Market Statistics - Finding a statistical Edge in your market

[fiftyeight]

Actually the ECB only reports 8 times per year on rate decisions....so the two days is mostly waiting for the NFP. There 3 occasions, including this week that the ECB reported the day before the NFP report.

I'm trying to look for trend days after they report, but not seeing much in it yet.

Are you just eyeballing an hourly chart on specific days?

 

[CanOz]

Are you just eyeballing an hourly chart on specific days?

Yup, we'll I was eyeballing a TPO / VP chart, a 5m and a 240 minute with the global cursor.... But I want to use that information to see if it's worth while going a step further, like getting a mean reversion strategy coded...then I can gather more stats.

 

[CanOz]

Seems a fitting night for one of these.....

 

[cynic]

View attachment 66917

Sins ewe knead tool urn mooring leash, mite eye sir jest ewe ewes ditch atoll know take err four trance lay shun!

Ore reef knot, reed allowed!



Rear salts forded axe egg salmon neigh shun ova too honey dread ant fore demons dart are:







Pea erse: Sum won comb planed eye ewes tomb anise ill labels.

 

[Quant]

"" Market Statistics - Finding a statistical Edge in your market ""

A thread with a purpose , watching with interest . Ultimately success in this endeavour is as close to the holy grail you will find , Mean reversion systems on indice are likely (definitely) the place to look . I believe a volatility filter is an essential part of this process ( hint hint ) .

Certain days are definitely statistically superior for highs and lows as had been discussed , certain days are statistically superior on being trending days . Intraday reversals similar traits timewise . There are a plethora of edges to be found , developing systems to exploit said edges is the hard part


Hope this thread gets traction without getting the typical derail ... carry on

In the Wise Words of Howard Bandy whom we are very lucky to have on ASF , he has the keys to finding the answers , pursue them with vigour

 

[Quant]

This podcast definitely worth a listen for those heading down this road


https://chatwithtraders.com/ep-070-jeff-davis/

 

[get better]

This podcast definitely worth a listen for those heading down this road


https://chatwithtraders.com/ep-070-jeff-davis/

Thanks for the link!

 

[fiftyeight]

I was trying to replicate some stuff I found on TH’s old blog (Re-reading it). I was looking at gaps and found this.

Has something structurally changed in the NKY that could cause the fall in the %gap or is something going on with my data?

This posed an interesting thought for me. If you do identify something while playing with stats, how do you make sure it is actually something interesting and not a change in data or some other structural change?

 

[History Repeats]

Look at time period of your chart and the index

 

[fiftyeight]

Look at time period of your chart and the index

Yeah it was range bound during this period so I had a look at the below.

The high low range although more volatile, remains relatively unchanged. Unlike %Gaps

 

[fiftyeight]

I am thinking about trading the close of the es-mini. I have some daily data so thought I would have a look at how far off the lows does the close finish on a down day.

There have been 810 down days since 27/01/2009.

About 50% of the time the market recovers 0-27% off the lows. 18% of the time it recovers 10% of the range or less. I think this probably falls in lines with people's general perception of the markets so not that fruitful but thought I would post anyway.

 

[cynic]

I am thinking about trading the close of the es-mini. I have some daily data so thought I would have a look at how far off the lows does the close finish on a down day.

There have been 810 down days since 27/01/2009.

About 50% of the time the market recovers 0-27% off the lows. 18% of the time it recovers 10% of the range or less. I think this probably falls in lines with people's general perception of the markets so not that fruitful but thought I would post anyway.

View attachment 67519

Interesting analysis, just the same.

One puzzling thought does spring to mind though. Until the market closes for the day, how do you determine that the daily low is already in place?

 

[Gringotts Bank]

Interesting analysis, just the same.

One puzzling thought does spring to mind though. Until the market closes for the day, how do you determine that the daily low is already in place?

Yeh there's look ahead bias built in. Not tradable.

The only thing you could do is use the (high-open) stats from the previous days, if today is likely to be a down day, and trade around the open. Or vice versa.

 

[cynic]

Yeh there's look ahead bias built in. Not tradable.

The only thing you could do is use the (high-open) stats from the previous days, if today is likely to be a down day, and trade around the open. Or vice versa.

Following on from your logic, instead of using EOD, one could use intraday data to obtain an interim daily low (at an earlier than close cut off time) and trade the residual period.

 

[fiftyeight]

One puzzling thought does spring to mind though. Until the market closes for the day, how do you determine that the daily low is already in place?

Yeah I did think of this, if I find anything of note I will have to investigate further.

I did find this interesting though.

After a recovery or a sell off back to the open, there is an almost 17% range where the es has not closed in 1854 trading days. (could also be a very tight range, but then I would not expect this gap)

Not sure if it my excel skills or data or something interesting yet? Further investigating required

 

[Kryzz]

I'm really surprised this thread hasn't gained a significant amount of attention. After listening/watching through the podcasts and videos in this thread, it is really interesting stuff.

From anyone's experience, is there a resource where such statistics are readily available? Or do most people generate these themselves?

I've been doing a lot more basic stuff with some major FX pairs, with an EOD approach.

As mentioned earlier in the thread, it's a not a trading system but definitely can provide a market bias. Using the example below, looking at the past 20 years of data there's a 70%+ chance that April will be a positive month for the AUDUSD.

In my mind this takes a lot of the guess work out, you would more inclined to keep your eyes peeled for long only setups.

I will be going through a lot of the major FX pairs and recording the statistics, if anyone is interested I will post up the spreadsheet accordingly.

 

[Gringotts Bank]

From anyone's experience, is there a resource where such statistics are readily available? Or do most people generate these themselves?

There's plenty of blogs around. Generally such information is not presented as a table of statistics, but rather as a complete trading system in a backtest.

Say you use AMIbroker, you can for example select/optimize buy conditions based on the month/week/day/hourofday/minuteofday in a single line of code, then match it with a corresponding set of sell parameters. Statistics are a good way to find potential edges, but a complete system requires much more. Knowing April will have a 70% of being positive is a very long way from being something you can trade, but it's possibly a useful piece of knowledge. You can accumulate such chunks of knowledge and try to put them into a tradable system.

Everyone will be interested in your stats - post 'em up.

 

[Wysiwyg]

I'm really surprised this thread hasn't gained a significant amount of attention. After listening/watching through the podcasts and videos in this thread, it is really interesting stuff.

A securities market statistician might be able to prove me wrong, but from my experience, statistics are bound to the data from which they are gathered. Extrapolating a guaranteed winning strategy from the gathered statistics is not a given.

 

[fiftyeight]

A securities market statistician might be able to prove me wrong, but from my experience, statistics are bound to the data from which they are gathered. Extrapolating a guaranteed winning strategy from the gathered statistics is not a given.

This is not why I like looking at stats at all.

The reason I use stats is because I do not have 20 years experience in any market. I have no idea of how a market tends to behave under certain conditions.

Even if I did have 20 years experience, the human memory is so fallible that it is not the best source of stats.

Simply trying to have a better understanding of how they "normally" move

 

[Kryzz]

This is not why I like looking at stats at all.

The reason I use stats is because I do not have 20 years experience in any market. I have no idea of how a market tends to behave under certain conditions.

Even if I did have 20 years experience, the human memory is so fallible that it is not the best source of stats.

Simply trying to have a better understanding of how they "normally" move

Agree with the above, I would be looking to use statistics to gain an insight/bias toward market behaviour not necessarily the basis of a trading system in its own right.