StockFetcher Forums · Filter Exchange · A different way of looking at the Guppy Multiple Moving Averages<< 1 ... 2 3 4 5 6 >>Post Follow-up
xarlor
553 posts
msg #155061
Ignore xarlor
12/22/2020 10:00:16 PM

I risk 2% of my account value on a single trade. I never risk more than 5% of my account value on a given ticker. For example, if I risk 2% on QQQ, but then see a set up that warrants adding to the position, the total amount risked can never exceed 5%.

If I am trying out a new strategy, I keep the amount risked under 1%.

Successful and smart people suggest using the Kelly Criterion when making trades. To use it, you need to know how successful a particular strategy is (e.g. wins 70% of the time).

Cheese
1,374 posts
msg #155062
Ignore Cheese
12/22/2020 10:27:34 PM

@xarlor,

THANK YOU for a very clear explanation.

I've tried paper trading some of the main bull market ETFs with Kelly Criterion
but somehow that formula didn't click with me, so I gave up on Kelly Criterion.



fotchstecker
304 posts
msg #155083
Ignore fotchstecker
modified
12/24/2020 3:39:33 PM

Using a flat percent of equity gets a bit tricky when target vol of the instruments varies. A true "2% risk" could be helped with some historical measure.

I think it makes more sense to use a dollar-vol-adjusted equity size.

Kelly is not overly complicated, but few traders I know implement Kelly as it was "on paper." Along with Optimal F, Kelly is easier to stomach when used fractionally. Google "Fractional Kelly or Fractional Optimal F sizing".

I've experimented a lot with sizing, so if there are questions, fire away. You can also trial Adaptrade MSA, which is a faster way to model the impact of sizing on your trade history or model.

These days I do something completely different -- dynamic sizing based on inputs that emerge as the system is running.



Cheese
1,374 posts
msg #155085
Ignore Cheese
12/24/2020 5:33:15 PM

THANK YOU, fotch !
and Merry Christmas to you and all members who celebrate.

I've been reviewing and re-learning money management and sizing.
You are the second person who mentioned Adaptrade Market System Analyzer (MSA), so I may give it a try. The "different thing" that you mentioned is probably Adaptrade Builder ?

I will Goggle "Fractional Kelly or Fractional Optimal F sizing" and review my past attempts at Kelly.

I will certainly ask for help if I need clarification.

THANKS AGAIN, fotch!
Have a Happy New Year.

fotchstecker
304 posts
msg #155088
Ignore fotchstecker
12/24/2020 7:17:07 PM

Hi, Cheese! Happy holidays!

No, the "new sizing" I use is not derived from MSA. MSA I use with historical trades to model differnt sizing methods. There is a feature for sizing the "next real-world" trade in MSA, which I used in the past. However, it's a somewhat manual process.

The "new sizing" I mentioned is different from any sizing method in MSA and uses a logic of its own. It's running automatically during the day on intraday bars on some algos I built. Categorically, it's more close to equity curve trading (which you can also model in MSA). In equity curve trading, the system performance itself is an input to next size. I'm doing this in a different environment completely, not daily bars and filters in SF.

fotchstecker
304 posts
msg #155089
Ignore fotchstecker
12/24/2020 7:19:32 PM

Oh, Cheese, one other thing -- I like MSA and wouldn't want to take anything away from it. That said, there is an excel-based sizing spreadsheet product that is less expensive. If I can find it, I will link it. It's not as integrated or good as MSA, but it does do a few different sizing types to give you an idea and for an excel product it's pretty good. I'll see I can find it. Too bad SF doesn't have messaging; this kind of thing would be easier to update...

Cheese
1,374 posts
msg #155092
Ignore Cheese
12/24/2020 9:27:37 PM

THANK YOU very much, fotch. Very kind of you.

and THANK YOU, graf for letting me borrow your thread for money management position sizing.

fotch, I will take advantage of your kind offer to help answer questions, and start with the first one here.

The same filter with the same formulas would tell me to do sightly different things,
with different ATR lengths ATR(14, ATR(21), ATR(63), ATR(252), etc
with different stocks SPY, QQQ, XLE, etc
and in different time periods, e.g. price advances or declines in different periods
(different short-term phases of the markets)

What I got was the position sizing based on ATR would tell me to do different things

* Sometimes, the ATR approach would tell me to reduce the number of shares when prices go up
and increase the number of shares when prices go down.

* But other times, he ATR approach tells me to do the opposite.

What are your insights from your experience with position sizing?

Thank you in advance to you, fotch and to all members for your inputs.


fotchstecker
304 posts
msg #155095
Ignore fotchstecker
modified
12/25/2020 12:55:30 PM

Cheese, I'll say some things you probably already know but just so we're on the same page:

All any sizing method does is methodically preserve capital in a way that is ideally proportional to the expectancy of the system itself. So, the size of stop (adverse move limit) is the direct lever you can pull for risk in general. However, stop limit size is also related to trade frequency. If you are stopped out frequently (say by design), you are implying that your system has positive expectancy over a longer series with more frequency (since some average winners will need to outweigh your losers by a margin that creates a positive return over time).

Without directly seeing what you're running into with ATR, what you mention makes me think of a couple of things off the bat. I didn't really have an appreciation for the sensitivity of things like ATR or historical vol among different tickers until I really started backtesting/WFO in earnest. In a different environment, I will optimize the exact same system on various tickers and come up with very different lookback periods for the same relative performance. If you initially construct a basket using correlation and beta, you can come up with a basket that suits your trading horizon with relatively differentiated stocks that still have some feature in common, like "high volatility." So, you can find 10 symbols of "high volatility" but due to the underlying market dynamics, they will be in various states of actual price dynamics for the period.

I wouldn't have had an appreciation for just how different symbols are, and how they are responsive to the parameters in a trading system, without running simulations. This same "eye opener" undid any rule-of-thumb values for many of the indicators and measures we have all come across for years (or even ship as "standard" with an indicator), like often-stated rule of "risking 1% equity" or using ATR14 (14 period atr), or 20-2 or bollinger (20 period, 2 standard dv). Backtesting showed me that finding the risk level and trade timeframe were things I needed to find to match my style, weird to say. I know we all know that parameters can be altered, and we all do, but I didn't have really an idea of HOW dynamic things are.

Earlier this year I had the goal of taking on a equal risk-adjusted position across the portfolio for any new buy. This required everything to be vol-normalized, then purchased per unit of vol by converting to dollars. This is a lot like a risk parity approach and in reality, it isn't really right for a retailer of my size. Real funds do things like vol carry trades for periods of instability (like futures curve trading) or trading the spread between the index and single tickers -- like arbitrage for then index vol is high but single ticker vol is low.

ATR-based position sizing just lets you calculate how much to put into a position if you know the a. volatility you will accept and b. the equity pool value. But without some testing/measurement, it won't tell you if your stop makes sense or if your system has positive expectancy. Say you have a 100K equity pool and want to target a 1.75% vol allocation, then your size is $1,750. In the same example, if you're using the 10-period ATR to define vol and ATR10 is $2.25 and you wanted to know how much you can buy to risk at your target allocation and at your target vol: 1750/2.25 = 777. So you can see if you're trading a basket and want to take on similar dollar risk across a portfolio, you have the levers to pull.

The above is why at some point a lightbulb went on and I realized that any "rule of thumb" was probably not very helpful. The elements you need in the background are positive expectancy in your approach and then an idea of how much risk you can take on and still have positive expectancy. Then you know what your own "1% of equity" size really would be...

Vol-based stop are just one of the many ways people stop risk, obviously. However, how one "limits risk" in a system is one of the features needed for optimizing position sizing in general. You can't really optimize how much you can risk if the method for stopping isn't math, something one can understand. Risk defines profit and losses, which if central to expectancy.

In the words of Ralph Vince, "A negative expectation set of data points has no optimal fraction to bet." That was in "Optimal f and the Kelly Criterion" in the IFTA Journal -- I have it around somewhere.

Sorry for the rambling here!



Cheese
1,374 posts
msg #155098
Ignore Cheese
12/25/2020 3:34:53 PM

THANK YOU very much, fotch.
I really appreciate your help.

Your posts on money management are the most extensive,
and I'm sure you will be helping a lot more people than just me.

Please feel free to continue to teach me with what you know about this stuff,
and you seem to know a lot.

There is so much I want to un-learn and re-learn.
I've Googled the "Fractional Kelly or Fractional Optimal F sizing" that you recommended.

The new material helps me get more insights from my old learning,
A study framework is emerging and taking shape, so something good will come out of this.
All this will take me some time through the Holy Day Season.

THANKS AGAIN, FOTCH for your time and for sharing.
.



graftonian
1,089 posts
msg #155454
Ignore graftonian
1/26/2021 4:15:58 PM

Need some help with a "pure" Guppy MMA exit
Fetcher[
/*************** A PURE GUPPY FILTER Jan 26,2021******************/
/*Buy when trader group is completely aligned (or stacked right as I think of it),
the investor group is not, and the traders cross above the investors
Sorting is by the width of "white space" between the 2 groups
Exit stratagy: I would like to use Daryl Guppy's three day count-back, do not know how to code it */
chart-time is 4 months
not otcbb
average volume(30) > 1234567
close > .50
/*Standard Guppy Averages*/
draw ema(3) draw ema(5) draw ema(7) draw ema(9) draw ema(11) draw ema(13) draw ema(15)
draw ema(30) draw ema(35) draw ema(40) draw ema(45) draw ema(50) draw ema(55) draw ema(60)

/*Identify when each pair of MAs cross and sum the crosses */
set{3x5, count(ema(3) > ema(5), 1)} draw 3x5 set{cc1, 3x5}
set{5x7, count(ema(5) > ema(7), 1)} draw 5x7 set{cc2, cc1 + 5x7}
set{7x9, count(ema(7) > ema(9), 1)} draw 7x9 set{cc3, cc2 + 7x9}
set{9x11, count(ema(9) > ema(11), 1)} draw 9x11 set{cc4, cc3 + 9x11}
set{11x13, count(ema(11) > ema(13), 1)} draw 11x13 set{cc5, cc4 + 11x13}
set{13x15, count(ema(13) > ema(15), 1)} draw 13x15 set{cc6, cc5 + 13x15}
set{traXinv, count(ema(15) > ema(30), 1)} traXinv > 0 plottype{traxinv, zerobar} draw traXinv/*15x30*/
set{cc6a, cc6 + traXinv} set{30x35, count(ema(30) > ema(35), 1)} draw 30x35
set{cc7, cc6a + 30x35} draw 30x35 set{35x40, count(ema(35) > ema(40), 1)}
draw 35x40 set{40x45, count(ema(40) > ema(45), 1)} draw 40x45
set{cc8, cc7 + 40x45} set{45x50, count(ema(45) > ema(50), 1)} draw 45x50
set{cc9, cc8 + 45x50} set{50x55, count(ema(50) > ema(55), 1)} draw 50x55
set{cc10, cc9 + 50x55} set{55x60, count(ema(55) > ema(60), 1)} draw 55x60
set{CrossCount, cc10 + 55x60}
add column CrossCount

/****************** Entry criteria ************/
3x5 > 0 and 5x7 > 0 and 7x9 > 0 and 9x11 > 0 and 11x13 > 0 and 13x15 > 0
traXinv crossed above .5 add column traxinv
and 30x35 < .5
and 35x40 < .5
and 40x45 < .5
and 45x50 < .5
and 50x55 < .5
and 55x60 < .5


PlotType{CrossCount, line}
draw CrossCount draw Crosscount line at 7
set{ccAvg, cema(CrossCount, 2)}
draw ccAvg on plot CrossCount
/****************** sort criterion ********************/
/***How far apart are ema15 and ema30? The so called "white space" ***/
set{15x30Spread1, ema(15) - ema(30)}
set{15x30Spread2, 15x30Spread1 / ema(15)}
set{15x30SpreadPCt, 15x30Spread2 * 100}
add column 15x30SpreadPct{White Space %}
sort on column 7 descending
]



StockFetcher Forums · Filter Exchange · A different way of looking at the Guppy Multiple Moving Averages<< 1 ... 2 3 4 5 6 >>Post Follow-up

*** Disclaimer *** StockFetcher.com does not endorse or suggest any of the securities which are returned in any of the searches or filters. They are provided purely for informational and research purposes. StockFetcher.com does not recommend particular securities. StockFetcher.com, Vestyl Software, L.L.C. and involved content providers shall not be liable for any errors or delays in the content, or for any actions taken based on the content.


Copyright 2022 - Vestyl Software L.L.C.Terms of Service | License | Questions or comments? Contact Us
EOD Data sources: DDFPlus & CSI Data Quotes delayed during active market hours. Delay times are at least 15 mins for NASDAQ, 20 mins for NYSE and Amex. Delayed intraday data provided by DDFPlus


This site is protected by reCAPTCHA and the Google Privacy Policy and Terms of Service apply.