International markets traders banter

[Quant]

This last week is the first instance of bearish range expansion on rising volatility since the Nov election , also noted the treasury to junk bond spread has widened also . Really is the first significant risk of to be seen since the Orange oracle took power

 

[kid hustlr]

Rate my trade - interested in thoughts and try to generate some discussion.

I recognise it's in a vacuum.

One other point to note is that i did pick up anything 'obvious' watching the tape. Price moved but it was orderly, didn't feel like a huge push. I would describe the volume as still above average thouh

 

[kid hustlr]

I would be REALLY interested in hearing from others who have done some statistical work on their respective equity indices if they trade futures. Looking at some FTSE data and I'm kind of blown away with what I'm seeing. Note this is cash hours only, 6 years of data.

- average range 75 points, slightly higher than I expected but there is a long tail in there. 72% of days have a range of between 33 and 95 points

- yesterday's high is broken 53% of the time
- when the high is broken the average new high is 38 points (skewed by several big moves). 64% of the time the new high is 40 points or less but there were far less false break's (say 5 points or less) then i was expecting. Overall I'm surprised how bullish it is when we break y/d high.

- yesterday's low is broken 46% of the time
- when the low is broken the average new low is 47 points (skewed by several big moves). 55% of the time the new low is 40 points or less but there were far less false break's (say 5 points or less) then i was expecting. Overall I'm surprised how bearish it is when we break y/d low.

The other thing I did do was the break data into up trending vs not uptrending. Now this was just looking at the run up between 2012 and 2015 vs the rest and i was shocked that the figures/statistics above didn't really change. Perhaps this is because it was a pretty choppy run up in the FTSE or perhaps regardless of the bull/bear market the figures quoted above remain the same.

Will continue to research but woudl enjoy thoughts on other mkt's and experiences.

 

[kid hustlr]

I have a statistics question.

If i flip a coin it's 50/50 whether its heads or tails. The chances of flipping 2 heads in a row is 25%.

Noting the above, if i flip a head the first time, the chances of a head the second time is still 50/50 as each is independent.

Now on the FTSE, i have data which tells me based on 1000 days, (fake data btw):

# of times last nights high is broken (1 day in a row) = 500
# of times the next nights high is broken (2 days in a row) = 200
# of times the next nights high is broken (3 days in a row) = 50
# of times the next nights high is broken (4 days in a row) = 10

If we had 3 new highs in a row, what's the chance of a high on the 4th day?

 

[skc]

I have a statistics question.

If i flip a coin it's 50/50 whether its heads or tails. The chances of flipping 2 heads in a row is 25%.

Noting the above, if i flip a head the first time, the chances of a head the second time is still 50/50 as each is independent.

Now on the FTSE, i have data which tells me based on 1000 days, (fake data btw):

# of times last nights high is broken (1 day in a row) = 500
# of times the next nights high is broken (2 days in a row) = 200
# of times the next nights high is broken (3 days in a row) = 50
# of times the next nights high is broken (4 days in a row) = 10

If we had 3 new highs in a row, what's the chance of a high on the 4th day?

20%?

 

[kid hustlr]

20%?

That's what I'm thinking as well - in the case of the market they aren't independent events? (I Think?)

Edit:

Based on my data there might be something there for me which will help my trading, but I want to make sure my understanding of the maths is correct

 

[OmegaTrader]

I have a statistics question.

If i flip a coin it's 50/50 whether its heads or tails. The chances of flipping 2 heads in a row is 25%.

Noting the above, if i flip a head the first time, the chances of a head the second time is still 50/50 as each is independent.

Now on the FTSE, i have data which tells me based on 1000 days, (fake data btw):

# of times last nights high is broken (1 day in a row) = 500
# of times the next nights high is broken (2 days in a row) = 200
# of times the next nights high is broken (3 days in a row) = 50
# of times the next nights high is broken (4 days in a row) = 10

If we had 3 new highs in a row, what's the chance of a high on the 4th day?

1) Assumption, there is no pattern and it is randomness, independence 500/1000 =50%

2) Assumption 3 already dependence given your stats 10/50= 20%

The assumption is the hard part...

Are you hoping that 80% of the time the trend will die out or...

 

[Quant]

I would be REALLY interested in hearing from others who have done some statistical work on their respective equity indices if they trade futures. Looking at some FTSE data and I'm kind of blown away with what I'm seeing. Note this is cash hours only, 6 years of data.

- average range 75 points, slightly higher than I expected but there is a long tail in there. 72% of days have a range of between 33 and 95 points

- yesterday's high is broken 53% of the time
- when the high is broken the average new high is 38 points (skewed by several big moves). 64% of the time the new high is 40 points or less but there were far less false break's (say 5 points or less) then i was expecting. Overall I'm surprised how bullish it is when we break y/d high.

- yesterday's low is broken 46% of the time
- when the low is broken the average new low is 47 points (skewed by several big moves). 55% of the time the new low is 40 points or less but there were far less false break's (say 5 points or less) then i was expecting. Overall I'm surprised how bearish it is when we break y/d low.

The other thing I did do was the break data into up trending vs not uptrending. Now this was just looking at the run up between 2012 and 2015 vs the rest and i was shocked that the figures/statistics above didn't really change. Perhaps this is because it was a pretty choppy run up in the FTSE or perhaps regardless of the bull/bear market the figures quoted above remain the same.

Will continue to research but woudl enjoy thoughts on other mkt's and experiences.

My experience with fading X consec green days is there is no edge BUT i do see an edge fading X RED days and the best edges come at extremes , i do believe taking context into it will refine the edge ( depth of move and position in range ) , when i got my net back ill share some data , Brisbane here and i still have no wired net and using phone hotspot and rationing data atm ..

 

[CanOz]

Did you lose Internet in the storm quant?

 

[kid hustlr]

20%?

Just to finish this discussion.

Looking at my data again there's nothing too crazy there. In effect there is no real tell that after x days we are likely to have y happen. It's pretty much 50/50 from one day to the next.

5 up days in a row to 6 up days in a row is 16/32 = 50% for example.

 

[skc]

Just to finish this discussion.

Looking at my data again there's nothing too crazy there. In effect there is no real tell that after x days we are likely to have y happen. It's pretty much 50/50 from one day to the next.

5 up days in a row to 6 up days in a row is 16/32 = 50% for example.View attachment 70611

These numbers look a bit too neat? It's so close to 50% even with the small sample size... probably worth a quick check of the excel formula used...

That's what I'm thinking as well - in the case of the market they aren't independent events? (I Think?)

Remember the difference between probability and statistics. In a coin there's a known probability and the statistics over the long term will be the same as the probability. In social science, there is only statistics. We might imply probability from statistics but we could easily be wrong when something changes fundamentally. I think that's how lots of quant funds blow up... thinking that they are trading with probability on their side.

How does that translate to your FTSE statistics? :dunno:

 

[kid hustlr]

These numbers look a bit too neat? It's so close to 50% even with the small sample size... probably worth a quick check of the excel formula used...



Remember the difference between probability and statistics. In a coin there's a known probability and the statistics over the long term will be the same as the probability. In social science, there is only statistics. We might imply probability from statistics but we could easily be wrong when something changes fundamentally. I think that's how lots of quant funds blow up... thinking that they are trading with probability on their side.

How does that translate to your FTSE statistics? :dunno:

I checked the data as I thought the same thing but I'm confident it's correct. It is very clean across the board isnt it.

There's some practical help for me here - the discussion above about highs and lows in a row doesn't give me much, but the stats I've looked at about how often y/d highs, lows (or both) are broken and the mean result when it occurs is telling.

I naturally want to fade most intraday and previous day highs and lows but working through this I'm a little more inclined to 'go with it' a fraction more when y/d highs & lows are broken

 

[OmegaTrader]

1) Is it random
2) Is it cause or an effect of an underlying process
3) Is the sample enough
4) Will the past repeat itself

60% prob looks good haha

but only 185 sample

looks like follow the high ...

Which in the long term makes sense if the market is going up long term anyway?

economy is inflating, population growth, technology growth, efficiency, profits, survivorship bias etc etc

Idk

 

[Quant]

Did you lose Internet in the storm quant?

Yeah still no wired net , Optus tech due today but not holding my breath , imagine they be very busy . My phone data fees will be through the roof .

Ok here is a chart with the consecutive days marked , pretty plain to see that fading green looks less meaningful but as i stated above fading extremes of red consec looks to have merit with quite a few marking significant multi day swing lows once the count ends , combined with a filter or 2 i can see a distinct edge potential . Once i have my internet back i will do some explorations .

The point to consider is not the result the next day but the result over the next 5-10 days

 

[Quant]

Yeah still no wired net , Optus tech due today but not holding my breath , imagine they be very busy . My phone data fees will be through the roof .

Ok here is a chart with the consecutive days marked , pretty plain to see that fading green looks less meaningful but as i stated above fading extremes of red consec looks to have merit with quite a few marking significant multi day swing lows once the count ends , combined with a filter or 2 i can see a distinct edge potential . Once i have my internet back i will do some explorations .

The point to consider is not the result the next day but the result over the next 5-10 days

OK nets back , spent about 5 minutes coding this and with no filters and zero money management the results are encouraging enough to continue on . this test is from april 2009 which seems cherry picking i know but it tests back to 1984 with positive returns , not difficult to filter out the crashes , with additional filters and risk management this has good potential i think . Curve is smooth and steady and will get significantly smoother with some risk control & additional filters , enough so as to Amp up size and get exponentially better returns.

This is using 5 and above consecutive red days as baseline , will expand the range in time , will also test the consec green days scenario ..

since 2009 2 years have outperformed 2009 fwiw with another 2 very close to 09 returns

 

[kid hustlr]

Here's one that threw me.

Chart below is 6 years of FTSE daily data.

Blue line = index
Orange line = close - open (cumulative)
Grey line = open - previous close (cumulative)

By definition Orange (during the day) + Grey (during the night) = Blue (and it does)

In mid 2013 something definitely changed. In effect the market has a HUGE sell bias during the cash hours then an EVEN BIGGER buy bias after the close.

USA dragging up the mkt out of hours?

Even more confusing is I did some study around gap fills (I defined a gap as above y/d range) and I found no clear edge or obvious tendencies (apart from the fact Monday gaps fill more often than the rest of the week).

I feel I need to go back and look at this more closely - especially gap up days.

Any thoughts/comments??

 

[Trembling Hand]

Here's one that threw me.

In mid 2013 something definitely changed. In effect the market has a HUGE sell bias during the cash hours then an EVEN BIGGER buy bias after the close.

That's a well-tested feature of markets. You get paid for taking on risk, holding while markets are closed, and get punished while taking the 'safe' route - day trading long.

You will find that showing up in most bull markets. In all types of instruments. Punters rushing to get in on the open after realising they should have bought yesterday or last month!

http://tremblinghandtrader.typepad.com/trembling_hand_trader/2007/07/spi.html

 

[skc]

Here's one that threw me.

Chart below is 6 years of FTSE daily data.

Blue line = index
Orange line = close - open (cumulative)
Grey line = open - previous close (cumulative)

By definition Orange (during the day) + Grey (during the night) = Blue (and it does)

You should also plot daily Grey minus Orange. That way you will pick up any change in this observation quicker than a cumulative plot (and month tickers on the X-axis just so one can see the dates)..

You will find that showing up in most bull markets. In all types of instruments. Punters rushing to get in on the open after realising they should have bought yesterday or last month!

http://tremblinghandtrader.typepad.com/trembling_hand_trader/2007/07/spi.html

I was going to point kid to this blogpost but just couldn't find it...

 

[kid hustlr]

Here I am doing work when all I had to do was read the blog.

Thanks for making me feel inadequate all

 

[kid hustlr]

What I do find is that gap fills aren't that profitable. I need to do more work here but you can't just sell the open when we gap up and profit, there's more to it than that