is there a way to calculate the probability of a market top?

[trader198]

you need a timeframe.otherwise it ismeaningless.
for example, from today to end of May.

if no price roofs off the recent high, that is top.

if you frame that way, you can easily caculate the probability.

you can pull out data from the market inception,examine each year.

April/May is notiriously for decline, temp top's odd is very high.

 

[meanVelocity]

??

If you could predict a top with a high degree of accuracy, then predicting a monthly or a yearly top would be no more difficult than predicting an intraday top, all you need to do is change the time frame accordingly.

Different timeframes have different statistical behaviors. The most highly consistent timeframe for predictability on the S&P is the weekly, and also the daily if combined with the weekly. Monthly timeframes and quarterly have much different behavior. Also, long term forecasting kind of got ended in 2009 as we have entered a world of constant fed/central bank intervention.

There is also the basic nature of nonlinear systems that are subject to butterfly effects. One person placing a order for 5,000 cars that shifts that market a few ticks through the day will cause the daily pattern to be completely different a few months away. Nonlinear systems like the stock market are subject to the butterfly effect to a high degree. One person like Shiff or Jim Rodgers going on CNBC and stating their opinion also induces a butterfly effect which completely changes the daily pattern's the market would have formed a couple months distant. The problem with some automated strategies that have been developed is if traded with too many contracts(in futures)... The trading activity changes the market pattern and renders the strategy ineffective. Heck, even 10 Emini's following a automated strategy can have enough of a butterfly effect to change very short term patterns an hour later.

We can only predict the weather accurately for three days out. Once you get past two weeks weather models all go to the butterfly effect. One variable being changed even slightly changes the entire forecast. Markets are pretty similar... Predictability horizon declines in a linear fashion with time.

It's a little more complicated in the markets though. Certain things are easy to predict while others are very hard. For example... Predicting that prices in the S&P six to eight months from now are still going to return to touch 1825 is a easy prediction. While... In the meantime price could rocket upto 2,000 then return to this area, or rocket downward to 1700 then return to this area. But, the pattern that prices will take a month from now until then on the daily/weekly charts? Completely unpredictable.

Predictability horizon for weekly is best under three weeks, and goes unpredictable beyond five generally. Same for the daily... It's much easier to trade the market by predicting it's behavior(trending, ranging, cycling), then taking trading signals that work provided that behavior prediction is correct. Market behaviors generally steadily change or are suddenly shifted by a emotional catalyst (Like that last jobs report) and can stay consistent for months.

This current market acts so different than the market of 2009. For example, the amount of momentum it needs to maintain to have a low probability of breakdown is MUCH lower.

 

[meanVelocity]

you need a timeframe.otherwise it ismeaningless.
for example, from today to end of May.

if no price roofs off the recent high, that is top.

if you frame that way, you can easily caculate the probability.

you can pull out data from the market inception,examine each year.

April/May is notiriously for decline, temp top's odd is very high.

Seasonality effects are valid. The boost of sunshine in spring increases peoples vitamin D levels resulting in a improved mood... Heightened optimism! Warmer temperatures also help relax people and reduce tension. This pushes demand up in spring, then it declines from there...

Many different reasons seasonality effects can be quantified and proven as a physiological factor. On the west coast there is a massive amount of "winter blues" during the rainy season from fall through winter. Depression is rampant and optimism goes in the toilet until spring. If there were more west coast traders the sell in May effect would likely be much stronger!

However, seasonality effects are just a slight push. If markets are trending strongly... Seasonality gets neutralized. It's still there, it just results in a slower/faster uptrend or downtrend or slightly higher or lower prices than would occur at a different time of year.

Seasonality effects end at temperatures and sunlight in my opinion... Moon cycles, planetary, and the other crap has no connection that is quantifiable and provable(yet).

Nothing like a vacation in Florida or Hawaii to relax and lower the cortisol. (Stress hormone) I could see behavioral finance and trading psychologists going overboard and requiring fund managers to have hormone boosting shots during high market volatility. Crazier things have occurred!

 

[xelite777]

Different timeframes have different statistical behaviors.

Really?

Do you have any serious statistical study to validate this claim?

We can only predict the weather accurately for three days out.

True (at least for now), but financial markets have nothing to do with the weather. People (traders) move the market, while on the other hand the weather is controlled by physical laws.

 

[meanVelocity]

Really?

Do you have any serious statistical study to validate this claim?




True (at least for now), but financial markets have nothing to do with the weather. People (traders) move the market, while on the other hand the weather is controlled by physical laws.

Weather and financial markets are both non-linear systems that form similar cycles and patterns. If you look at daily average temperature and highs and lows for a location over a year... It's very similar to stock patterns. Even statistically. High mean return bias when temperatures extend too high/low from norms. Long steady trends with cycles.

You can generally assume that when a strong heatwave occurs that temperatures will moderate after a amount of time. The same in the markets... If there is a very strong rally, probability that it will moderate and consolidate is very high after an amount of time. The time both phenomena take to moderate is actually similar... LoL!

But anyway, going back to the statistical difference between patterns on 5m, 15m, daily, weekly, monthly... Huge behavioral differences which can easily be proven statistically.

I recently just statistically proved the superiority of Tick(Trade) charts over time charts. I was actually kind of surprised... I knew they were a bit better, but when I looked at the stability of their behaviors with straight statistics it was a little jaw dropping...

 

[meanVelocity]

Ill give you tip on one of the biggest differences between monthly and lower timeframes... Sustainable trend strength on monthly is half of what lower timeframes like the 15m can sustain. Trend strength defined as velocity divided into volatility (ATR).

I'm talking about the DOW/S&P. Commodities behave very differently. Forex is also kind of unique. Similar behavior to commodities, but not quite.

 

[xelite777]

Ill give you tip on one of the biggest differences between monthly and lower timeframes... Sustainable trend strength on monthly is half of what lower timeframes like the 15m can sustain. Trend strength defined as velocity divided into volatility (ATR).

Interesting.

But how do you define "velocity" in mathematical terms?

Thanks.

I'm talking about the DOW/S&P. Commodities behave very differently.

True, it seems that the equity market (stock and indexes) is essentially mean reverting, while commodities can trend for a very long time.

 

[meanVelocity]

Just average the rate of movement over a number of days.

If a market has risen for an average of five points per day over five days, that's the velocity. Divide it into the volatility (ATR over those days, that is recorded trend strength).

That's how I do it. Well, I have my own proprietary formulas and smoothing techniques, but you get the idea.

Interesting.

But how do you define "velocity" in mathematical terms?

Thanks.



True, it seems that the equity market (stock and indexes) is essentially mean reverting, while commodities can trend for a very long time.

 

[meanVelocity]

Ok, since you all are so interested in probability... I will give you a nibble.

Right now I estimate we have around a 75% probability of closing this week below 1899, but we also have around a 60% probability of touching prices above 1899.

Not Shown: Also, we have a 50% probability of touching 1858 next week. Probability of touching 1878 was a minimum of 75% this week... Probability of touching 0.5ATR above this weekly mean has kinda maintained at 70%+ since early 2013. LoL...

70% probability doesn't sound like much unless you know how to do compound statistics. A 70% OR probability compounds to a 97.3% probability over three bars. So, this market has been maintaining a 97.3% probability of touching prices 0.5ATR over what is essentially a 6SMA over each three weeks.

Ok, so... Let me just state some facts simpler. This market has been maintaining a 90%+ probability of deep sixing any shorts done 0.5ATR below average... Within three weeks... By 1ATR+. Probability of price staying below -.5ATR week to week has maintained around 10%.

So, if we are -0.5 ATR below average, we have a 90%+ probability of going 0.5 ATR higher than average. Over two weeks, that compounds to a 99% probability.

The OP's question is very valid. Yes, there are ways to judge the probability of a bottom with some accuracy. But, market behaviors do eventually change. Markets pre-2013 were heavily cyclic. This market is currently behaving normally. Just like the 2007 bull market, or the .com bubble runup.


There are more factors at play if you want to get close to calculating current true probability... Number of bars up/down, wave counts, momentum, and momentum divergence. My probability estimates are 90% mechanical, and 10% subjective. I do watch news closely as certain events like fed policy changes can cause probability to be shifted a bit different than what the numbers say.

Note: This chart is a bit stripped down... My normal chart setups and code I keep secret nowadays. xelite777, you can PM me if you want a glance. Will probably put you into a state of belief shock... Jk, lol!

 

[athlonmank8]

Different timeframes have different statistical behaviors. The most highly consistent timeframe for predictability on the S&P is the weekly, and also the daily if combined with the weekly. Monthly timeframes and quarterly have much different behavior. Also, long term forecasting kind of got ended in 2009 as we have entered a world of constant fed/central bank intervention.

There is also the basic nature of nonlinear systems that are subject to butterfly effects. One person placing a order for 5,000 cars that shifts that market a few ticks through the day will cause the daily pattern to be completely different a few months away. Nonlinear systems like the stock market are subject to the butterfly effect to a high degree. One person like Shiff or Jim Rodgers going on CNBC and stating their opinion also induces a butterfly effect which completely changes the daily pattern's the market would have formed a couple months distant. The problem with some automated strategies that have been developed is if traded with too many contracts(in futures)... The trading activity changes the market pattern and renders the strategy ineffective. Heck, even 10 Emini's following a automated strategy can have enough of a butterfly effect to change very short term patterns an hour later.

We can only predict the weather accurately for three days out. Once you get past two weeks weather models all go to the butterfly effect. One variable being changed even slightly changes the entire forecast. Markets are pretty similar... Predictability horizon declines in a linear fashion with time.

It's a little more complicated in the markets though. Certain things are easy to predict while others are very hard. For example... Predicting that prices in the S&P six to eight months from now are still going to return to touch 1825 is a easy prediction. While... In the meantime price could rocket upto 2,000 then return to this area, or rocket downward to 1700 then return to this area. But, the pattern that prices will take a month from now until then on the daily/weekly charts? Completely unpredictable.

Predictability horizon for weekly is best under three weeks, and goes unpredictable beyond five generally. Same for the daily... It's much easier to trade the market by predicting it's behavior(trending, ranging, cycling), then taking trading signals that work provided that behavior prediction is correct. Market behaviors generally steadily change or are suddenly shifted by a emotional catalyst (Like that last jobs report) and can stay consistent for months.

This current market acts so different than the market of 2009. For example, the amount of momentum it needs to maintain to have a low probability of breakdown is MUCH lower.

Not to disagree, but there's been traders here who have been able to predict turning points months out with high degrees of accuracy. One of the best examples.......

http://www.elitetrader.com/vb/showthread.php?p=2160361#post2160361
http://www.elitetrader.com/vb/showthread.php?p=2337356#post2337356

As for linear cycles there's plenty of people who agree with you there. Hurst is one. Even Miner.

Ironically, he also trades moon cycles