Well, following the candlestick thread I have been thinking a lot about buying short term weakness. Anyway just for fun I have been trying out a few optimisations, and I have come up with something that is far from perfect but intruiging nevertheless:
I checked the Realized Transactions of your original Sim and the charts of 15 out of 20 of the stock with the highest return were stocks that had very large declines over the next year or two when the market was trending up. It looks to me like these are poorly managed companies (not just value companies), and your buy rules were catching some big volatility moves right after a big but rare move up. I’m not sure what to make of this, or how to take better advantage of it. It is just strange to me.
I also noticed that your Sim was trading some penny stocks. If I add minimum liquidity buy rules of AvgDailyTot(20) > 100000 & Price > 1, the annual return falls to 178% and 68% drawdown.
I agree with much of your comments Denny, it would appear that buying on very short term pullbacks is a good idea.
The “value” ranking system only adds a small amount, this works pretty well with my “random” ranking system too. Interestingly, the more stocks added, the higher the profit per transaction, but the lower the overall returns, which is interesting.
Incidentally, I am picking up some unusual effects with the “price” function, I have studied some of the transactions and some of them are definately bought below the 50 moving average.
What is price exactly? If you substitute “price” with “close(0)” it produces very different results, to my mind this shouldn’t be.
It is my understanding that Price is not supposed to be used anymore. There could be a lookahead bias if you use price in conjunction with SMA() or some other functions. Price was supposed to be mapped to Close(0) but maybe this hasn’t happened. ??
One change, I added the buy rule: low(0) < low(1) AND low(1) < low(2). I added this rule to Stitts challenge also. While it reduced annualized return of the original I felt that it was more robust when running variations. I am also running only 5 years back but the original had similar results over 5 years. Based on the number of trades and avg hold time it looks like it was invested 86% of the time.
Originally I tried this with a different ranking system thinking it wouldn’t make any difference, and initially it didn’t, annualized returns were roughly 300%. But after adding this one buy rule annualized return was 424%/yr (with Dans TF1205 ExcellentO). I was stunned to see returns go up even more by adding back the original ranking.
For kicks I tried a 1 stock sim because I don’t want to trade too much: 480% annualized but 86% drawdown.
These results really seem too good to be true. My first thought was that to improve the sim the focus should be on improving the win rate. Typically counter-trend type systems can have win rates much better than 50%. Perhaps still room for improvement?
This is some really cool stuff! Profit of nearly 5% per trade, almost in business. What made you think of adding the rule about lows? I am so glad you did because I would not have thought of it. I have to admit I am a bit of a novice at this “counter trend” trading thing.
525% annualized - realistic? Haha well if its I will be retiring to Bermuda soon! One issue is of course hindsight bias, the results are getting an uplift from the fact we know what happened, and we can develop rules that work perfectly with the historical data set.
Going forward they may still work, but not quite as well as advertised, so for example, the profit per trade may knock a few percentage points off, and then there is the issue of slippage, it may turn out these stocks have big spreads and poor liquidity, all of which chip away at the annualised returns. But still worth looking into!
I think one issue that has put me off swing trading so far is that its quite “fiddly”, you have to get the timing right and the executions have to be good. Trend following (as with Adv Mom Val) is much easier as you hop on and off trends without worrying too much were they are going to reverse, and there are nice juicy profits that make slippage all but irrelevant.
Incidentally, I think my “price > sma(50)” rule may not be doing much, the “close(0) > sma(50)” rule certainly does, and is far more restricting.
I did want to post a link showing the effects of removing it, however this evening P123 seems to be running very slowly with its simulations, so I think I will have to get back to this tommorow.
Indeed, this is interesting stuff. I just experimented with it a little with the screener, as I’m running out of ranking systems. With the screen my results improved somewhat by increasing the ATRN slightly from 10-14.
I am new to ATRN, but so far find if its too short it does do its job well. So, yes short - but not too short. Of course, I am not using the sell rules and with the screener there is only weekly rebalance. (Still, I like to check systems without the sell rules just to see that the system does not rely too heavily on them. I am always somewhat suspucious of systems that pump results with sell rules that may or may not work going forward.
Whats also noteable about this system is that its so selective. First, when the market is heading south you are not invested. Which is like money in your pocket, Ching! Second, you don’t have to make many trades which is always nice. Looking forward to more discussion and input. This system has potential!
Hi Olikea,
“What made you think of adding the rule about lows?”
I’ve heard that a 3 to 5 day pullback in a trending stock is supposed to be buyable. I think of a pullback as lower lows, lower highs, the last days close < open, and the last days close some level lower then the high a few days ago, probably at least 1 ATR. Followed your example and started with lows and quit while I was ahead.
I was incorrect in my prior post re. Stitts Challenge, the rule I used there was Close(0) < Lowest(#Close,3,1). Same general idea - finding a pullback.
I agree that trend trading with P123 seems to be easier then technical trading. I have had some experience with technical trading and there is so much randomness - what works now doesn’t necessarily work next month. I prefer trading with the more typical P123 portfolio’s, but I find sims like these very intriguing.
I found that increasing the length of the ATR variable doesn’t really seem to help.
I have been doing some revisions, with the aim of making the system more “tradable”. In goes the liquidity rule, and also a minimum price. I found that excluding penny stocks is too restrictive, so I exclude stocks under 10cents. Although IB charge on a per share basis they cap their commisions to 0.5%. The problem is that some of the illiquid issues may have spreads of 2%, meaning the effective slippage is 3%, so this leaves little room for “error”.
I also added rule, 4week relative strenght positive; improves profit per transaction. I reduced the ATRN to > 12 to allow more trades to pass through the filters/
I was also unsure if the sell rules were the best, instead of “stoploss” and “takeprofits” points, i just sold after 3 days:
All of these adjustments are in my revision 2 of the sim:
I am trying to get to grips with the “price” function, it seems to have quite different properties to “close(0)” and this worries me.
I also have experimented with getting the profit per transaction over 5%. It is quite possible, but at the expense of limiting the number of available trades, and thus reducing annualised return. (Though it is worth bearing in mind the annualised return AFTER slippage may be higher).
Here is the sim, with a profit/transaction over 5%:
Yes I am thinking about it. Though my opinion is good traders should be prepared to go “where angels fear to tread”. QQQQ may be very liquid but I have my doubts as to what serious trading opportunities it offers, at least in percentage terms.
By going to places ignored by others, micro caps, illiquids, OTCs, penny stocks, it is my view that the profit potential is higher because you are competing against fewer pros.
So its swings and roundabouts. I tend to think the optimal amount of minimum liquidity is where it does just push your comfort zone a little. That’s what I aim for.
I am going to monitor penny stocks for a while to get a feel for how tradeable they are.
IB charge half a cent per share capped at 0.5% of transaction value. I.e. trading penny stocks costs 1% for a round trip. Its not great but its not a showstopper either.
Interesting this did not improve things. Excluding the sell rules, does anyone have an explaination why the screeners results and sim results might differ so much? (I’ve noticed this before, and frankly it puzzles me).
At first, I thought it might be the ranking system but in this case the results are so restrictive that it really should not make a difference. I have also tested exact same screen and ranking on different days and got vastly different results. If you are using the screener it might be a good idea to track your sim results and compare them.
Olikea, this equity curve is impressive! You do not seem to be too far away from a tradeable system. I really never thought anything like this was remotely possible, but both you and Stitts are blazing P123 readers into a completely new territory. (I’ve always been a buy and hold investor). Press on!
Hi,
Re. the Price function, this is how Marco described it in an old post:
““Price” and “PriceH” are updated once a week, Close(0) everyday.”
So I agree that Price should not be used, but PriceH and PriceL should be ok.
Atrn(10,0) > 15 seems to be the key to this system. If buying pullbacks is profitable, then buying pullbacks on very volatile stocks should be more profitable. This is an area that I have been meaning to explore further. I would think the concept could be applied to other strategies as well.
I have not yet focused on making the system more tradeable but it’s interesting what happens with removing a couple rules to allow for more trades and then the MktCap rule so now it’s really trading penny stocks. System went into a drawdown in Jan so I stopped it then just to see what the high was: 2570% annualized return, after Jan there was a significant drawdown to 1700% annual. http://www.portfolio123.com/port_summary.jsp?portid=266593
I don’t even like trading OTC and this is into penny stocks, but it does show that the system tends to be more profitable the smaller the stocks are.
Hi,
Regarding PriceH or PriceL, I am thinking that using weekly data over daily is not that big a deal, but I suppose a better variation might be: (Highest(#High,250,1) / Lowest(#Low,250,1)) > 2. I tested Circolo’s variation with this and it improved returns slightly. http://www.portfolio123.com/port_summary.jsp?portid=266798
I agree, the sim in my last post is entirely impractical … but it was fun.
After reviewing many of the trades in your Sim I found that a high % of the stocks were poor companies that were trending down. Although some of them were caught by the buy rules just right on a rebound for a very good gain, many of them just continued lower and lower for a big loss. The Transactions; All page shows that there were many, many stocks that were bought with very low rank values, sometimes in the teens.
I thought that there might be a better chance of a higher % of good stocks if the ranking system ranked stocks with some momentum factors higher than the poor stocks and therefore they would be selected much more often than the poor stocks. So I created a simple Ranking System with only Momentum factors and a liquidity filter: AvgDailyTot(20) > 100000 & Price > 1 & mktcap > 25.
I then re-ran Olikea’s Sim with this Ranking System as a 1 stock Sim so it would be more fully invested. The results were 667% annual return, 63% drawdown, 59% winners, and the magic 5.0% average gain.
Haha that sim had me in hysterics! 499 TRILLION, roflol.
However, I think you were wrong when you said the system is more profitable with the minnows, the profit per transaction was only about 4%, its just that because it had more to chose from it remained fully invested for longer, and if you can consistently maintain 4% every 3 days then your equity will go to the moon.
The problem is I can easily imagine these minnows having spreads in excess of 4%, so the gains shown in the sim are purely illusory.
However, it would be nice if there would be a way of slacking up the buy rules without sacrificing the profit per transaction, I am currently working on it.
I wanted to run your sim myself and play with it but I guess that is going to have to wait until tommorow, p123 seems to be a little busy right now!
I am not totally convinced a 1 stock port is a good idea, the drawdowns are potentially pretty large, and a single big drawdown could wipe out many years of gains, better to keep some in cash and bet a fixed % of your equity each time.
Circolo buy and hold investing isn’t dead yet! This could all yet be pie in the sky. I think of it as a little “spice” on the side, not a replacement
