NexusFi: Find Your Edge


Home Menu

 





Probability 101


Discussion in Traders Hideout

Updated
      Top Posters
    1. looks_one Lornz with 23 posts (49 thanks)
    2. looks_two monpere with 11 posts (32 thanks)
    3. looks_3 Fat Tails with 10 posts (21 thanks)
    4. looks_4 trendisyourfriend with 4 posts (1 thanks)
      Best Posters
    1. looks_one monpere with 2.9 thanks per post
    2. looks_two Lornz with 2.1 thanks per post
    3. looks_3 Fat Tails with 2.1 thanks per post
    4. looks_4 Hotch with 1.3 thanks per post
    1. trending_up 39,036 views
    2. thumb_up 114 thanks given
    3. group 26 followers
    1. forum 67 posts
    2. attach_file 6 attachments




 
Search this Thread

Probability 101

  #61 (permalink)
 
bobwest's Avatar
 bobwest 
Western Florida
Site Moderator
 
Experience: Advanced
Platform: Sierra Chart
Trading: ES, YM
Frequency: Several times daily
Duration: Minutes
Posts: 8,162 since Jan 2013
Thanks Given: 57,341
Thanks Received: 26,267


Fat Tails View Post
The simplest statement containing a paradox caused by self-reference is the statement known as the liar paradox:

"This sentence is false."

If it is false it must be true, and if it is true it must be false. No way out of this.

Well, the way that logicians get out of this is to declare that you just can't do self-reference. I've always thought that was a neat way out of the trap: make it against the rules.

Bob.

Reply With Quote
Thanked by:

Can you help answer these questions
from other members on NexusFi?
Are there any eval firms that allow you to sink to your …
Traders Hideout
ZombieSqueeze
Platforms and Indicators
NT7 Indicator Script Troubleshooting - Camarilla Pivots
NinjaTrader
My NT8 Volume Profile Split by Asian/Euro/Open
NinjaTrader
Futures True Range Report
The Elite Circle
 
Best Threads (Most Thanked)
in the last 7 days on NexusFi
Get funded firms 2023/2024 - Any recommendations or word …
61 thanks
Funded Trader platforms
38 thanks
NexusFi site changelog and issues/problem reporting
27 thanks
GFIs1 1 DAX trade per day journal
19 thanks
The Program
18 thanks
  #62 (permalink)
 
Fat Tails's Avatar
 Fat Tails 
Berlin, Europe
Market Wizard
 
Experience: Advanced
Platform: NinjaTrader, MultiCharts
Broker: Interactive Brokers
Trading: Keyboard
Posts: 9,888 since Mar 2010
Thanks Given: 4,242
Thanks Received: 27,102


danielk View Post
Exactly! A classic example of a question i'd normally simply reply to with "dude, your question sucks!"

However would you agree that changing answer C. would be irrelevant? an all round 'false' result is false, regardless of it being 60%, 0% or 250%

You could change option A to 0%.

Reply With Quote
Thanked by:
  #63 (permalink)
miktrading
Los Angeles, California
 
Posts: 24 since Mar 2017
Thanks Given: 4
Thanks Received: 8


Hey everyone. I read all the pages about statistics and probabilities, yet can't find the information I'm looking for.

How to use that information to develop a strategy that has at least a 1:2 risk/reward and large enough probabilities to be profitable.

I've traded based on multiple timeframes with moving average crossovers and macd. Many say these are lagging indicators. You should be entering at the source of a move. That's great but one needs statistics that will keep you from entering in the middle of no where without an "edge".

Recently I began using Linnsoft Investor RT program. It seems great for Volume Profiling and Delta, but not of much use since I haven't traded with that information before.

If think in reverse I know that an ideal day would have 2-4 trades strictly intraday. Each trade should not risk more than 0.25% of the account, therefore with 4 trades the maximum daily risk is 1%. Risk/Reward should be 1:2. This criteria is too vague though. What's an efficient way to find entries have an edge that pay off in the long run?

Accounting for average daily range, 10 day range, and current range could give some clues to expectations. Relative volume might help in regards to range expansion?

With a 100k account the risk parameters are $250 risk per trade and $1,000 max daily risk. That's 5 points in the E-mini SPX and 25 ticks in Crude oil.

Yes, it's also known that one traders system might not work for another trader, but that's not the point. The point is in finding a system and its statistics.

Because I'm sure many would like to know this, has anyone already done such analysis and is able to trade based on that analysis, achieving similar results as tested after at least 1 year?

If something has already been done, it would save many from trying the same with trial and error which takes even more time.

I appreciate anyone's help in answering this.

Reply With Quote
  #64 (permalink)
 tpredictor 
North Carolina
 
Experience: Beginner
Platform: NinjaTrader, Tradestation
Trading: es
Posts: 644 since Nov 2011

@miktrading This is typically referred to as backtesting or simulation. The simulations or backtest produce the statistics. Most statistics are based on the normal distribution which is found in nature. However, markets are not normally distributed. This is just one reason backtesting doesn't always work because the statistical properties of the market can change, called non-stationarity. However, backtesting does allow you to see whether a trading method would have been profitable in the past. If you are trading on the idea that a pattern produces a profit statistically in the market then eliminating the false belief is beneficial. As for "finding entries efficiently", one possibility would be to try the generative method for system development. It is, also, possible to simply optimize the stop and target and review the results.

Reply With Quote
  #65 (permalink)
 iantg 
charlotte nc
 
Experience: Advanced
Platform: My Own System
Broker: Optimus
Trading: Emini (ES, YM, NQ, ect.)
Posts: 408 since Jan 2015
Thanks Given: 90
Thanks Received: 1,147


miktrading View Post
Hey everyone. I read all the pages about statistics and probabilities, yet can't find the information I'm looking for.

How to use that information to develop a strategy that has at least a 1:2 risk/reward and large enough probabilities to be profitable.

.


The statistics you are looking for to solve for this are surprisingly simple. Here is a little background information.

1. The main statistics you will be using will be quantifying the ranges of the market. You can do this a variety of ways, Avg range, STD, but I think the easist to understand is just the Max(sample size) - Min(sample size). If you are sampling 10 minutes, or 2000 ticks, or whatever... it doesn't matter, just keep one unit of measure constant throughout the statistical analysis. What you are trying to define is what the range of movement is typically. Then you can classify this statistically like this: *** I just made up the 25% splits, you may find in practice this skews more towards one of these categories than the others.**

A: 25% of the time the market moves inside of this range (Low Volatility)
B: 25% of the time market moves in a wider range than A, but still inside of this range (Medium Volatility)
C: 25% of the time the market moves in a wider range and A and B. (High Volatility)
D: 25% of the time the market moves in a range greater than A-C. (Crazy High Volatility)

2. The second part you will need is some statistics on direction. This is even easier to quantify. You just need some flavor of is the market moving up or down,or is the movement so small it may as well be flat. This can be measured, daily, hourly, weekly, it doesn't matter. But ultimately what you need is something like this:

A: Trending up %
B: Trending down %
C: Flat %

So let's say that after testing a particular instrument on a specific time series across several months, quarters, years, etc, you determine that for this instrument you have part 1 defined as:

A: Low volatility = 10 ticks or <
B: Medium Volatility = 15 ticks up to 20 ticks
C: High volatility = 20 ticks up to 30
D: Crazy High volatility > 30 ticks

You also observe that the market is evenly split between long and short and flat, and there is no real long term bias. (It's a flip a coin type of market on direction)

With this information and your requirement that you want a Risk / Reward ratio of 1 x 2 or > Here is the market condition and bet you will be making.

When the market is in cycle C (High Volatility 20 to 30 ticks) or D (Crazy High Volatility > 30 ticks) You will set your profit target at 1.5 to 2x your stop loss. Specifically you will employ the following settings:

High Volatility periods: (The market moves inside of a 20-30 tick range). Set your profit target somewhere between 20 to 30 ticks and set your stop loss somewhere between 10 to 15 ticks. Here is why this will work. Volatility periods unlike directional trends, tend to last longer, so even if you pick the wrong direction, the odds of the market moving up and down in 20 to 30 tick moves, will be = to the odds of the market moving in 10 to 15 tick moves. Right or wrong on direction you will hit roughly the same amount of winners as losers, but your winning value will be 2x your losing value. *** It is important to realize why this works. It only works like this in high volatility periods when the market is just as like to make 20 tick moves as it is 10 tick moves. We saw several weeks of this already this years for days straight.

By Contrast when the market contracts down to low volatility you need to employee the exact opposite strategy. Use a Risk / Reward of 2x1. In this example if your analysis tells you that 10 ticks is the typical range in low volatility periods, then set your stop loss at 12 ticks and your profit target at 6 ticks. Even if you pick the direction wrong, you will likely see the market bounce up and down inside of the 10 tick range and never hit your SL. So yes the value of your SL is higher, but the odds of you hitting it are very low because you did your homework, studied your ranges, and know historically the behavior of your particular instrument on your favorite settings.

So while I will stop short of telling you how to do this type of analysis work, I can tell you that using statistics, this is exactly one type of model that you can easily solve for.

There are tons of other ways to apply statistical probabilities to trading, but this is one of my favorite use cases.

Best of luck!

Ian

In the analytical world there is no such thing as art, there is only the science you know and the science you don't know. Characterizing the science you don't know as "art" is a fools game.
Visit my NexusFi Trade Journal Reply With Quote
Thanked by:
  #66 (permalink)
miktrading
Los Angeles, California
 
Posts: 24 since Mar 2017
Thanks Given: 4
Thanks Received: 8


iantg View Post
The statistics you are looking for to solve for this are surprisingly simple. Here is a little background information.

1. The main statistics you will be using will be quantifying the ranges of the market. You can do this a variety of ways, Avg range, STD, but I think the easist to understand is just the Max(sample size) - Min(sample size). If you are sampling 10 minutes, or 2000 ticks, or whatever... it doesn't matter, just keep one unit of measure constant throughout the statistical analysis. What you are trying to define is what the range of movement is typically.

If its daytrading, wouldn't I want to measure the days range then? I didn't understand the max-min sample size. Say I wanted to go back 2,000 days of trading. What did you mean by 10 minutes or 2,000 ticks? The range of movement within those 10 minutes or every 2,000 ticks? Something new to me with Investor RT was eliminating time frames and looking at 8 tick volume profile renko charts for example.


"Volatility periods unlike directional trends, tend to last longer, so even if you pick the wrong direction, the odds of the market moving up and down in 20 to 30 tick moves, will be = to the odds of the market moving in 10 to 15 tick moves."

I understand why that's true, but is there a way to test whether or not volatility period last longer, and if so, by how much longer than directional trends?

Thanks again. These are the specific answers I've been looking for.

Reply With Quote
  #67 (permalink)
 iantg 
charlotte nc
 
Experience: Advanced
Platform: My Own System
Broker: Optimus
Trading: Emini (ES, YM, NQ, ect.)
Posts: 408 since Jan 2015
Thanks Given: 90
Thanks Received: 1,147

This is a great question. There are several ways to go about it, but what you are solving for is to find the following:

1. A range that the market operates in for a reasonable enough time frame to establish several bets, but not too long. Your measuring stick should be anywhere from a few minutes to a few hours, because in practice if you are looking at longer time frames then you will end up with less actionable data. Here is why: If you look at the full range for an entire day it might be 100, 200, 300 ticks... So how would you use that in the context of what we are discussing? There is really only one bet you could make in the context of this type of strategy and that is to place an enormous stop loss beyond this range and put a profit target inside this range. There are a whole host of traders that literally don't use stops and just get hit on the daily close, and they often crush it using this type of principle, because they hit 10,20,50 targets daily and only get stopped out once every few days and eat a giant loss. But it is hell of risky playing this game and what we are referring to here is way more scientific and nuanced than this.

2. You want to find a range that suits a reasonable betting strategy where you can take 5, 10, 20, 50 trades per day against these statistics. So you need a range that will allow you to be able to enter and exit enough times. So this implies you need to be measuring smaller time frames vs. an entire day.

3. You want to be able to see a volatility range establish itself and hold for a fairly long period of time, but you also need be able to observe subtle changes so you can build the type of classification I referred to. (Low, Medium, High, Very High) etc.

By sample size what I mean is this. Let's take the ES for example. I trade this instrument, so I can give you so real life examples:

if you collect a sample size of 25 ticks for example, you won't really capture that much delineation. Every 25 tick bar you sample will likely have a range of between 1 tick and 2 ticks. It just takes more trades to move the ES. So this is too small of a sample size. Conversely if you use a sample size of 10 hours, then the range distribution will be all over the place and often times so high that you won't be able to make any bet other than the extreme case I mentioned before. What you want is a range closer to a few hundred ticks or several minutes. So if you get a data dump of every HLOC for each 5 minute bar and then take a 20 period Max - a 20 period Min of each of these bars, you would get a pretty workable range. For the ES I think you could see something like this.

For every 20 period Max(5 minute bar) - Min(5 Minute bar) you might see something like

Low Volatility: The market has moved less than 10 ticks in total over this time period
Medium Volatility: The market has moved less than 20 ticks in total over this time period
High Volatility: The market has moved 20 to 30 ticks over this period
Very High Volatility: The market has moved more than 30 ticks over this period

Now for this to work, you need to quantify that these ranges typically hold for a fairly decent period of time before they change to the next cycle. So if your cycles are moving like this:

Low Volatility (1 minute) > Medium Volatility (30 seconds) > High Volatility (1 minute) Then you are likely too zoomed in and may have your definitions a little off.

If you have your cycles flipping every 2 to 3 hours, then you are likely too zoomed out and you are not capturing any delineation and the only option you will have betting wise, is to just take a ridiculously high stop outside of the range which I am not advising.

What you want are cycles that hold for anywhere from a few minutes, to maybe an hour. Depending on how you sample the market, and what your thresholds are in your classification system you can find this fairly easily. But the key is to make sure that you can size this to have both a reasonable bet size to match your trading needs, as well as to make sure the flip between cycles gives you enough time to fire off a few bets. Because even if you pick the direction wrong, you can still win by just using the right risk / reward in the right volatility cycle. The ideal scenario is to be able to fire off 5, 10 bets per cycle and if you are sizing things right, you will get a 50% / 50% split on picking direction right, but you will get 2 to 1 odds on the volatility cycle sizing if you know how to play this game.

Hope this explains it a little better. It's one of the lesser known, but more statistical edges out there if you are willing to put in the work.

Best of luck

Ian




miktrading View Post
If its daytrading, wouldn't I want to measure the days range then? I didn't understand the max-min sample size. Say I wanted to go back 2,000 days of trading. What did you mean by 10 minutes or 2,000 ticks? The range of movement within those 10 minutes or every 2,000 ticks? Something new to me with Investor RT was eliminating time frames and looking at 8 tick volume profile renko charts for example.


"Volatility periods unlike directional trends, tend to last longer, so even if you pick the wrong direction, the odds of the market moving up and down in 20 to 30 tick moves, will be = to the odds of the market moving in 10 to 15 tick moves."

I understand why that's true, but is there a way to test whether or not volatility period last longer, and if so, by how much longer than directional trends?

Thanks again. These are the specific answers I've been looking for.


In the analytical world there is no such thing as art, there is only the science you know and the science you don't know. Characterizing the science you don't know as "art" is a fools game.
Visit my NexusFi Trade Journal Reply With Quote
Thanked by:
  #68 (permalink)
miktrading
Los Angeles, California
 
Posts: 24 since Mar 2017
Thanks Given: 4
Thanks Received: 8

Here's something interesting that Chad put together for me using Investor RT.

It's taking current volume relative to average of last 10 days... (bottom pane).

I saw it from a trader that trades crude oil. He takes the average 10 day range and then compares it to the current relative volume for the past 10 days. If rvol is up 30%, then you could expect range expansion 30% over the average range.

Brett Steenbarger talks about using volume, relative volume, and volatility.

TraderFeed: [AUTOLINK]Relative Volume[/AUTOLINK] and [AUTOLINK]Volatility[/AUTOLINK]: Understanding Who is in the Market

TraderFeed: The Importance of [AUTOLINK]Relative Volume[/AUTOLINK]: Is It a Slow Day or a Go Day?


Most of the time less is more, but I think the quality of tools is even more important. Has this been valuable to anyone?


Reply With Quote




Last Updated on May 30, 2018


© 2024 NexusFi™, s.a., All Rights Reserved.
Av Ricardo J. Alfaro, Century Tower, Panama City, Panama, Ph: +507 833-9432 (Panama and Intl), +1 888-312-3001 (USA and Canada)
All information is for educational use only and is not investment advice. There is a substantial risk of loss in trading commodity futures, stocks, options and foreign exchange products. Past performance is not indicative of future results.
About Us - Contact Us - Site Rules, Acceptable Use, and Terms and Conditions - Privacy Policy - Downloads - Top
no new posts