Statistical proof of the 90% rule | Page 2 | Forex Factory

Options

Go to
Permalinks
Print Thread

Similar Threads

FX statistical arbitrage 11 replies

Statistical Probability Trading With Price Action 576 replies

Statistical arbitrage, hedge, or bluffing myself? 16 replies

Statistical Validation on trading 24 replies

Statistical Report #2 on EAs 1 reply

Attachments: Statistical proof of the 90% rule
Exit Attachments

Statistical proof of the 90% rule

  • #21
  • Jun 27, 2017 12:13pm Jun 27, 2017 12:13pm
Quoting MoneyZilla
{quote} the cubic system decreases the forward movement by 50%. tp10 / sl10 ~ 50% / 50% tp20 / sl10 ~ 25% / 75% tp30 / sl10 ~ 13% / 87% tp40 / sl10 ~ 7% / 93% tp50 / sl10 ~ 4% / 96% tp60 / sl10 ~ 2% / 98% tp70 / sl10 ~ 1% / 99% i am sorry to disappoint you, but i have scanned 25 pairs for the last 8 years and they all show similar data. it does not matter how big the cube is. everything is proportional. the base cube size really does not matter... EDIT: also, the pair does not matter. everything is the same on all pairs. many will disagree with...
Thank you. Do you base your data off of movement from a position taken by a system or just general forward/backward movement of the market
  • #22
  • Edited 8:14am Jun 28, 2017 7:14am | Edited 8:14am
Quoting Piphunter00
{quote} Thank you. Do you base your data off of movement from a position taken by a system or just general forward/backward movement of the market
Every chart explicitly shows what is allowed. I am measuring what is allowed only. Cannot be specific, as I prefer to keep this for myself only, but if you can trust me a little? They are limiting the RR ratios. Like manipulating them is more properly said... in their one and only favor.

For example tp70 / sl10 should has 12.50% / 87.50% probability. In reality, on the chart, the probability is 1.00% / 99.00%. The chance of entering anywhere on the chart and hitting 70 pips tp vs 10 pips sl is around 1%. The change the sl10 to get hit is 99%. Do this 1,000 times on a pair and you will see what I mean.

There is a genius mind behind this cubic system. Whoever figured it out, long time ago, when there were no computers, is an absolute genius for me...
Maadik Hugiis. IQ 69.
  • #23
  • Jun 28, 2017 7:19pm Jun 28, 2017 7:19pm
So it sounds like we can devise a strategy doing the opposite of what Oanda clients are doing and make money?
  • #24
  • Jun 29, 2017 12:40am Jun 29, 2017 12:40am
Quoting MoneyZilla
{quote} Every chart explicitly shows what is allowed. I am measuring what is allowed only. Cannot be specific, as I prefer to keep this for myself only, but if you can trust me a little? They are limiting the RR ratios. Like manipulating them is more properly said... in their one and only favor. For example tp70 / sl10 should has 12.50% / 87.50% probability. In reality, on the chart, the probability is 1.00% / 99.00%. The chance of entering anywhere on the chart and hitting 70 pips tp vs 10 pips sl is around 1%. The change the sl10 to get hit is...
Thanks again. This is what I presumed it was. Movement from any given point on the chart. It makes sense.

I'll be taking a look into it.
  • #25
  • Edited 1:13pm Jul 22, 2017 12:50pm | Edited 1:13pm
Quoting MoneyZilla
{quote} the cubic system decreases the forward movement by 50%. tp10 / sl10 ~ 50% / 50% tp20 / sl10 ~ 25% / 75% tp30 / sl10 ~ 13% / 87% tp40 / sl10 ~ 7% / 93% tp50 / sl10 ~ 4% / 96% tp60 / sl10 ~ 2% / 98% tp70 / sl10 ~ 1% / 99% i am sorry to disappoint you, but i have scanned 25 pairs for the last 8 years and they all show similar data. it does not matter how big the cube is. everything is proportional. the base cube size really does not matter... EDIT: also, the pair does not matter. everything is the same on all pairs. many will disagree with...
Hi, MoneyZilla, I got interested in it.
So coded a program to get these stats. All stats are with real tick data from dukascopy broker. I did tests in c#.
I wanted to do the stats with that algo: On every new tick open buy and sell orders with the same sl and tp. But I didn't do that because it gets too long to calculate and too much data and can't export to excel. So to speed up calculations I ended with that algo: On every new minute, I open buy and sell orders with the same SL and TP levels.
So with the same SL and TP targets (10p and 10p) I got 50%. In these stats I don't add spread costs.
With tp10p and sl70p I got 12.74% for TP and 87.26% for SL. All trades are ~ 1 million. I think it's enough for stats. I can get more and add to SQL server but I don't think it'll change the results. So I don't get these cubic system results. And I am pretty sure, if I did with tick orders I would have get similar results. Is my algo wrong to get that?
Quote
The chance of entering anywhere on the chart and hitting 70 pips tp vs 10 pips sl is around 1%.
I can't get these stats.
Thanks, IenDzi.
  • #26
  • Jul 22, 2017 1:13pm Jul 22, 2017 1:13pm
I got stats for 1p TP and 100p SL and to win 1p it is 98.8% ~ 99%. I had similar results on tick data and traded several years ago with high leverage and getting 2p TP per trade. But it was without edge.
  • #27
  • Jul 23, 2017 11:51am Jul 23, 2017 11:51am
Quoting MoneyZilla
{quote} Every chart explicitly shows what is allowed. I am measuring what is allowed only. Cannot be specific, as I prefer to keep this for myself only, but if you can trust me a little? They are limiting the RR ratios. Like manipulating them is more properly said... in their one and only favor. For example tp70 / sl10 should has 12.50% / 87.50% probability. In reality, on the chart, the probability is 1.00% / 99.00%. The chance of entering anywhere on the chart and hitting 70 pips tp vs 10 pips sl is around 1%. The change the sl10 to get hit is...
Very powerfull truth few people understand; so nice to find someone that have real experience and doesnt "copy/paste" the same crap over and over again

Its easier to hit a "small" Pips target from a given entry point than a larger one; over time you stack the odds AGAINST your trade and your "edge" (if you had any) start to vanish.

"Fixed Risk:reward" its a dumb concept from the beggining

but dont say that over here, they will call you noob, you dont have experience, etc etc..
  • #28
  • Jul 23, 2017 4:28pm Jul 23, 2017 4:28pm
Quoting IenDzi
{quote} Hi, MoneyZilla, I got interested in it. So coded a program to get these stats. All stats are with real tick data from dukascopy broker. I did tests in c#. I wanted to do the stats with that algo: On every new tick open buy and sell orders with the same sl and tp. But I didn't do that because it gets too long to calculate and too much data and can't export to excel. So to speed up calculations I ended with that algo: On every new minute, I open buy and sell orders with the same SL and TP levels. So with the same SL and TP targets (10p and...
a "low" risk reward target like 1:3 its actually a negative proposition because its harder to hit the target over time

and i think now you understand it =)
  • #29
  • Jul 24, 2017 12:20pm Jul 24, 2017 12:20pm
Quoting havo
{quote} a "low" risk reward target like 1:3 its actually a negative proposition because its harder to hit the target over time and i think now you understand it =)
What I understand, that risk reward without win% is nothing. I like to think in expectancy per trade. And for my previously written algo, 1:3 risk : reward has 25% win %, and 1:1 risk:reward has 50% win%. So expectancy per trade is the same. The different is the drawdown.
I don't know how MoneyZilla get these cubic statistics. But he said "The chance of entering anywhere on the chart and hitting 70 pips tp vs 10 pips sl is around 1%.". For example tp70 / sl10 should has 12.50% / 87.50% probability. In reality, on the chart, the probability is 1.00% / 99.00%.
When I backtested, I got 12.5% probability.
  • #30
  • Jul 24, 2017 4:21pm Jul 24, 2017 4:21pm
Quoting IenDzi
{quote} What I understand, that risk reward without win% is nothing. I like to think in expectancy per trade. And for my previously written algo, 1:3 risk : reward has 25% win %, and 1:1 risk:reward has 50% win%. So expectancy per trade is the same. The different is the drawdown. I don't know how MoneyZilla get these cubic statistics. But he said "The chance of entering anywhere on the chart and hitting 70 pips tp vs 10 pips sl is around 1%.". For example tp70 / sl10 should has 12.50% / 87.50% probability. In reality, on the chart, the...

1.- "risk reward without win% is nothing" fixed risk reward its all theory in paper.. in practice there is no consistency because the higher the reward vs the risk the harder its to hit; of course you can be lucky sometimes or have a wining streak but its not the standar !!

2.- Its all about the entry vs exit point BOTH are very important; understand that the entry its about "expected" Direction (higher than "random" or "50/50" chance of going up or down) from a given point in time AND the exit its about % of LOCKING UP an x amount of pips/points ( the higher the amount the harder to hit)

REMEMBER in the end, trading its about an entry point and securing the profit; YOU NEVER KNOW HOW MUCH THE MARKET ITS GOING TO MOVE FOR YOU OR AGAINST YOU PERIOD!!!

the answer its right in your nose.. keep searching you are almost there
  • #31
  • Jul 25, 2017 2:12am Jul 25, 2017 2:12am
The Fallacy of Using SL/TP as Proxy Risk-Reward

Forex trading forums are full of well meaning, yet rather misguided advice about risk-reward setups and how to set your stop losses. Unfortunately, many of these people fail to understand the true meaning of risk or reward.
The idea that simply setting your stop loss smaller than your take profits will achieve a certain risk reward is complete nonsense.
Using risk/reward to set your trade entry and exits does not make any sense unless you know the probability of outcomes in a given trade.
Take this simple example. Suppose there is a lottery costing $1 to enter. The prize is $1m. By the definition of the nave trader, this gives:
Risk: $1
Reward: $1m
Reward/Risk ratio: 1,000,000
By that definition, this would seem a fantastic game to play. However, suppose we know that two million people enter the lottery. This makes the odds of winning 1:2,000,000 (one in two million). Now we know the odds, we can calculate the true risk reward:
True risk: p(loss) x E(loss) = (1- 1/2000000) x ($1)
True reward: p(win) x E(win) = (1/2000000) x ($1,000,000)
True reward/risk ratio: 0.5
In other words, for every $1 you put into this lottery, youd expect to get 50 cents back. Most would now agree this is not a very good game. Even though on the nave traders reckoning, it had a reward to risk ratio of one million.
This example highlights the fallacy of using stops and take profits as a measure of your risk/reward.
In a trade, we have the real risk/reward defined by:
Reward: p(win) x E(win)
Risk: p(lose) x E(lose)
Reward/Risk ratio: p(win) x E(win) / p(lose) x E(lose)

E(win) is the expected payoff in the trade, namely your take profit amount. E(lose) is your stop loss amount.
  • #32
  • Jul 27, 2017 12:30am Jul 27, 2017 12:30am
Quoting havo
{quote} 1.- "risk reward without win% is nothing" fixed risk reward its all theory in paper.. in practice there is no consistency because the higher the reward vs the risk the harder its to hit; of course you can be lucky sometimes or have a wining streak but its not the standar !! 2.- Its all about the entry vs exit point BOTH are very important; understand that the entry its about "expected" Direction (higher than "random" or "50/50" chance of going up or down) from a given point in time AND the exit its about % of LOCKING UP an x amount of pips/points...
Do you increase lot size in some trades, or are your trades always with the same lot size(risk)? I try to find an edge in the markets with the same lot and get positive results.Maybe it is my problem.
  • #33
  • Jul 27, 2017 5:49am Jul 27, 2017 5:49am
Quoting IenDzi
{quote}
What I understand, that risk reward without win% is nothing.
I don't know how MoneyZilla get these cubic statistics. But he said "The chance of entering anywhere on the chart and hitting 70 pips tp vs 10 pips sl is around 1%.". For example tp70 / sl10 should has 12.50% / 87.50% probability. In reality, on the chart, the probability is 1.00% / 99.00%. When I backtested, I got 12.5% probability.
There is another problem that is time constraint with this rule. when or how long could this target (70 pips) be hit? when or how long could this stop loss (10 pips) be hit?

Turn around, if tp 10 / sl 70 rule, what is probability? (may be more interested by traders). Can it still be in cubic statistics? and how long could tp 10 be reached? in few bars, within a day session?
  • #34
  • Jul 27, 2017 5:56am Jul 27, 2017 5:56am
Quoting MoneyZilla
{quote} Every chart explicitly shows what is allowed. I am measuring what is allowed only. Cannot be specific, as I prefer to keep this for myself only, but if you can trust me a little? They are limiting the RR ratios. Like manipulating them is more properly said... in their one and only favor. For example tp70 / sl10 should has 12.50% / 87.50% probability. In reality, on the chart, the probability is 1.00% / 99.00%. The chance of entering anywhere on the chart and hitting 70 pips tp vs 10 pips sl is around 1%. The change the sl10 to get hit is...
If I show you plenty of examples with 10p sl and 70p tp or more,will you quit trading?
  • #35
  • Jul 27, 2017 6:11am Jul 27, 2017 6:11am
Quoting MoneyZilla
{quote} For example tp70 / sl10 should has 12.50% / 87.50% probability. In reality, on the chart, the probability is 1.00% / 99.00%. The chance of entering anywhere on the chart and hitting 70 pips tp vs 10 pips sl is around 1%. The change the sl10 to get hit is 99%. Do this 1,000 times on a pair and you will see what I mean. There is a genius mind behind this cubic system. Whoever figured it out, long time ago, when there were no computers, is an absolute genius for me...
So, if this was true, that would mean that, by reversing this 'system' and entering anywhere on the chart with a TP 10 and a SL 70, we would hit 99% of the times ? That would be an easy HG system
Don't give up on your dreams. Keep sleeping.
  • #36
  • Jul 27, 2017 2:43pm Jul 27, 2017 2:43pm
Quoting timos
{quote} If I show you plenty of examples with 10p sl and 70p tp or more,will you quit trading?
Never. There are less SLs around the curves. I am trying to catch those. Just warning those who hope for the best, while thinking rr7 (or any rr) is just a wonderful idea...
Maadik Hugiis. IQ 69.
  • #37
  • Jul 27, 2017 2:50pm Jul 27, 2017 2:50pm
Quoting Artcool
{quote} So, if this was true, that would mean that, by reversing this 'system' and entering anywhere on the chart with a TP 10 and a SL 70, we would hit 99% of the times ? That would be an easy HG system
Yes, you are quite right about that. The problem comes from knowing the direction, cause a simple reverse, simply does not work, too. Timing is crucial. When they will hold it? For how long, exactly? When is the next moment of movement?

It looks like chaos, but it is not a chaos. Not at all. The rate movement is based on obvious, ingeniously smart and intelligent mathematical system and at the end (might be some day, week, month or a year), everything comes down to zero... Catch that timing and what you have suggested will work exactly as expected, but it is very unlikely... to catch the timing.
Maadik Hugiis. IQ 69.
  • #38
  • Jul 29, 2017 8:29am Jul 29, 2017 8:29am
Quoting Jack168
{quote} There is another problem that is time constraint with this rule. when or how long could this target (70 pips) be hit? when or how long could this stop loss (10 pips) be hit? Turn around, if tp 10 / sl 70 rule, what is probability? (may be more interested by traders). Can it still be in cubic statistics? and how long could tp 10 be reached? in few bars, within a day session?
Thanks for the tip. Yes, I got cubic stats. I add [0-20) minute time for tp 10 / sl 70 and I got 99%. The longer the time, the percentage gets smaller and when you take [0,infinity) time, then it gets 12.5, 87.5%. But I don't know how to use these stats, because without knowing the direction it is useless.
  • #39
  • Jul 30, 2017 12:58am Jul 30, 2017 12:58am
Quoting IenDzi
{quote} Thanks for the tip. Yes, I got cubic stats. I add [0-20) minute time for tp 10 / sl 70 and I got 99%. The longer the time, the percentage gets smaller and when you take [0,infinity) time, then it gets 12.5, 87.5%. But I don't know how to use these stats, because without knowing the direction it is useless.
You are comming dangerously close... They are telling the direction, btw. Not as straigh as just pointing it out, but with few flips and you get well positioned...

I saw your posts on the 12.50/87.50 ratio, but would not comment them. Everything I say is exactly as I say it. It is fact based. Your bt results does not contradict to my data... You are looking them differently and therefore cant reach to the 1/99 ratio, which is their final ratio, actually. Few more steps are to be made, to get there. If you can make them - good for you!
Maadik Hugiis. IQ 69.
  • #40
  • Edited 5:52am Jul 30, 2017 5:40am | Edited 5:52am
Quoting illiquid
The Fallacy of Using SL/TP as Proxy Risk-Reward
Good link. I reached the same conclusion some years ago, as follows:

First, ASSUME that (everything else being equal), price has an equal probability of rising 1 tick, or falling 1 tick. Then you can simulate this by flipping a coin, where heads represents a 1 tick rise, or a win on a buy trade; and tails a 1 tick fall.

Now, suppose you set a TP of 1 tick, and SL of 1 tick (a 1:1 trade). Then, to win, all you need to do is flip one head, for which the probability is 50%. Flip tails first, and price hits your SL (you lose).

Now, suppose you set a TP of 2 ticks, and a SL of 1 ticks (a 2:1 trade). Then it may take several flips before price reaches either the SL or TP, but the TP is always reached immediately the number of heads flipped exceeds the number of tails by 2; while the SL is reached immediately the number of tails exceeds the number of heads by only 1. No doubt there's a math formula that calculates the probability, but I solved it by coding a Monte Carlo simulation, and running it over several million trials; the probability of the TP being reached first is 33%, or 1 in 3.

Then if you repeat the process with a TP of 3 ticks, and SL 1 tick (a 3:1 trade), the TP is always reached immediately the number of heads flipped exceeds the number of tails by 3; while the SL is reached immediately the number of tails exceeds the number of heads by only 1. The MC simulation gave a probability of 25%, or 1 in 4.

And so on.

Hence the expectancy (or profit factor) at 1:1 is: (50% x 1) / (50% x 1) = 1.
At 2:1 it is (33% x 2) / (67% x 1) = 1.
At 3:1 it is (25% x 3) / (75% x 1) = 1.
At 4:1 it is (20% x 4) / (80% x 1) = 1.
And so on.

Conclusion: varying the RR of your trades has no effect on overall expectancy.

Some provisos:

1. This does not take transaction costs into account. Hence, in reality, expectancy is less than 1, and the system will lose.

2. Back to the original assumption: that price has an equal probability of rising 1 tick, or falling 1 tick. That is not necessarily true in all situations, creating the possibility of bias. For example, to whatever extent once price starts to rise, there is a greater than 50/50 probability of the next move also being an uptick, then there is a small edge in a 'trend following' approach (taking higher than 1:1 trades). Or vice versa, in situations where there is higher probability of an uptick being followed by a downtick -- or upmove followed by a downmove -- then a 'mean reversion' strategy will work (TP<SL). In other words, only by finding situations of bias, can you gain an edge, and possibly even one that's big enough to overcome br0ker costs. But everything else being equal, simply increasing your RR will have negligible long term effect.
0 traders viewing now