Are there indicators that help spot breakouts?
- Thread starter crgarcia
- Start date
Nice try but you obviously lack even a basic understanding of backtesting... and trying to malign me won't change the fact that your method is BROKEN and simply doesn't test out. First, look at your own chart. Entries are supposed to be at the 0 to 7 turn (NOT 0 to 1) and exits are supposed to be at the 4 to 3 turn (NOT 1 to 0). But since entries outnumbered exits to the extent that trades would last YEARS , I realized your model is FATALLY FLAWED and instead tested buying at the 0 to 7 turn and exiting 1,2,3,4 and 5 days later. Five days worked best (of those choices) so that's what I used. Second, the AVERAGE trade was a loss of almost $20 (not $16) but that's an average. Do you know what average means? Some trades lost thousands and some gained. But the net effect was that your model is an unprofitable churning machine.
Quote from Trader666:
Nice try but you obviously lack even a basic understanding of backtesting... and trying to malign me won't change the fact that your method is BROKEN and simply doesn't test out. First, look at your own chart. Entries are supposed to be at the 0 to 7 turn (NOT 0 to 1) and exits are supposed to be at the 4 to 3 turn (NOT 1 to 0).
snip....
Sorry I wasn't clear.
In scoring there are three components that you back tested.
Price, the most significant
Volume, the leader of price and less significant
and A/D, Accumulation /Distribution an independent variable that addresses trader Sentiment.
The scoring is shown here:
the pic is a .doc so ET will not let it show as a pic.
I'm sure your backtesting included all three of these and you tested as the P, V and A/D turned from all 0's to all 1's to simulate the 0 score going to the 7 score, (an entry) where all scores are expressed in decimal numbers. similarly the exit is going from 4 to 3 where again all binary values change (100 to 011) The raw scoring, of course follows my algorithm requirement for certainty which, de facto, only allows binary coding so common to computer programming in its simplest state.
Your text shown above seems to be very strange and possibly indicates you do notknow what I have been talking about.
I'm sure that may be part of the picture when you fail to understand that I am asking a question about those wins and losses in a trade. I am asking the basis of a trade. I share, 100 shares?? If you know what I mean please let the world know. By stating thousands of dollars lost, you are IMPLYING more that one share per trade. If and when you answer, we all get to find out that the equity curve is just on a noise and statisitcally insignificant value. It is taking me a lot of posts to nail down what your equity curve isn't.
You may not have included A/D in your testing for the entry. You did not do an exit of 4 to 3 for the backtsting since some stocks just kept increasing in price (the 1 staying a 1) so those stocks had to be excluded by cutting off the trade at five days.
If you're talking about these, I can understand how you find simplicity and clarity freaky.
http://www.elitetrader.com/vb/attachment.php?s=&postid=2014868
http://www.elitetrader.com/vb/attachment.php?s=&postid=2014869
What I find freaky is this byzantine mess courtesy of your star student, spydertrader (attached).
http://www.elitetrader.com/vb/attachment.php?s=&postid=2014868
http://www.elitetrader.com/vb/attachment.php?s=&postid=2014869
What I find freaky is this byzantine mess courtesy of your star student, spydertrader (attached).
Well I find your questions very strange.
Of course my tests accounted for P, V, and A/D. You should have known that because 1) I told you I used Spydertrader's code and 2) a score of 7 is impossible without all three variables and a score of 3 is impossible without A/D.
As for your other questions, I did portfolio level backtesting (do you understand what that is), staring with $500,000, so the number of shares bought at any time was not constant... it depended on the equity available, price per share, etc. I don't recall the max % of equity I'd set for any single trade, but the majority of trades was made with more than 100 shares, some with several thousand. So you're barking up the wrong tree. The REAL problem is that your paradigm is fatally flawed.
Of course my tests accounted for P, V, and A/D. You should have known that because 1) I told you I used Spydertrader's code and 2) a score of 7 is impossible without all three variables and a score of 3 is impossible without A/D.
As for your other questions, I did portfolio level backtesting (do you understand what that is), staring with $500,000, so the number of shares bought at any time was not constant... it depended on the equity available, price per share, etc. I don't recall the max % of equity I'd set for any single trade, but the majority of trades was made with more than 100 shares, some with several thousand. So you're barking up the wrong tree. The REAL problem is that your paradigm is fatally flawed.
Let's see here, hmm, indicators for breakouts... Volume. If volume mounts but the overall bias is to the downside, you can expect to see a few prop traders test the strength of longs by creating a false breakout. The volume will indicate buying strength by virtue of it being higher than the previous high, but it's a ruse created by a few trading wicked high size moving the market. When they're done scalping their own created move their covering will create panic in the momentum traders realizing they've been sharked, which fuels a strong move down.
Guage the bias, watch volume and expect a few false starts. But in the false starts watch volume even more carefully.
As much as I hate to admit it the breakout due in the markets presently will likely be to the upside stronger than many here have seen. But watch volume.
Guage the bias, watch volume and expect a few false starts. But in the false starts watch volume even more carefully.
As much as I hate to admit it the breakout due in the markets presently will likely be to the upside stronger than many here have seen. But watch volume.
Thanks so much for your response.
You started with 500K and as you curve shows you lost money during three years and making 24,000 trades, you wound up with 100,000 dollars. The trading was done with a 100 share minimum and up to several thousand shares.
The band within which the equity curve fluctuates shows the range of levels of loss fluctuation and as the capital decreased by a factor of four, the portfolio total losses stayed fairly constant over the trading.
The equity curve based on time lost about 1/2 the total portfolio capital available each of the 3 years from 2002 to 2005, inclusive: 200K, 100K and 50K. The number of trades done each year could have changed or stayed flat according to the opportunities. I would guess the capital per trade changed as time passed and the number of trades were flat since a three year period was chosen.
Using the middle of the period and the capital at that time, the 32 trades on that day had an average of 1,000 bucks** worth of stock in the trade and as usual, your numbers, the trade lost 20 bucks by holding it 5 days. Closes (exits) accounted for 1/5th of losses and holds were declining and accounting for 4/5ths of the losses that day and as yet unrealized.
My conclusion, unlike yours, is that trading the 1,000 stocks you chose from the S&P 1500 for a constant five day hold using portfolio management and only buying on a day when all three, P, V and A/D shifted from 0's to 1's is not a good portfolio management technique (if you were testing, as you finally say, portfolio management).
I find buying stocks* on the day they shift from 0 to 7, (skip the portfolio management) and selling them when they go from 4 to 3 (when ever that happens) using streams of available capital (divide capital in equal streams) to average 3% profits reasized or unrealized Day. See "Putting the Pieces Together" as an example of such planned trading for one day.
My results, differ from yours based on using a trading technique instead of portfolio management; including in the technique a Universe selection based upon stock quality assessment and ranking of stocks; and automating as much of the FA/TA techique and Universe selection as possible (100% at this point).
For anyone who wishes to understand my comments, again see "Putting the Pieces Together"
*High beta only and with priority ranking for selection.
**On the middle day a year and one half has passed. 200K was lost the first year and I estimated at least half the 100K loss of the second year had occurred. the trading of 24,000 turns over 750 days amounts to about 32 trades a day to close trades, all held for five days. 1/5 of the portfolio is traded daily in the portfolio based backtesting. The exisitng 150,000 dollars divided first by 32 to get about 5,000 dollars represents two groups of held stocks: 1/5 to be sold that day and 4/5 still in holds going on for 5 days. So each stock held and traded on a given day is a portion of total capital. 5,000 dollars aggregate divided by 5 is 1,000 bucks. All of the stocks, traded or not, lose 4.00 dollars a day per group, each worth a 1,000 dollars at this point. A 20 buck lose on 1,000 dollars worth of stock occurs with each exit. All entries five day before wee on rising price and the intervening period does not show a score of 0 (binary) for price (or an exit would occur if monitored). Scoring is designed to preclude trading losses. In the backtest a timeout was substituted instead. This is not a good trading idea to incorporate in testing or trading but it was done. Nothing is proven with time out trading rules or testing. This represents a dilemma for trader666 and he is assigning the screwup to the fact that the P, V Relationship doesn't work. Others Granville and Dodd used it in the late 20's. Wycoff et al are others who used it in the next generation. Darvas used it in "box" trading in the 50's when I began to use it. This rejection by trader666 is monumental moslty because he only used half of the P, V relation to make his case for rejection. How he can use even portfolio management to prove buying stocks on increasing P, V and A of A/D and is five days time out, successfully lose 400,000 dollars of 500,000 dollars in three years (and never have a gaining period even) is not a repeatable backtest in my opinion. It cetainly is was not for me or anyone else in the literature from the 20's forward. The opposite is generally true.
Again thank you for posting a chart for 21JUL08 that is annotated.
This is a five minute chart not a 1 minute chart nor a 1 sec chart such as you posted in the PA journal recently.
I'll go through 4 annotation of this chart forn for successive levels of beginner (2) and intermediate (2).
The beginner trades from FTT to FTT on channels. This was a trending DAY as we all see from the outer envelope of price movement all day long.
Coarse trading of beginers is infrequent but they are at work on the byzantine mess since they use the simplest system of cases of adjacent bars to build the nesting trends of tapes to get traverses to get channels which they trade.
Trends of any sort overlap.
We also get to see the relative importance of "3BR's" and "matching lows" and "matching highs"
These are the seven cases the beginner uses on adjacent bars.
All of pool extraction is based upon these cases. It applies to PVT, SCT and sector rotation.
Certainty in these cases is what eliminates "probability", as a necessity, for all trading.
Also notice the degree of complexty does not include the 3BR nor the 2 "matching" bottoms or tops
All of pool extraction is based upon these cases. It applies to PVT, SCT and sector rotation.
Certainty in these cases is what eliminates "probability", as a necessity, for all trading.
Also notice the degree of complexty does not include the 3BR nor the 2 "matching" bottoms or tops