- just their position size adjusts over time in relation to the markets price movement. Swing trades have fairly static cost basis though and my system has a very active movement of each sides cost basis as each week goes by.
How many systems can have a 100% profit rate per trade, week after week for all closed out positions? This is the only way I have concluded to get an end of year profit factor or per trade rate that is very very high.
Re last paragraph. Try to give consideration to using the way you determine to change the cost basis over time as a strategy that could be applied to modifying orders for the purpose of reversing holds on non closed out positions. Say for example you decide to close out; then you could in the appropriate account "affect" a reversal instead of a "close out". I feel that this is, in effect, a way to keep your stratey decision making the same and at the same time continue to be in the market instead of sideling that capital until you later rotate it back nto the balancing process that you do. It is the difference between considering profits per trade and profits per available capital.
You do not fix things that are not busted. Terrific. There is still the aditional element of sidelined capital that you have.
I really recognize how you do keep a lot of capital working and chunck off steady profits as the directional nature permits. I do see you feeding that "taken profit" and subsequently reapplying it. My suggestion is to do a "pseudo reversal" across accounts and utilize the exit decision as a more powerfully leveraged decision (twice the financial power).
First off, the system does not need to do anything special on trend days and that is very important or this system would have a big weakness in my opinion. If you have a big price swing like Thursday then sure I will try to take advantage off this, but I do not need to for robustness of returns.
What I was looking for when I started to develop this ratio non-linear trading method was three primary points.
- I wanted to have a system that could return 3% a month or more on the account balance.
- I wanted to not "need" the market to go any particular direction on a daily basis.
- I wanted a method that would benefit from growing position size and not be hindered by this.
- I did not want the system to degrade in performance from an increase in position sizing.
Running the short and long cost basis away from the current actual trading range is a function of the way I trade all the intraday constant up and down 1 point profits that continually accumulate----------you would have to watch me for 30 minutes to see what I am talking about, then it would make perfect sense.
Quote from tradingbug:
I am currently working on an ES system that does a simliar thing. I look on the daily timeframe to find what would probably be the most likely outcome for the day. Then I use the 30 minute timeframe to determine which way I should take my position trade and scalps. Then I look at the 5 minute and buy pullbacks within my overall position trade to scalp within my position. The key thing is determining when to switch ones position trade direction. You have to get callibrated to the volume of the 30 min and 5 min timeframe along with an absolute indicator.
I think its best to look at the big picture and try to firgure out how the small timeframe influences the longer timeframe.
Your ratio of fractal durations (30/5) is good (B+); for the same purposes, try using the 15 and 2 for a ratio of (15/2).. This will shift your timing ahead. What I mean it that you will make the same set of decisions but they will be made sooner than before.
There are three threads here now: this one; the clininical hypnothreapy one (crying..etc) and the trends stuff that collectively show how persons who are able to entertain with a depth of understanding and can get past being "too smart" and really be able to consider how one "grows" to be able to, optimally, take out of the market what is being offered by the market.
The salient ingredients involve:
- Knowing concurrent independant trades are required; the trades must be able to run independantly of each other so no netting of positions is taking place.
- All measures of performance are focused upon return of total leveraged capital being in the market at all times (Thus, edge trading is passe and not a possibility for makng money).
- The mind and its growth, maintenance and repair is a 24/7 obligation based upon the fact that there is no alternative. Your subconscious mind works all the time and, particularly, when you sleep. Its function is to "organize" your immediate past EXPERIENCE into the subconscious belief system your life has given you. Imagine going from "edge trading" to continuous seamless trading where coordinated independant accounts continuously extract money as price change occurs.
- Knowing that all sensory inputs are continually paired with the emotions (and the biochemical generation of what "greases" the facile operation of the mind and/or blocks you from thinking about that which endangers you).
Today both my short and long side made booked profits, but what was todays direction------ a little down, some sideways, a nice move up, a little down. My system picks the direction of the market on a macro basis by the ratio----------the higher the daily price levels the more the short side is accumulating positions and just the opposite when the market moves downward. Of course I look at these ratio's in relation to where I have my daily chart fib levels plotted.
On strong intraday trends I am playing one side heavier then the other side based on the indicators I use. At the end of each day though, if I have played the long side real heavy intraday, I must make sure my ratio is in line just prior to the cash session close. For instance, yesterday I had to close out extra longs to be at my proper ratio right at the end of the day. Actually there has been numerous days in the past several weeks where I have had to close out extra positions because of very strong trends running to the finish of the day.
There is several different techniques that I use to move both sides cost basis away from the current trading range-------- but the primary event that moves the cost basis up for the shorts and down for the longs is the way I trade all the constant intraday up/down 1 point profit covers---------some of this is done with the play of position size at these levels and some of it has to do with what gets covered when---------but 1 point of profit is the minimum I take per trade when handling the intraday repetitive trades.
See the "direction" of the day is not always clearly defined or what is the primary means to collecting profits day after day. Daily direction is at times the wrong focus or the wrongly weighted focus for a trade system--------- again I have to comeback to the power of the repetitive and constant intraday price movement within each candlestick and within each intraday trend.
I really do not know of specific references for this method, it is just what I have gravitated too from frustration with the limitations of linear based trading----------getting into a trade with a target and a stop, then "needing" the market to go a specific direction to claim profits for that individual trade. Apparently this is what some traders call non-linear trading, so maybe under NON-LINEAR in google there is some additional ideas for this--------not sure though?
It is wonderful to see someone using an unconventional method successfully - especially on these boards. This technique itself has been around for a while though. The following was written more than a hundred years ago (around 1895) by Charles Dow (founder, Wall Street Journal).
"Catching the Fluctuations- During a "Traders' market:, or a market without any pronounced trend one way or the other, any active stock will move over certain points dozens of times. The plan is to place a net that will catch these daily fluctuations. Buy 100 shares of, say, St. Paul, at the market price, and 100 more every half point up or down, but don't hold more than a 100 at a time at the same figure, and don't accumulate more than 600 shares altogether. Treat every purchase as a separate transaction, and whenever a profit of one point net is shown , sell that 100 shares, buying back on a one point reaction, When a purchase and sale are both indicated at the same figure, do nothing - simply hold that 100 shares, but for convenience assume that 100 has been sold and 100 bought. If St. Paul should keep on going up without a reaction, you would thus always be long 200 shares. Don't get frightened because of a temporary downward tendency. The fluctuations are what bring you profit. Great care must, of course, be taken not to work this system on the bull side if the general trend is downward, or on the bear side if the trend is upward."
Several variations of the above general method are being used in the ES market. An Exchange member of CME I met last year was using a very similar method and doing as much as 8K to 10K roundturns per day. He, though, viewed his method as an attempt to 'make the spread'.
See the "direction" of the day is not always clearly defined or what is the primary means to collecting profits day after day. Daily direction is at times the wrong focus or the wrongly weighted focus for a trade system--------- again I have to comeback to the power of the repetitive and constant intraday price movement within each candlestick and within each intraday trend.
Very interesting comment here. I have found that when my position trading bias changes, it can change into either a reversal OR congestion. If its a reversal, the position trade works out great and I continue to scalp in the direction of the reversal. If its congestion, then I can scalp a little in my position biased direction and the day will still end up a little positive as I am essentially buying down and exiting when a more favorable swing comes in. More often than not I can wait in a lateral trend to exit my position trade at a minor loss.
After congestion comes a BO and then the whole process above starts again.
A net position {if you picked the right direction that day----IF!} will rarely on a daily basis beat a system that literally profits from almost every single gyration of price movement during that day. There is more potential POINTS of profit stored in every single candlestick and every little fluctuation of price movement then there is in most trends from beginning to end during the day. Price moves so erratic through out the day that every little fluctuation has the potential for profits and the only way I have found to take advantage of this, is to be in both sides of the market at all times.
I pull off any extra contracts that I played intraday just prior to the close to be within a formula of the ratio I need to be at for the current price level at that decision time. So yes it is a formula that I use to get my overnight position set properly for the ratio at the current price level when the market closes.
Several years and bigger then what you have listed. Started this with a 70,000 account with bigger spacing between the levels I would play. Over time I am now playing every es point in both directions just at different ratios.
As of right now both sides of the system made some very good booked profits today-----------the short side made some money the first part of the day, then the long side did very well up to the close. Now the short side is profiting again with the earnings news.
To safely start this trade from scratch, I would say that a 100K account could trade every 4 points of es movement. Over time and with the account size increasing you then move to 3 point levels, 2 point levels, etc. The turning on and the start point for this method are critical for immediate success.
I hit 100% to 0% shorts to longs for a period of time in the overnight for my ratio trade----------the china currency news run-up finally touched the price level for this to occur. I may hold off adding any longs until the es penetrates 1135 with the lack of clear news out of GB.
When at the bottom of what I have projected of the current range for my ratio I would be 35 to 65 short to long-------- I never go below this and then I keep a stop for the long side all the time when I have 40% or more longs for news spikes. When at the lower levels of my price range when the longs are heavier, I will close everything down just before the close on Fridays at 50 to 50. Sunday night I will rebalance everything back to the proper ratio once the market reopens.
Where you start this trade from {price level} is very important and yes one side goes into the negative on the very first day if that was a heavy trend day. This trade is best started where the 50 to 50 price level is {right now I have 1160 to 1170 as the range for where my 50/50 is located}. As every little up and down movement of the es is making 1 point profits then the two sides cost basis are adjusting away from the current trading price-----------this is very powerful as the 1 point accumulations are racking up minute by minute. When you start trading for 1 point from .25 to .25, and .50 to .50, and .75 to .75, and .00 to .00 on both the short and long side at the same time the profit accumulation affect is very fast.
You can't start this trade with a small account size in the beginning-----------the minimum would be a 70,000 account and then the spread between levels traded would have to be wide in the beginning weeks of trading.
Not necessarily----------If you were to start this trade and the market traded in a tight range or was choppy for several days then you would have the perfect situation. Your long and short side accounts would both be building up profits as the trading price stayed in lets say a 12 point range over these start up days. Yes at any moment there will be one side of the trade with positions in the negative, but you are hedged by the other side--------------so the rapid and repetitive profit hits are outpacing the amount of loss from the side that is "technically" negative {not a realized negative}.
To focus on the side that is negative will mask your ability to comprehend the amount of realized profits from all the 1 point accumulations. This is the aspect of ratio trading which is difficult to "see" unless you watch this realtime. I will give you a better example later----------------have to go meet friends now.
My company developed a high-efficiency market depth processor for stock/index arbitrage. We deploy it for interlisted arbitrasge trading - by human traders and machines. Basically, it is high-frequency scalping. We developed it (as a part of our trading system) because market data providers fail to delivery necessary quality of market depth we need, especially for Island and Arca ECNs (we tried Reuters, eSignal, Realtick, etc - all they suck). Anyway, our market depth processor has latency less than 1/10 of milisecond and it runs on a cheap single-CPU PC with throughput up to 100,000 messages a second. For example, currently for all Island stocks CPU load is less than 3%. Processing includes parsing direct feed, sorting of depth (with optional aggregating) and sending it into a network for distribution. In terms of money we have traders who quite often are making 5K a day because of the superior speed of the solution. Machines are making 7-10K daily. To summarise : without high-quality data (what implies minimum latency) high-frequency trading is impossible
To my knowledge, reference to high frequency *analytics* is mostly of interest to academics and researchers, who can now study market microstructure matters on a tick-by-tick scale, which result in new implications for risk management, dynamic hedging, agents behavior, etc.
High frequency *trading*, however, is mostly of interest to hedge funds and banks which separate their operations between low frequency ones and high frequency ones. Low frequency usually implies heavier/deeper analytics. The edge is in the numbers, not in the trading. High frequency implies more simple analytics, but the timing and the volume are crucial. They rely more on market microstructure than on acurate solving of PDEs. Yet it does not mean easier to execute. The arbs are risky because you can get stuck on only one leg, and information can turn against you within a second. Unclosed arbs are your loss.
High frequency trades can go from statistical arbs (most commonly heard), to volatility frequency trades/price mean reversals. Most simple arbs, like mergers, correlation triangle arbs, and such, are executed at high frequency because they go vapor quickly.
There would be many other examples I cant think of at the moment. But that should give a good picture. To my belief, HF vs. LF is not about how many trades you shoot per day, but what your strategy is about. HF has been there for a long time. But it's only recently, because of the required skills that distinguish them from low frequency/deeper analytics trades, that a separation has been made. Mostly in hedge funds.
Of course information risk exists!
Information risk is why trading is "expensive". Homogeneous information explains liquidity, and heterogeneous information explains market frictions and trading gridlocks. The more heterogeneous the information is, the highest the informational cost of a trade will be. This informational cost is translated into the bid/ask spread. This is the basis of market microstructure information models.
This is why most of day traders who think they can make money with "trends" get screwed. While they think their trades only cost them broker commissions, they are actually affected by a much higher cost: that of superior information.
Using limit orders on markets does not avoid you the cost of the bid/ask spread. Bid and ask quotes are values conditional to trades based on the information structure of the market. Placing a limit order will only get you hit by traders potentially having superior information, and leave you screwed with losses.
In terms of informational value, technical trading models are not far higher than placing blind trades at random on a market on which vultures with accurate insider information or arbitrage models are waiting for their lunch.
For 11 months. As we are acquiring experience we are adding different trading models. But foundation remains the same - low latency of data feeds and fast orders execution.
Please be reasonable. Our approach does not tolerate the risk of this magnitude. It is an arbitrage - the system takes position for a period of fraction of second and tries to unwind it in another market as soon as we got a fil or partial filll. If it fails to unwind it for profit for a reasonable period (usually 5 seconds), a stop order is issued. As a result losses are pretty limited. So as you can see for our model speed is everything. That is why we process market data in-house because other solution fail to meet necessary latency requirements.
"My trading is kind of high frequency in the futures with either a profit fill order or an new entry order every 1 point level in the es sp mini. No mater what price is doing in the es at each 1 point level I have short and long orders pending---------buy to covers, sell to covers, new long orders, or new short orders. As price moves up and down on a daily basis my ratio between my short and long positions constantly changes........"
Macro - are you still trading this? As best I understand your method of trading, it seems to be a short volatility (mean reversion/option selling, etc) type of strategy with a martingale element embedded into it. What do you do with the accumulating unrealized losers? It seems like in a period of prolonged trending (accentuated by large directional moves), you can easily blow up (or build substantial loses). I simulated your approach (from what I could gather) and it did poorly in the late 90s and during the Bear market of 2000-2002 (as expected). You must have some other signal you are relying on - pure position sizing won't give you a true edge. Like with any martingale strategy, the drawdowns could be huge and provided no artificial limits are hit (forcing liquidation) the capital requirements needed to withstand the drawdown make the resulting returns unattractive. In terms of the mean reversion element, you can probably do better by avoiding the costs associated with frequent trading (i.e. average down less frequently). I don't mean to be critical of your strategy, just trying to understand it (and sharing my concerns in the process). Thanks for posting.