Trading What I see !?

[ShaunWhite]

Edit the last post Shaun answered it .

You can teach tech to anyone prepared to put in the work, but thinking of ways to dig into your results is the pay dirt. 90% thinking 10% doing. I've always sorted my pl by things other than date because date doesn't seem especially interesting unless there's a cycle. Picturing the races as weirdly happening in something like total volume sequence, or price gap to the next nearest in the betting is more interesting. Don't sort by what shade of brown the horse was of course, it's got to be things that have some logical connection to market.

[ShaunWhite]

I'd like to make a point here.

People have this false impression that there is something special going on, if only they knew that one thing it would transform their trading. But it's not one thing, its an amalgam of many.

I have traded live in front of people many times, but nobody really learns anything. It all happens too fast. The trading process is just much deeper than that. What do you in those final moments, is the product of what you are doing beforehand. So teaching people the components allows them to come up with the final result.

I've actually had the privilege of trying many things and examining all areas. I've even taught my children.

When I used to run courses I would try all manner of different things to try and get people to head in the right direction. But it often felt the harder you tried the harder it became. You need to start at the basics and create some building blocks and work out from there.

I remember sitting two people down, telling them what to do and trading live in front of them. One ended in a profit, the other a loss and at that point I realise not only was trading the issue, but psychology as well.

You can't guarantee, even if you give people the knowledge they need that they can use it.

To give you a case in point I posted two really good YouTube videos both of which resulted in some fab results the week after I posted them. Neither was viewed or liked much at all in comparison to other videos.

I think this thread has fractured. A few of us are trying to help goat with his analysis/automation and others have started talking about live trading videos for some reason.

[goat68]

Is the red line the pnl at each value? Can you plot cumulative pnl instead?

ok i'm been dumb!, what do you mean by "cumulative" ?!
What I did was put bets into "buckets", so every bet that was 0.94-0.949999 was the red data point on the graph for 0.94, etc ...

Cummulative is just a running total. Eg you usually sort your results by date/time when you want a pl chart for the month, but you can sort your results by anything you want, price, price gap, time till it was matched, overall vol or anything your creativity can come up with. As though the races happened in some other sequence other than date/time. Then look for upward slopes and see what then corresponded to on the x-axis.

For small samples you may want to group the x-axis into chunks but one step at a time for niem and if the sample is reasonable some sort of pattern should visible.

It would be a useful exercise to take all of your bet history (pl) for a strategy and sort it by as many things as you can think of. I thought you were doing that and that's why I'm always suprised you do the analysis so quickly. It should take days even just to come up with a list of anything and anything you can break it down by (the creativity part) ....... But once you've done that you'll have a reusable spreadsheet you can put any results into to analyse. Again, I suggested you did that work while your data was accumulating so you'd be ready to churn ideas and analyse them easily. I'm pretty sure you didn't do that, maybe I just didn't explain it very well.

What I produced was cummulative PnL by Vwap gap. I achieved this by throwing results into 0.01 vwap gap buckets as I went along.
I think we're all agreeing with different words about the same thing which is the chart above.

By the way I do sort my backtests in many different ways. The way I do this is I wrote an analyzing program that takes multiple params of what I want to sort or filter by, and it pumps out the answer from the raw result...

[goat68]

I'd like to make a point here.

People have this false impression that there is something special going on, if only they knew that one thing it would transform their trading. But it's not one thing, its an amalgam of many.

I have traded live in front of people many times, but nobody really learns anything. It all happens too fast. The trading process is just much deeper than that. What do you in those final moments, is the product of what you are doing beforehand. So teaching people the components allows them to come up with the final result.

I've actually had the privilege of trying many things and examining all areas. I've even taught my children.

When I used to run courses I would try all manner of different things to try and get people to head in the right direction. But it often felt the harder you tried the harder it became. You need to start at the basics and create some building blocks and work out from there.

I remember sitting two people down, telling them what to do and trading live in front of them. One ended in a profit, the other a loss and at that point I realise not only was trading the issue, but psychology as well.

You can't guarantee, even if you give people the knowledge they need that they can use it.

To give you a case in point I posted two really good YouTube videos both of which resulted in some fab results the week after I posted them. Neither was viewed or liked much at all in comparison to other videos.

I think this thread has fractured. A few of us are trying to help goat with his analysis/automation and others have started talking about live trading videos for some reason.

agree, i'm not interested in live videos, although I do watch all yours Peter

[Euler]

I'll butt out of the thread!

[goat68]

But it often felt the harder you tried the harder it became.

Yeah!
I'm just wondering who's going to go mad first me or Shaun (I think I am leading!)

[arch4672]

ok i'm been dumb!, what do you mean by "cumulative" ?!
What I did was put bets into "buckets", so every bet that was 0.94-0.949999 was the red data point on the graph for 0.94, etc ...

Cummulative is just a running total. Eg you usually sort your results by date/time when you want a pl chart for the month, but you can sort your results by anything you want, price, price gap, time till it was matched, overall vol or anything your creativity can come up with. As though the races happened in some other sequence other than date/time. Then look for upward slopes and see what then corresponded to on the x-axis.

For small samples you may want to group the x-axis into chunks but one step at a time for niem and if the sample is reasonable some sort of pattern should visible.

It would be a useful exercise to take all of your bet history (pl) for a strategy and sort it by as many things as you can think of. I thought you were doing that and that's why I'm always suprised you do the analysis so quickly. It should take days even just to come up with a list of anything and anything you can break it down by (the creativity part) ....... But once you've done that you'll have a reusable spreadsheet you can put any results into to analyse. Again, I suggested you did that work while your data was accumulating so you'd be ready to churn ideas and analyse them easily. I'm pretty sure you didn't do that, maybe I just didn't explain it very well.

What I produced was cummulative PnL by Vwap gap. I achieved this by throwing results into 0.01 vwap gap buckets as I went along.
I think we're all agreeing with different words about the same thing which is the chart above.

By the way I do sort my backtests in many different ways. The way I do this is I wrote an analyzing program that takes multiple params of what I want to sort or filter by, and it pumps out the answer from the raw result...

I don't think that you are producing a cumulative pnl line. The point at the end of the line on the right hand side would be the total pnl of all your bets, which on your plot is zero. I find it hard to believe that after 27000 bets you'd come out exactly break even.

[goat68]

Cummulative is just a running total. Eg you usually sort your results by date/time when you want a pl chart for the month, but you can sort your results by anything you want, price, price gap, time till it was matched, overall vol or anything your creativity can come up with. As though the races happened in some other sequence other than date/time. Then look for upward slopes and see what then corresponded to on the x-axis.

For small samples you may want to group the x-axis into chunks but one step at a time for niem and if the sample is reasonable some sort of pattern should visible.

It would be a useful exercise to take all of your bet history (pl) for a strategy and sort it by as many things as you can think of. I thought you were doing that and that's why I'm always suprised you do the analysis so quickly. It should take days even just to come up with a list of anything and anything you can break it down by (the creativity part) ....... But once you've done that you'll have a reusable spreadsheet you can put any results into to analyse. Again, I suggested you did that work while your data was accumulating so you'd be ready to churn ideas and analyse them easily. I'm pretty sure you didn't do that, maybe I just didn't explain it very well.

What I produced was cummulative PnL by Vwap gap. I achieved this by throwing results into 0.01 vwap gap buckets as I went along.
I think we're all agreeing with different words about the same thing which is the chart above.

By the way I do sort my backtests in many different ways. The way I do this is I wrote an analyzing program that takes multiple params of what I want to sort or filter by, and it pumps out the answer from the raw result...

I don't think that you are producing a cumulative pnl line. The point at the end of the line on the right hand side would be the total pnl of all your bets, which on your plot is zero. I find it hard to believe that after 27000 bets you'd come out exactly break even.

ok, thank you , you've explained it in a way I now understand, You're good at this

[goat68]

Cummulative is just a running total. Eg you usually sort your results by date/time when you want a pl chart for the month, but you can sort your results by anything you want, price, price gap, time till it was matched, overall vol or anything your creativity can come up with. As though the races happened in some other sequence other than date/time. Then look for upward slopes and see what then corresponded to on the x-axis.

For small samples you may want to group the x-axis into chunks but one step at a time for niem and if the sample is reasonable some sort of pattern should visible.

It would be a useful exercise to take all of your bet history (pl) for a strategy and sort it by as many things as you can think of. I thought you were doing that and that's why I'm always suprised you do the analysis so quickly. It should take days even just to come up with a list of anything and anything you can break it down by (the creativity part) ....... But once you've done that you'll have a reusable spreadsheet you can put any results into to analyse. Again, I suggested you did that work while your data was accumulating so you'd be ready to churn ideas and analyse them easily. I'm pretty sure you didn't do that, maybe I just didn't explain it very well.

What I produced was cummulative PnL by Vwap gap. I achieved this by throwing results into 0.01 vwap gap buckets as I went along.
I think we're all agreeing with different words about the same thing which is the chart above.

By the way I do sort my backtests in many different ways. The way I do this is I wrote an analyzing program that takes multiple params of what I want to sort or filter by, and it pumps out the answer from the raw result...

I don't think that you are producing a cumulative pnl line. The point at the end of the line on the right hand side would be the total pnl of all your bets, which on your plot is zero. I find it hard to believe that after 27000 bets you'd come out exactly break even.

Here we go (again!)

Screenshot 2021-09-20 at 14.50.31.png

[goat68]

You know this conversation is making a great point about what Euler was on about, in that both "teacher" and "student" can both 100% agree they both understand each other, when in fact they're totally different !

[ShaunWhite]

Brill thanks. That's more like it.

Shame the x values are missing. Tiny tiny range of values where it goes up. Doesn't look especially statistically significant unfortunately. Do you get the same shape when you run subsets of those results?

If you'd shown me that before I could have saved you 300 quid. Feels like progress though.... Now go sort crazy.....

I'd bin the bet analysis program, there's this nifty new thing called Excel that's pretty good.

[ShaunWhite]

both "teacher" and "student" can both 100% agree they both understand each other, when in fact they're totally different !

One or both is making an assumption they understand each other. Poor teaching and/or poor studenting is the only reason for miscommunication.

..

[goat68]

I'd bin the bet analysis program, there's this nifty new thing called Excel that's pretty good.

fine stick your fingers into my efforts again......!
I'm very nifty at the cmd line and like to dial in what I want through my own programming..., I also personally don't get on with Excel/Sheets

[decomez6]

Edit the last post Shaun answered it .

You can teach tech to anyone prepared to put in the work, but thinking of ways to dig into your results is the pay dirt. 90% thinking 10% doing. I've always sorted my pl by things other than date because date doesn't seem especially interesting unless there's a cycle. Picturing the races as weirdly happening in something like total volume sequence, or price gap to the next nearest in the betting is more interesting. Don't sort by what shade of brown the horse was of course, it's got to be things that have some logical connection to market.

That’s why tech is a hobby career, giving one the ability to approach arts with a scientific method.
Logic and algorithms are hard wired into our brains but it takes effort to even know what the brain is doing .
There is also the problem of exchanging time for money . One must weigh the benefits of learning new things if you can just pay for the services and launchpad your artistic side with the help of the gifted scientist.
Regarding the data you collect it
Seems like you have excel pivot tables and excel Ai in mind ( or any other machine learning program) .it can do wonders with such kind of data . Exposing all sorts correlations.

[decomez6]

I
To give you a case in point I posted two really good YouTube videos both of which resulted in some fab results the week after I posted them. Neither was viewed or liked much at all in comparison to other videos.

Due to the wealth of information at your disposal You posses the big advantage of hindsight , almost like the sixth sense of sports trading , your muscle memory Will actually see anomalies manifesting into trading opportunities .
Things like the historical effects of weather and the likelihood of such a re occuring phenomenal , or a big horse owners helicopter landing on a racecourse, horses travelling long distances just to take part in a race …etc…will be only valuable to those who know what to do with it . So if you make a videos highlighting such occurrences it will only be regarded as a one off lucky coincidence and very few traders keep their ears that close to the ground.
That being said I wonder which videos ?
I find all your videos informative , either as a reminder or an initiator to new trading strategies.